Category Archives: Environmental Health

The Association between Lifelong Greenspace Exposure and 3-Dimensional Brain Magnetic Resonance Imaging in Barcelona Schoolchildren

Author Affiliations open

1ISGlobal, Barcelona, Catalonia, Spain

2Pompeu Fabra University, Barcelona, Catalonia, Spain

3CIBER Epidemiología y Salud Pública (CIBERESP), Madrid, Spain

4MRI Research Unit, Radiology Department, Hospital del Mar, Barcelona, Catalonia, Spain

5Centro Investigacion Biomedica en Red de Salud Mental, Barcelona, Catalonia, Spain

6Department of Environmental Health Sciences, Fielding School of Public Health, University of California, Los Angeles, Los Angeles, California, USA

7IMIM-Parc Salut Mar, Barcelona, Catalonia, Spain

PDF icon PDF Version (1.1 MB)

  • Background:
    Proponents of the biophilia hypothesis believe that contact with nature, including green spaces, has a crucial role in brain development in children. Currently, however, we are not aware of evidence linking such exposure with potential effects on brain structure.
    Objective:
    We determined whether lifelong exposure to residential surrounding greenness is associated with regional differences in brain volume based on 3-dimensional magnetic resonance imaging (3D MRI) among children attending primary school.
    Methods:
    We performed a series of analyses using data from a subcohort of 253 Barcelona schoolchildren from the Brain Development and Air Pollution Ultrafine Particles in School Children (BREATHE) project. We averaged satellite-based normalized difference vegetation index (NDVI) across 100-m buffers around all residential addresses since birth to estimate each participant’s lifelong exposure to residential surrounding greenness, and we used high-resolution 3D MRIs of brain anatomy to identify regional differences in voxel-wise brain volume associated with greenness exposure. In addition, we performed a supporting substudy to identify regional differences in brain volume associated with measures of working memory (d′ from computerized n-back tests) and inattentiveness (hit reaction time standard error from the Attentional Network Task instrument) that were repeated four times over one year. We also performed a second supporting substudy to determine whether peak voxel tissue volumes in brain regions associated with residential greenness predicted cognitive function test scores.
    Results:
    Lifelong exposure to greenness was positively associated with gray matter volume in the left and right prefrontal cortex and in the left premotor cortex and with white matter volume in the right prefrontal region, in the left premotor region, and in both cerebellar hemispheres. Some of these regions partly overlapped with regions associated with cognitive test scores (prefrontal cortex and cerebellar and premotor white matter), and peak volumes in these regions predicted better working memory and reduced inattentiveness.
    Conclusion:
    Our findings from a study population of urban schoolchildren in Barcelona require confirmation, but they suggest that being raised in greener neighborhoods may have beneficial effects on brain development and cognitive function. https://doi.org/10.1289/EHP1876
  • Received: 9 March 2017
    Revised: 22 December 2017
    Accepted: 22 December 2017
    Published: 23 February 2018

    Address correspondence to P. Dadvand, ISGlobal, Doctor Aiguader 88, 08003 Barcelona, Catalonia, Spain. Telephone: 34 93 214 73 29. Email: payam.dadvand@isglobal.org

    Supplemental Material is available online (https://doi.org/10.1289/EHP1876).

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (150 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (37 KB)

Introduction

Currently, approximately half of the world’s population lives in cities, and it is predicted that by 2050, nearly 66% of people will live in urban areas worldwide (UN Department of Economic and Social Affairs 2015). Urban areas are characterized by a network of nonnatural built-up infrastructures where residents often have limited access to natural environments (Escobedo et al. 2011). Proponents of the biophilia hypothesis believe that contact with nature may have a defining role in human brain development (Kahn and Kellert 2002; Kellert 2005).

In a recent longitudinal study of >2,200 Barcelona schoolchildren (7–9 y old) (Dadvand et al. 2015a), we found that over a 12-month period, children who attended schools with higher outdoor greenness had a greater increase in working memory and a greater reduction in inattentiveness than children who attended schools with less surrounding greenness. The brain develops steadily during prenatal and early postnatal periods, which are considered the most vulnerable windows for effects of environmental exposures (Grandjean and Landrigan 2014). In this context, exposure to greenness early in life could be associated with beneficial structural changes in the developing brain. Accordingly, the overarching aim of this study was to determine whether lifelong exposure to residential surrounding greenness is associated with regional differences in brain volume based on 3-dimensional magnetic resonance imaging (3D MRI) among children attending primary school (the principal substudy). Toward this aim, we sought to confirm the beneficial nature of these differences by investigating the overlap between them and brain regions associated with cognitive function (supporting substudy I) and by studying the association between peak tissue volumes in these regions and objective measures of cognitive function (supporting substudy II).

Materials and Methods

We estimated lifelong exposure to residential surrounding greenness using satellite-based normalized difference vegetation index (NDVI) for all residential addresses of each participant since birth. High-resolution 3D MRIs of brain anatomy used to measure voxel-wise brain volume in the present study population were first obtained for studies of air pollution exposures and neurodevelopmental outcomes (Pujol et al. 2016a, 2016b). Measures of working memory and inattentiveness from a previous study of the entire Brain Development and Air Pollution Ultrafine Particles in School Children (BREATHE) cohort were based on computerized n-back and attentional network task (ANT) tests, respectively, that were repeated four times over a 12-month period (Dadvand et al. 2015a).

Study Participants

This study was developed in the context of the BREATHE project (Dadvand et al. 2015a). The general project design has been fully described elsewhere (Dadvand et al. 2015a). Briefly, all schoolchildren (n=5,019) without special needs in the second to fourth grades (7–10 y old) of 39 representative schools were invited to participate by letters or presentations in schools for parents, of whom 2,897 (59%) agreed to take part in BREATHE. All children had been in the school for >6 months (and 98% >1 y) before the beginning of the study. Participating schools were similar to the remaining schools in Barcelona in terms of the neighborhood socioeconomic vulnerability index (0.46 versus 0.50, Student’s t-test p=0.57).

Of 2,897 original BREATHE families, 1,564 (54%) agreed to receive information about the MRI study via post, email, or telephone, and 263 completed the imaging protocol in 2013. Ten children were excluded on the basis of image quality criteria (Pujol et al. 2016b); thus, 253 participants were included in the analyses. All parents or legal guardians signed the informed consent form approved by the Research Ethical Committee of the Hospital del Mar Medical Research Institute (IMIM)-Parc de Salut Mar, Barcelona, and the FP7-ERC-2010-AdG Ethics Review Committee.

Exposure to Greenness

To assess outdoor surrounding greenness, we applied NDVI derived from RapidEye data at 5 resolution. NDVI is an indicator of greenness and is calculated as follows:

where NIR and R are the land surface reflectance of the near-infrared and red (visible) parts of the spectrum, respectively. NDVI ranges between 1 and 1 with low values (e.g., 0.1) indicating water bodies, snow, and barren areas of rock and sand; and higher values indicating photosynthetically active vegetation (USGS 2015). RapidEye images were acquired from a constellation of five satellites 630 km above ground in sun-synchronous orbits. We generated our NDVI map using an image obtained on 23 July 2012. We used this map to estimate greenness surrounding all residential addresses for each participant from birth until the time of MRI evaluation, reported by parents at baseline enrollment into BREATHE. Given the highly developed built environment in Barcelona and the relatively short period of follow-up, the greenness in most areas would have been unlikely to change substantially during the study period. Lifelong residential surrounding greenness was abstracted as the average of NDVI in a buffer of 100 m (Dadvand et al. 2012, 2015a) around all the home addresses of each study participant since birth, weighted by the time (years) the participant stayed at each address.

Neuroimaging

A 1.5-Tesla Signa Excite system (General Electric) equipped with an eight-channel phased-array head coil was used for neuroimaging. High-resolution 3D anatomical images were obtained using an axial T1-weighted 3D fast spoiled gradient inversion recovery-prepared sequence (Pujol et al. 2016a, 2016b). A total of 134 contiguous slices were acquired with an inversion time of 400 ms, repetition time of 11.9 ms, echo time of 4.2 ms, flip angle of 15°, field of view of 30 cm, 256×256 pixel matrix, and slice thickness of 1.2 mm.

All of the anatomical images were visually inspected before analysis by a trained operator to detect any motion effect. Gray and white matter tissue volume [i.e., the volume proportion of gray matter to white matter and cerebrospinal fluid (CSF) and the volume proportion of white matter to gray matter and CSF] at a voxel level was measured using statistical parametric mapping (SPM8; FIL Methods Group 2013). SPM voxel-based morphometry (VBM) algorithms with DARTEL registration (FIL Methods Group 2013) were used to carry out the following processing steps (Pujol et al. 2016b): segmentation of anatomical images into gray and white matter tissue probability maps in their native space; normalization of images (warping) to a common group template and, later, from the common template to Montreal Neurological Institute (MNI) space by iteratively registering the individual segmented images with their average; scaling the MNI-normalized tissue probability maps by the Jacobian determinants estimated during the normalization procedure; and finally, reslicing images to 1.5 mm resolution and smoothing the resulting images with a 10×10× full width at half maximum (FWHM) Gaussian filter before entering into the study analyses. The scaling of tissue probability maps by Jacobian determinants, a procedure known as modulation, ensures that the total volume of the tissue is preserved even though the normalization to a common template implies the stretching and shrinking of structures. Modulated VBM can be interpreted as gray/white matter local volume, affected by differences in both tissue concentration and volume of local structures. Analyses for both gray and white matter were conducted within the same standard MNI space whole-brain mask.

Cognitive Function

Details on the BREATHE cognitive assessment methodology have been published elsewhere (Dadvand et al. 2015a). Briefly, cognitive function was assessed through repeated evaluations of working memory and attention. We selected these functions because they grow steadily during preadolescence (Rueda et al. 2005; Jaeggi et al. 2010). The n-back test and Attentional Network Task (ANT) have been frequently used to assess working memory (Jaeggi et al. 2010; Shelton et al. 2010) and inattentiveness (Rueda et al. 2004), respectively. The 2-back predicts general mental abilities (hereafter referred to as working memory), and the 3-back predicts superior functions such as fluid intelligence (hereafter referred to as superior working memory) (Forns et al. 2014). We have previously shown that for the BREATHE participants, the n-back and ANT indicators have statistical relationships with age, school performance, attention deficit hyperactivity disorder (ADHD) clinical criteria, behavioral problems, and maternal education (Forns et al. 2014).

From January 2012 to March 2013, children were evaluated every three months, using computerized n-back tests (with number stimuli) and ANTs, over four repeated visits in sessions lasting approximately 40 min. Having repeated measures is expected to improve the precision of the characterization of cognitive function. Groups of 10–20 children wearing ear protectors were assessed together and supervised by one trained examiner per 3–4 children (Dadvand et al. 2015a). The n-back parameter analyzed was d prime (d′), a measure of detection that subtracts the normalized false alarm rate from the hit rate (Z hitrateZfalsealarm×). A higher d′ value indicates more accurate test performance. In line with previous BREATHE studies [e.g., (Dadvand et al. 2015a)], from among the ANT measures, we chose hit reaction time standard error (HRT-SE), a measure of response speed consistency throughout the test (Conners and Multi-Health Systems Staff 2000; Dadvand et al. 2015a). A higher HRT-SE value indicates highly variable reactions related to inattentiveness.

Statistical Analysis

Principal substudy.

After individual preprocessing of each 3D anatomical image (Pujol et al. 2016a, 2016b), analyses were carried out to map, voxel-wise, the association between brain tissue measures and greenness exposure by conducting a separate regression analysis for each voxel in the brain using statistical parametric maps (SPM8; FIL Methods Group 2013). The statistical map, in effect, shows clusters as accumulations of individual voxels that were statistically significantly associated with the predictor of interest. Within SPM8, voxel-wise correlation analyses were performed with the lifelong greenness variable as the predictor without adjustment for any covariates. The resulting statistical maps require a correction for multiple comparisons because the statistical test for every voxel is strongly dependent on the tests of the neighboring voxels. We addressed this issue by performing 2,000 Monte Carlo simulations using AlphaSim as implemented in the SPM REST toolbox (Song et al. 2011). Input parameters to AlphaSim included an individual voxel threshold probability of 0.01, a cluster connection radius of 5 mm, 10 mm FWHM smoothness, and a comprehensive gray matter mask volume of 301,780 voxels (1.02 L). Results were considered significant with clusters of 2.2 mL (650 voxels) at a height threshold of p<0.01, which satisfied the family wise error (FWE) rate correction of pFWE<0.05.

Supporting substudy I.

For the first supporting substudy, we reanalyzed the MRI data using working memory (2-back d′), superior working memory (3-back d′), and inattentiveness (ANT HRT-SE) as predictors to identify brain regions with volumes that were significantly correlated with each predictor.

Supporting substudy II.

For the second supporting substudy, we performed a series of analyses with the three cognitive measures (2-back d′, 3-back d′, and HRT-SE) modeled as the outcome (dependent) variables and each brain region (cluster) that was significantly correlated with greenness as the predictor (one cognitive measure and one brain region per model). We used linear mixed effects models with child and school as random effects to account for the four repeated outcome measures within each child and the nonindependence of children within each school (Dadvand et al. 2015a). Each brain region (significant cluster) was modeled using the measured tissue volume of the voxel with the strongest association with greenness exposure (peak value) within the cluster as the independent variable.

Sensitivity Analyses

We repeated the principal analysis to identify brain regions associated with lifelong residential greenness using a 500-m versus a 100-m buffer to derive the exposure variable. We also evaluated, separately, the influences of maternal education [an indicator of individual socioeconomic status (SES)] and Urban Vulnerability Index (an indicator of neighborhood SES at the census-tract level [median area of 0.08km2 for the study area)] on the findings of our principal substudy. Urban Vulnerability Index is based on 21 indicators of urban vulnerability grouped into four themes: sociodemographic vulnerability (five indicators), socioeconomic vulnerability (six indicators), housing vulnerability (five indicators), and subjective perception of vulnerability (five indicators) (Spanish Ministry of Public Works 2012). For this study, we used the Urban Vulnerability index based on the 2001 Spanish Census (Spanish Ministry of Public Works 2012). Moreover, we explored the impacts of sex, age, and maternal education on our findings for supporting substudy II.

Results

Descriptive sociodemographic characteristics for the participants of the original BREATHE cohort and for those included in the present study are presented in Table 1. Participants included in the present study were similar to the original BREATHE cohort in terms of age and sex; however, they tended to have mothers with higher educational attainments, higher scores for working memory (2-back d′) and superior working memory (3-back d′), and a lower score for HRT-SE (consistent with less inattentiveness). In the present study, participants’ age, sex, and maternal education were not associated with their lifelong greenness exposure (see Table S1).

Table 1. Description of characteristics of the study participants in the current MRI study and the original BREATHE cohort.
Characteristic MRI study (n=253) BREATHE cohort (n=2,897) p-Valuea
Age (y) 8.4 (1.3) 8.5 (1.5) 0.13
Sex (female) 49.4% 49.5% 0.97
Maternal educational attainment 0.05
None or primary school 9.6% 12.7%
Secondary school 23.9% 28.7%
University 66.5% 58.6%
Residential surrounding greennessb 0.10 (0.06) 0.09 (0.07) 0.25
2-back (d′)c 2.5 (1.3) 2.4 (1.3) 0.06
3-back (d′)c 1.3 (1.0) 1.2 (0.9) 0.03
HRT-SE (ms)c 241.0 (112.1) 249.0 (110.9) 0.02

Note: For continuous variables, the median [interquartile range (IQR)] has been reported, and for categorical variables, the percentage of each category has been reported. BREATHE, Brain Development and Air Pollution Ultrafine Particles in School Children project; HRT-SE, hit reaction time standard error; MRI, magnetic resonance imaging.

ap-Value of chi-squared test for categorical variables and Mann–Whitney U test for continuous variables.

bAverage of normalized difference vegetation index (NDVI) across a buffer of 100 m around the residential address(es) since birth, weighted by the time the participant spent in each address.

cAverage of four repeated test values. 2-back d′ and 3-back d′ are indicators of working memory and superior working memory, respectively, and HRT-SE is an indicator of inattentiveness.

Principal Substudy

The principal analysis identified clusters in several brain regions with volumes that were significantly associated with lifelong exposure to greenness, including clusters mapped to gray matter in the right and left prefrontal cortex and in the left premotor cortex (Table 2 and Figure 1A) and to white matter in the right prefrontal region, in the left premotor region, and in both cerebellar hemispheres (Table 2 and Figure 1B). Adjusting for maternal education or for neighborhood SES (in separate models) reduced the size of all of the clusters, with the largest reduction after adjustment for maternal education (see Table S2). Cluster sizes for clusters in the left prefrontal cortex and for the superior white matter cluster in the right prefrontal region were no longer significant after adjustment for either variable, and the size of the left premotor region white matter cluster was no longer significant after adjustment for maternal education (see Table S2). However, for all clusters and both models, associations with the peak voxel volume remained significant. When we defined residential surrounding greenness as the NDVI average in a 500-m buffer around residential addresses of the study participants (instead of the 100-m buffer used in the aforementioned analyses), the voxel-wise Pearson’s correlation coefficient of the t-values of the statistical maps associated with these two sets of exposures was 0.90 for the white matter and 0.86 for the gray matter. The overlap (intersection) of clusters associated with greenness exposure across 100-m and 500-m buffers relative to the extent of the clusters associated with the 100-m buffer was 72.8% for white matter and 53.8% for gray matter.

Table 2. Regional clusters associated with lifelong exposure to greenness.
Location Cluster size,a voxels (mL) p-Valueb x y z Coordinatesc td p-Valuee
Gray matter
Left premotor cortex 1338 (4.5) <0.0005 36 2 66 3.29 0.0006
Left prefrontal cortex 1980 (6.7) <0.0005 38 33 27 3.03 0.001
Right prefrontal cortex
 Superior prefrontal 3233 (10.9)f <0.0005 32 45 21 3.09 0.001
 Inferior prefrontal (operculum) 3233 (10.9)f <0.0005 47 38 0 3.46 <0.0005
White matter
Cerebellum
 Left hemisphere 15938 (53.8)f <0.0005 29 60 47 3.46 <0.0005
 Right hemisphere 15938 (53.8)f <0.0005 30 74 39 3.14 0.0009
Left premotor region 981 (3.3) 0.006 32 2 61.5 3.22 0.0007
Right prefrontal region
 Superior cluster 840 (2.8) 0.017 44 41 12 3.29 0.0006
 Inferior cluster (operculum) 1373 (4.6) <0.0005 57 27 3 4.26 <0.0005

Note: All clusters reported correspond to cluster size p-values<0.05. Lifelong exposure to greenness was abstracted as the average of normalized difference vegetation index (NDVI) across a buffer of 100 m around the residential address(es) since birth, weighted by the time the participant spent at each address.

aThe number of voxels each showing statistically significant association with lifelong residential surrounding greenness.

bCluster size p-value that establishes the probability of the occurrence of a cluster of the specified voxel size or larger under the null hypothesis of a brain made of voxels with only spatially autocorrelated noise.

cx y z coordinates of the voxel with maximum (peak) t-value inside the corresponding cluster provided in Montreal Neurological Institute (MNI) space.

dMaximum (peak) t-value within the cluster. The t-values are generated from the voxel-wise regression model.

ePeak p-value that establishes the probability of occurrence of the specified t-value or greater generated by the voxel-wise regression model. The p-value of each row corresponds to the maximum (peak) t-value of each cluster.

fTwo parts of a single cluster.

Figure 1A is a 3D rendering of gray matter clusters in the right and left prefrontal and left promotor cortices associated with lifelong greenness exposure. Figure 1B is an orthogonal display of the brain including white matter clusters in cerebellum and left premotor regions associated with lifelong greenness exposure.
Figure 1. Regional gray and white matter volumes associated with lifelong residential surrounding greenness. Results are displayed using conventional canonical templates [Cortex_20484 surface mesh in (A) and MNI152_T1 template in (B)] in Montreal Neurological Institute (MNI) space with statistical parametric mapping (SPM8; FIL Methods Group 2013) software. Yellow and white areas indicate regional clusters with volumes positively associated with greenness (larger t-statistics). (A) Three-dimensional (3D) renderings of cortical gray matter (Cortex_20484 render) showing significant clusters in the right and left prefrontal cortex (left and right panels, respectively) and in the left premotor cortex (right panel). Results were considered significant with clusters of 2.2 mL (650 voxels) at a height threshold of p<0.01, which satisfied the family-wise error (FWE) rate correction of pFWE<0.05. (B) Orthogonal displays (sagittal, coronal, and axial views in the left, middle, and right panels, respectively, MNI152_T1 template) showing significant white matter clusters in the cerebellar hemispheres (all panels) and in the left premotor region (sagittal view). The right hemisphere appears on the right side of the axial and coronal views. Clusters with inverse associations between volumes and greenness would appear in cold colors (none identified.) See Table 2 for numeric data for each significant region. Residential greenness exposure was quantified based on the average Normalized Difference Vegetation Index (NDVI) within a 100-m buffer around all residences since birth, weighted by the time the participant spent at each address.

Supporting Substudy I

With regard to gray matter, smaller ANT HRT-SE (i.e., less inattentiveness) values were associated with higher gray matter volume across a broad area of the prefrontal lobes and inferior parietal lobules bilaterally, in addition to areas in the opercula and the inferior temporo-occipital cortex (Figure 2A, red color; see also Table S3). Similarly, higher d′ values in 2-back and 3-back tests (i.e., better working memory and better superior working memory, respectively) were associated with higher gray matter volumes in prefrontal, inferior parietal, and lateral temporal areas, in addition to the dorsal premotor cortex (Figure 2C and 2E, red color; see also Table S2). The gray matter clusters associated with greenness and those associated with cognitive outcomes partially overlapped at the dorsal prefrontal cortex (Figure 2A, 2C, and 2E, yellow color). This overlap was 37.4%, 22.2% and 32.2% for ANT HRT-SE, 2-back d′, and 3-back d′, respectively, relative to the area of the clusters associated with greenness exposure, or 6.6%, 6.3%, and 5.6%, respectively, relative to the area of the clusters associated with the corresponding cognitive test.

Figures 2A, 2C, and 2E are 3D renderings of right and left prefrontal cortex including clusters associated with lifelong greenness exposure as well as performance in cognitive tests. Figures 2B, 2D, and 2F are orthogonal displays of the brain including clusters associated with lifelong greenness exposure as well as performance in cognitive tests.
Figure 2. Regional gray and white matter volumes associated with lifelong residential surrounding greenness and cognitive performance. Results are displayed using conventional canonical templates [Cortex_20484 surface mesh in (A), (C), and (E) and MNI152_T1 template in (B), (D), and (F)] in Montreal Neurological Institute (MNI) space with statistical parametric mapping (SPM8; FIL Methods Group 2013) software. Green areas indicate regional volumes significantly associated with greenness (see Figure 1). Results were considered significant with clusters of 2.2 mL (650 voxels) at a height threshold of p<0.01, which satisfied the family-wise error (FWE) rate correction of pFWE<0.05. Red areas indicate regional clusters with volumes significantly associated with cognitive functions: hit reaction time standard error (HRT-SE; an indicator of inattentiveness) in (A) and (B); 2-back d′ (an indicator of working memory) in (C) and (D), and 3-back d′ (an indicator of superior working memory) in (E) and (F). The overlaps between regions associated with greenness and those associated with cognitive functions are shown in yellow. Voxels with significant results were binarized to the corresponding single color. (A) Three-dimensional (3D) renderings of cortical gray matter showing clusters negatively associated with HRT-SE in the right and left cortex (left and right panels, respectively). (B) Orthogonal displays (sagittal, coronal, and axial views in the left, middle, and right panels, respectively) showing white matter clusters negatively associated with HRT-SE. (C) 3D renderings of cortical gray matter showing clusters positively associated with 2-back. (D) Orthogonal displays showing white matter clusters positively associated with 2-back. (E) 3D renderings of cortical gray matter showing clusters positively associated with 3-back. (F) Orthogonal displays showing white matter clusters positively associated with 3-back. The right hemisphere appears on the right side of the axial and coronal views. See Table S3 for numeric data for each significant region. Residential greenness exposure was quantified based the average Normalized Difference Vegetation Index (NDVI) within a 100-m buffer around all residences since birth, weighted by the time the participant spent at each address. Bs, brainstem; Cer, cerebellum; Hp, hipoccampus; IPc, inferior parietal cortex; Op, operculum; PFc, prefrontal cortex; PM, premotor cortex; SM, sensorimotor cortex; Tc, temporal cortex; TOc, temporo-occipital cortex.

As shown in Figure 2, better performance on the ANT task and the n-back tests was mainly associated with greater white matter volumes in the cerebellum, the brainstem, the thalamus, part of the parietal lobe, the hippocampus, and the sensorimotor cortex extending to the premotor cortex (Figure 2B, 2D, and 2F, red color; see also Table S3). The gray matter clusters associated with greenness and those associated with cognitive outcomes partially overlapped at the cerebellar white matter and in a small portion of the premotor white matter (Figure 2B, 2D, and 2F, yellow color). This overlap was 51.2%, 17.0% and 64.2% for ANT HRT-SE, 2-back d′ and 3-back d′, respectively, relative to the area of the clusters associated with greenness exposure, or 1.3%, 1.2%, and 2.6%, respectively, relative to the area of the clusters associated with the corresponding cognitive test.

Supporting Substudy II

Of the gray matter clusters associated with greenness, peak volume measured in the left premotor cortex was positively associated with scores for working memory and superior working memory (2-back and 3-back d′, respectively) and inversely associated with inattentiveness (HRT-SE). In addition, peak volume in the superior cluster of the right prefrontal cortex was associated with working memory; peak volume in the inferior cluster of the right prefrontal cortex was associated with superior working memory and was inversely associated with inattentiveness; and peak volume in the left prefrontal cortex was inversely associated with inattentiveness (Table 3). Of the white matter clusters associated with greenness, peak volume in both the right and left cerebellar hemispheres was inversely associated with inattentiveness; peak volume in the right cerebellar hemisphere was positively associated with working memory; and peak volume in the left cerebellar hemisphere was positively associated with superior working memory (Table 3). In addition, peak volume in the left premotor region was positively associated with working memory and superior working memory (p=0.06 for the latter), and peak volume in the superior cluster of the right prefrontal region was inversely associated with inattentiveness (Table 3).

Table 3. Crude associations between gray/white matter volume in the peak voxel of clusters significantly associated with lifelong residential greenness (independent variable) and cognitive test scores measured on four occasions over 12 months (dependent variables) derived using separate linear mixed effect models with random effects for child and school.
Location Working memorya Superior working memoryb Inattentivenessc
Regression coefficient (95% CI) p-Value Regression coefficient (95% CI) p-Value Regression coefficient (95% CI) p-Value
Gray matter
Left premotor cortex 4.1 (1.2, 7.0) 0.01 4.1 (1.7, 6.6) <0.01 367 (639, 95) 0.01
Left prefrontal cortex 1.0 (0.4, 2.4) 0.18 1.0 (0.3, 2.1) 0.12 161 (292, 31) 0.02
Right prefrontal cortex
 Superior cluster 3.1 (0.6, 5.6) 0.02 1.3 (0.9, 3.4) 0.25 198 (434, 38) 0.10
 Inferior cluster 0.8 (0.7, 2.3) 0.30 1.7 (0.4, 3.0) 0.01 150 (287, 13) 0.03
White matter
Cerebellum
 Left hemisphere 0.4 (0.7, 1.4) 0.50 0.9 (0.0, 1.8) 0.05 97 (190, 4) 0.04
 Right hemisphere 1.9 (0.1, 3.7) 0.04 1.3 (0.2, 2.9) 0.09 226 (395, 58) 0.01
Left premotor region 2.8 (0.0, 5.7) 0.05 2.3 (0.1, 4.8) 0.06 176 (442, 89) 0.19
Right prefrontal region
 Superior cluster 1.4 (1.2, 4.1) 0.28 1.3 (0.9, 3.6) 0.24 244 (486, 2) 0.05
 Inferior cluster 1.9 (12.5, 8.7) 0.73 6.7 (2.3, 15.7) 0.15 180 (798, 1158) 0.72

Note: Adjustment was conducted for age, sex, and maternal education. Volume refers to the volume proportion of gray matter to white matter and cerebrospinal fluid (CSF) and the volume proportion of white matter to gray matter and CSF in each voxel. Voxel-wise volumes are expressed so that the total amount of tissue volume in the different brain structures is preserved during the process of normalization, which involves local stretching and shrinking of the brain structures. This is accomplished by the modulation of the segmented tissue probability maps using the Jacobian determinants derived from the spatial normalization step. CI, confidence interval.

aCharacterized using 2-back d′. A higher d′ indicates more accurate test performance.

bCharacterized using 3-back d′. A higher d′ indicates more accurate test performance.

cCharacterized using Attentional Network Task (ANT) Hit Reaction Time Standard Error (HRT-SE). A higher HRT-SE indicates more inattentiveness.

After adjustment of the analyses of supporting substudy II for age, sex, and maternal education, the associations for working memory and superior working memory remained generally consistent in terms of direction and statistical significance; however, the associations between white matter at the right cerebellar hemisphere and working memory and left cerebellar hemisphere and superior working memory lost their statistical significance (Table 4). Our observed associations for inattentiveness remained similar after this adjustment in terms of their direction; however, these associations became weaker and lost their statistical significance with the exception of the association for the left prefrontal area, which remained nearly statistically significant (p=0.06).

Table 4. Adjusted associations between gray/white matter volume in the peak voxel of clusters significantly associated with lifelong residential greenness (independent variable) and cognitive test scores measured on four occasions over 12 months (dependent variables) derived using separate linear mixed effect models with random effects for child and school.
Location Working memorya Superior working memoryb Inattentivenessc
Regression coefficient (95% CI) p-Value Regression coefficient (95% CI) p-Value Regression coefficient (95% CI) p-Value
Gray matter
Left premotor cortex 4.3 (1.2, 7.4) 0.01 3.8 (1.2, 6.5) <0.01 226 (502, 49) 0.11
Left prefrontal cortex 1.0 (0.5, 2.5) 0.17 0.8 (0.5, 2.1) 0.23 127 (2260, 7) 0.06
Right prefrontal cortex
 Superior cluster 3.0 (0.4, 5.6) 0.02 0.9 (1.4, 3.1) 0.44 108 (338, 123) 0.36
 Inferior cluster 0.8 (0.8, 2.5) 0.32 1.6 (0.2, 3.1) 0.02 78 (222, 67) 0.29
White matter
Cerebellum
 Left hemisphere 0.3 (0.8, 1.3) 0.62 0.7 (0.2, 1.6) 0.12 53 (146, 39) 0.26
 Right hemisphere 1.8 (0.2, 3.7) 0.08 0.9 (0.8, 2.6) 0.31 124 (296, 49) 0.16
Left premotor region 3.1 (0.2, 6.0) 0.04 2.1 (0.4, 4.6) 0.10 112 (371, 147) 0.40
Right prefrontal region
 Superior cluster 0.9 (1.8, 3.7) 0.52 0.5 (1.9, 2.9) 0.68 39 (284, 206) 0.76
 Inferior cluster 1.8 (12.6, 8.9) 0.74 5.8 (3.4, 15.0) 0.21 461 (481, 1403) 0.34

Note: Adjustment was conducted for age, sex, and maternal education. Volume refers to the volume proportion of gray matter to white matter and cerebrospinal fluid (CSF) and the volume proportion of white matter to gray matter and CSF in each voxel. Voxel-wise volumes are expressed so that the total amount of tissue volume in the different brain structures is preserved during the process of normalization, which involves local stretching and shrinking of the brain structures. This is accomplished by the modulation of the segmented tissue probability maps using the Jacobian determinants derived from the spatial normalization step. CI, confidence interval.

aCharacterized using 2-back d′. A higher d′ indicates more accurate test performance.

bCharacterized using 3-back d′. A higher d′ indicates more accurate test performance.

cCharacterized using Attentional Network Task (ANT) Hit Reaction Time Standard Error (HRT-SE). A higher HRT-SE indicates more inattentiveness.

Discussion

We evaluated whether an estimate of lifelong residential surrounding greenness was associated with differences in MRI-based measures of regional brain volumes in primary schoolchildren. Greenness exposure was positively associated with gray matter volume in clusters located in the left and right prefrontal cortices and in the left premotor cortex and with white matter volume in clusters located in the right prefrontal region, in the left premotor region, and in both cerebellar hemispheres. Clusters associated with the residential greenness exposure partly overlapped with more numerous and spatially extensive clusters that were positively associated with measures of working memory and inversely associated with a measure of inattentiveness.

Interpretation of Results

As the first investigators to evaluate such an association, we did not have an a priori hypothesis about specific brain regions that might be affected by exposure to residential greenness. However, considerable consistency existed between the regions in white and gray matter that were identified to be associated with greenness exposure in our principal substudy. For all cortical regions found to be associated with greenness exposure (with the exception of the left prefrontal cortex), we also observed changes in their adjacent white matter region. Furthermore, clusters associated with greenness overlapped by larger clusters associated with measures of working memory and inattentiveness, and the peak volumes measured in some of the clusters associated with greenness were positively associated with working memory and inversely associated with inattentiveness, particularly before adjustment for confounding by age, sex, and maternal education. These findings are in line with the available body of evidence showing that both premotor and prefrontal areas are key elements of the dorsal attentional network and are consistently activated during working memory tasks, specifically during the n-back test used in our study (Owen et al. 2005). In a functional MRI (fMRI) study of a sample of nine adult males, the n-back test was reported to activate the cerebellum (Stoodley et al. 2012), where we found an increase in white matter volume associated with greenness exposure as well as increases in 2-back and 3-back d′. A recent review noted that evidence from MRI studies suggests that children and adults with ADHD have lower prefrontal and premotor cortex and cerebellum volumes than those without ADHD (Friedman and Rapoport 2015). Consistently, we observed negative associations of inattentiveness with volumes of cerebellar vermis and hemispheres (supporting substudy I) and with peak values of greenness exposure–related clusters in cerebellar hemispheres and prefrontal cortex (supporting substudy II). However, the latter associations lost their statistical significance after controlling for age, sex, and maternal education.

Underlying Mechanisms

The biophilia hypothesis suggests that humans have important evolutionary bonds to nature (Wilson 1984; Kellert and Wilson 1993). Accordingly, contact with nature has been postulated to be essential for brain development in children (Kahn 1997; Kahn and Kellert 2002). Proponents of the biophilia hypothesis postulate that green spaces provide children with opportunities such as prompting engagement, discovery, creativity, risk taking, mastery, and control; bolstering sense of self; inspiring basic emotional states; and enhancing psychological restoration, which in turn are suggested to positively influence different aspects of brain development (Kahn and Kellert 2002; Kellert 2005; Bowler et al. 2010). In addition to exerting direct influence on brain development, green spaces might have indirect impacts mediated by other factors. For example, greener areas often have lower levels of traffic-related air pollution (Dadvand et al. 2015b) and noise (Gidlöf-Gunnarsson and Öhrström 2007). Moreover, people living in proximity to green spaces or in greener areas have been reported, albeit inconsistently, to be more physically active (James et al. 2015). Furthermore, green spaces are postulated to enrich microbial input from the environment (Rook 2013). Reduced exposure to air pollution and noise, increased physical activity, and enriched microbial input could lead to a beneficial impact of green spaces on brain development (Fedewa and Ahn 2011; Klatte et al. 2013; Rook 2013; Sunyer et al. 2015).

Limitations of the Study

Participants in the present study tended to have a better SES and to perform better on cognitive tests than the original BREATHE cohort, which might have resulted in selection bias. Although there are no known adverse effects of MRI exposure, the present study was conducted using a 1.5-Tesla magnet, which results in lower exposure but also generates lower-resolution images than would be obtained using a 3-Tesla magnet. We estimated greenness exposure at all residential addresses since birth, but we did not account for greenness in the vicinity of schools, friends’ homes, or other locations. Therefore, our residential exposure metric did not capture all possible exposure to greenness. Using high-resolution satellite data on greenness enabled us to account for small-area green spaces (e.g., home gardens, street trees, and green verges) in a standardized way. However, NDVI does not distinguish the types of vegetation or provide information on the quality of green spaces or on access to these spaces, which might have had implications in our study. By using an NDVI map obtained at a single point in time (2012), we effectively assumed that the spatial distribution of NDVI across our study region remained constant over the study period. The findings of our previous studies support the stability of the NDVI spatial contrast over years (Dadvand et al. 2012, 2014). We did not have data on parental cognitive status or on geographical factors such as walkability, which might have resulted in residual confounding in our results.

Conclusions

We identified several brain regions that had larger volumes in urban children with higher lifelong exposure to residential surrounding greenness. Brain regions whose volumes were increased in association with better cognitive test scores partly overlapped with some of the regions associated with greenness. In addition, peak volumes in some of the clusters associated with greenness also predicted better scores for some cognitive tests. Our findings provide new perspectives on how connections with the natural environment could potentially contribute to brain development. Further studies are needed to confirm our findings in other populations, settings, and climates; to evaluate other cognitive and neurological outcomes; to examine differences according to the nature and quality of green spaces (including specific types of vegetation) and children’s access to and use of them. Moreover, whether developmental effects on the structure of the brain contribute to associations between greenness exposure and cognitive development remains an open question to be evaluated by future studies.

Acknowledgments

The research leading to these results received funding from the European Community’s Seventh Framework Program (ERC-Advanced Grant) under grant agreement number 268479: The BREATHE project. The research leading to the methodology used for the exposure assessment in this study received funding from the European Community’s Seventh Framework Program (FP7/2007-2013) under grant agreement number 282996: The PHENOTYPE project. P.D. is funded by a Ramón y Cajal fellowship (RYC-2012-10995) awarded by the Spanish Ministry of Economy and Competitiveness. The sponsors and funding organizations had no role in the design or conduct of this research.

References

Bowler D, Buyung-Ali L, Knight T, Pullin A. 2010. A systematic review of evidence for the added benefits to health of exposure to natural environments. BMC Public Health 10:456, PMID: 20684754, 10.1186/1471-2458-10-456.

Conners CK, Multi-Health Systems Staff. 2000. Conners’ Continuous Performance Test II: Computer Program for Windows Technical Guide and Software Manual. North Tonwanda, NY:Multi-Health Systems.

Dadvand P, Nieuwenhuijsen MJ, Esnaola M, Forns J, Basagaña X, Alvarez-Pedrerol M. 2015a. Green spaces and cognitive development in primary schoolchildren. Proc Natl Acad Sci U S A 112(26):7937–7942, PMID: 26080420, 10.1073/pnas.1503402112.

Dadvand P, Rivas I, Basagaña X, Alvarez-Pedrerol M, Su J, De Castro Pascual M, et al. 2015b. The association between greenness and traffic-related air pollution at schools. Sci Total Environ 523:59–63, PMID: 25862991, 10.1016/j.scitotenv.2015.03.103.

Dadvand P, Sunyer J, Basagaña X, Ballester F, Lertxundi A, Fernández-Somoano A, et al. 2012. Surrounding greenness and pregnancy outcomes in four Spanish birth cohorts. Environ Health Perspect 120(10):1481–1487, PMID: 22899599, 10.1289/ehp.1205244.

Dadvand P, Wright J, Martinez D, Basagaña X, McEachan RRC, Cirach M, et al. 2014. Inequality, green spaces, and pregnant women: roles of ethnicity and individual and neighbourhood socioeconomic status. Environ Int 71:101–108, PMID: 24997306, 10.1016/j.envint.2014.06.010.

Escobedo FJ, Kroeger T, Wagner JE. 2011. Urban forests and pollution mitigation: Analyzing ecosystem services and disservices. Environ Pollut 159(8–9):2078–2087, PMID: 21316130, 10.1016/j.envpol.2011.01.010.

Fedewa AL, Ahn S. 2011. The effects of physical activity and physical fitness on children’s achievement and cognitive outcomes. Res Q Exerc Sport 82(3):521–535, PMID: 21957711, 10.1080/02701367.2011.10599785.

FIL (Functional Imaging Laboratory) Methods Group (and Honorary Members). 2013. SPM8 Manual. http://www.fil.ion.ucl.ac.uk/spm/doc/spm8_manual.pdf [accessed 26 June 2017].

Forns J, Esnaola M, López-Vicente M, Suades-González E, Alvarez-Pedrerol M, Julvez J, et al. 2014. The n-back test and the attentional network task as measures of child neuropsychological development in epidemiological studies. Neuropsychology 28(4):519–529, PMID: 24819069, 10.1037/neu0000085.

Friedman LA, Rapoport JL. 2015. Brain development in ADHD. Curr Opin Neurobiol 30:106–111, PMID: 25500059, 10.1016/j.conb.2014.11.007.

Gidlöf-Gunnarsson A, Öhrström E. 2007. Noise and well-being in urban residential environments: the potential role of perceived availability to nearby green areas. Landsc Urban Plan 83(2):115–126, 10.1016/j.landurbplan.2007.03.003.

Grandjean P, Landrigan PJ. 2014. Neurobehavioural effects of developmental toxicity. Lancet Neurol 13(3):330–338, PMID: 24556010, 10.1016/S1474-4422(13)70278-3.

Jaeggi SM, Buschkuehl M, Perrig WJ, Meier B. 2010. The concurrent validity of the N-back task as a working memory measure. Memory 18(4):394–412, PMID: 20408039, 10.1080/09658211003702171.

James P, Banay RF, Hart JE, Laden F. 2015. A review of the health benefits of greenness. Curr Epidemiol Rep 2(2):131–142, PMID: 26185745, 10.1007/s40471-015-0043-7.

Kahn PH. 1997. Developmental psychology and the biophilia hypothesis: children’s affiliation with nature. Dev Rev 17(1):1–61, 10.1006/drev.1996.0430.

Kahn PH, Kellert SR. 2002. Children and Nature: Psychological, Sociocultural, and Evolutionary Investigations. Cambridge, MA:MIT Press.

Kellert SR. 2005. Building for Life: Designing and Understanding the Human-Nature Connection. Washington, DC: Island Press.

Kellert SR, Wilson EO. 1993. The Biophilia Hypothesis. Washington, DC:Island Press.

Klatte M, Bergström K, Lachmann T. 2013. Does noise affect learning? A short review on noise effects on cognitive performance in children. Front Psychol 4:578, PMID: 24009598, 10.3389/fpsyg.2013.00578.

Owen AM, McMillan KM, Laird AR, Bullmore E. 2005. N-back working memory paradigm: a meta-analysis of normative functional neuroimaging studies. Hum Brain Mapp 25(1):46–59, PMID: 15846822, 10.1002/hbm.20131.

Pujol J, Fenoll R, Macià D, Martínez-Vilavella G, Alvarez-Pedrerol M, Rivas I. 2016a. Airborne copper exposure in school environments associated with poorer motor performance and altered basal ganglia. Brain Behav 6(6):e00467, PMID: 27134768, 10.1002/brb3.467.

Pujol J, Martínez-Vilavella G, Macià D, Fenoll R, Alvarez-Pedrerol M, Rivas I, et al. 2016b. Traffic pollution exposure is associated with altered brain connectivity in school children. NeuroImage 129:175–184, PMID: 26825441, 10.1016/j.neuroimage.2016.01.036.

Rook GA. 2013. Regulation of the immune system by biodiversity from the natural environment: an ecosystem service essential to health. Proc Natl Acad Sci U S A 110(46):18360–18367, PMID: 24154724, 10.1073/pnas.1313731110.

Rueda MR, Fan J, McCandliss BD, Halparin JD, Gruber DB, Lercari LP, et al. 2004. Development of attentional networks in childhood. Neuropsychologia 42(8):1029–1040, PMID: 15093142, 10.1016/j.neuropsychologia.2003.12.012.

Rueda MR, Rothbart MK, McCandliss BD, Saccomanno L, Posner MI. 2005. Training, maturation, and genetic influences on the development of executive attention. Proc Natl Acad Sci U S A 102(41):14931–14936, PMID: 16192352, 10.1073/pnas.0506897102.

Shelton JT, Elliott EM, Matthews RA, Hill BD, Gouvier WD. 2010. The relationships of working memory, secondary memory, and general fluid intelligence: working memory is special. J Exp Psychol Learn Mem Cogn 36(3):813–820, PMID: 20438278, 10.1037/a0019046.

Song X-W, Dong Z-Y, Long X-Y, Li S-F, Zuo X-N, Zhu C-Z, et al. 2011. REST: A toolkit for resting-state functional magnetic resonance imaging data processing. PLoS One 6(9):e25031, PMID: 21949842, 10.1371/journal.pone.0025031.

Spanish Ministry of Public Works. 2012. Atlas de la Vulnerabilidad Urbana en España 2001 y 2011: Metodología, contenidos y créditos (Edición de diciembre de 2015) [in Spanish]. http://www.fomento.gob.es/NR/rdonlyres/40668D5E-26B6-4720-867F-286BD55E1C6B/135960/20160201METODOLOGIAATLASVULNERABILIDAD2001Y2011.pdf [accessed 27 May 2016].

Stoodley CJ, Valera EM, Schmahmann JD. 2012. Functional topography of the cerebellum for motor and cognitive tasks: an fMRI study. NeuroImage 59(2):1560–1570, PMID: 21907811, 10.1016/j.neuroimage.2011.08.065.

Sunyer J, Esnaola M, Alvarez-Pedrerol M, Forns J, Rivas I, López-Vicente M, et al. 2015. Association between Traffic-related air pollution in schools and cognitive development in primary school children: a prospective cohort study. PLoS Med 12(3):e1001792, PMID: 25734425, 10.1371/journal.pmed.1001792.

UN Department of Economic and Social Affairs. 2015. World Urbanization Prospects: The 2014 Revision. New York, NY:United Nations.

USGS (U.S. Geological Survey). 2015. NDVI, the foundation for remote sensing phenology. https://phenology.cr.usgs.gov/ndvi_foundation.php [accessed 21 July 2017].

Wilson EO. 1984. Biophilia. Cambridge, MA:Harvard University Press.

Another Potential Risk Factor for ALS: Exposure to Traffic-Related Air Pollutants

PDF icon PDF Version (1.8 MB)

  • Published: 22 February 2018

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

Related EHP Article

Long-Term Air Pollution Exposure and Amyotrophic Lateral Sclerosis in Netherlands: A Population-based Case–control Study

Meinie Seelen, Rosario A. Toro Campos, Jan H. Veldink, Anne E. Visser, Gerard Hoek, Bert Brunekreef, Anneke J. van der Kooi, Marianne de Visser, Joost Raaphorst, Leonard H. van den Berg, and Roel C.H. Vermeulen

Little is known about what causes amyotrophic lateral sclerosis (ALS), a rare and debilitating neurological condition affecting approximately 450,000 individuals worldwide.1,2 Research suggests that the disease, which has a median survival time of just under three years after the onset of symptoms3 results from a complex set of genetic and exogenous factors;4 the vast majority of cases occur in people with no family history.4 To date, the best-established risk factor is smoking,5 but a report in Environmental Health Perspectives offers evidence that exposure to traffic-related air pollutants may also be an important risk factor.6

The study included 917 Dutch ALS patients and 2,662 controls from the general population. Using home addresses, the researchers estimated exposures of the participants to six measures of air pollution: the nitrogen oxides NO2 and NOx; three measures of particulate matter ( PM2.5, PM10, and PMcoarse, which is the fraction of PM calculated as the concentration of PM10 minus that of PM2.5); and fine particulate matter absorption ( PM2.5absorption, a marker for black soot or carbon).

For all six measures, estimated exposures were higher for ALS patients than for controls. Similarly, for the three measures most closely associated with traffic— NOx, NO2, and PM2.5absorption—individuals in the most-exposed group were more likely to have been diagnosed with ALS than those in the least-exposed group. All estimated pollutant levels fell below current European limits.6

The size of the estimated effects of NO2 and PM2.5absorption was similar to or higher than what previous studies have shown for smoking, says lead author and physician Meinie Seelen, who performed the research while earning her PhD at Utrecht University. The stronger association with traffic-related particles, which are the smallest of the pollutants measured, makes biological sense, she says.

Aerial photograph of traffic on a street in Amsterdam
Smoking is currently the best-established risk factor for amyotrophic lateral sclerosis (ALS). In a new study, exposures to three traffic-related pollutants had estimated effects on ALS that were similar to or higher than smoking. Image: © georgeclerk/iStockphoto.

“It has been demonstrated that ultrafine particles can circumvent the blood–brain barrier,” Seelen says. She explains that the tiny particles are deposited in the lining of the nose, and there is evidence that they may travel from there to the brain via the olfactory nerve.7 Previous research has shown that this may, in turn, cause chronic brain inflammation, oxidative stress, and other outcomes that could contribute to ALS.8,9,10

But there may be something else going on as well, says Jane Parkin Kullmann, a University of Sydney postgraduate researcher and toxicologist who studies behavioral and environmental factors in ALS. Traffic-related pollution often contains metals, including lead and mercury, as a result of processes such as the wearing of brakes and tires.11 These metals are known to be toxic to the brain.12,13

“As far as the biological rationale, it is very different for lead or mercury versus ultrafine particles; their mode of action is different,” says Kullmann, who was not involved with the present study. But ultimately, she notes, exposures to a combination of metals and fine particles—as opposed to just one pollutant or the other—could potentially play a role in ALS.

The new research adds to the results from an epidemiological study published in 2015 that investigated the relationship between air pollution and ALS occurrence.14 That study also reported an association, though in a far smaller population of 51 patients. Evelyn Talbott, senior author of the 2015 paper and a professor of epidemiology at the University of Pittsburgh, says the new study is a landmark in the field.

“The methodology was impressive, and they certainly performed a number of different sensitivity analyses,” she says. “It is a strong paper. Now that this has been done once, I am sure other people are going to look at the same thing.” In addition to replicating the study among different populations, future work could use animal models to investigate potential mechanisms, says Talbott, who also was not involved with the present study.

Future research could also seek to shed light on the still-shrouded etiology of ALS by investigating not only the potential role of pollution but also critical windows of exposure, says Roel Vermeulen, a professor at Utrecht University and senior author of the new paper. “Besides replication, the more nuanced questions also still have to be answered,” he says. “Is air pollution earlier in life or later in life more important? Is it a trigger, or does it accelerate? These are [aspects] that we do not know.”

Some previous epidemiological studies have already linked exposure to air pollution with incidence of Parkinson and Alzheimer diseases, the two most common neurodegenerative diseases.15,16,17,18 “It is possible,” suggests lead author Seelen, “that air pollution represents the first in a chain of events, although not necessarily the most important one.”


Nate Seltenrich covers science and the environment from Petaluma, California. His work has appeared in High Country News, Sierra, Yale Environment 360, Earth Island Journal, and other regional and national publications.

References

1. CDC (Centers for Disease Control and Prevention). 2017. National Amyotrophic Lateral Sclerosis (ALS) Registry. https://www.cdc.gov/als/Default.html [accessed 20 September 2017].

2. ALS Therapy Development Institute. ALS Frequently Asked Questions. https://www.als.net/about-als-tdi/als-faq/ [accessed 20 September 2017].

3. Huisman MHB, de Jong SW, van Doormaal PTC, Weinreich SS, Schelhaas HJ, van der Kooi AJ, et al. 2011. Population based epidemiology of amyotrophic lateral sclerosis using capture-recapture methodology. J Neurol Neurosurg Psychiatry 82(10):1165–1170, PMID: 21622937, 10.1136/jnnp.2011.244939.

4. Al-Chalabi A, Hardiman O. 2013. The epidemiology of ALS: a conspiracy of genes, environment and time. Nat Rev Neurol 9(11):617–628, PMID: 24126629, 10.1038/nrneurol.2013.203.

5. Armon C. 2009. Smoking may be considered an established risk factor for sporadic ALS. Neurology 73(20):1693–1698, PMID: 19917993, 10.1212/WNL.0b013e3181c1df48.

6. Seelen M, Toro Campos RA, Veldink JH, Visser AE, Hoek G, Brunekreef B, et al. 2017. Long-term air pollution exposure and amyotrophic lateral sclerosis in the Netherlands: a population-based case–control study. Environ Health Perspect 125(9):097023, 10.1289/EHP1115.

7. Elder A, Gelein R, Silva V, Feikert T, Opanashuk L, Carter J, et al. 2006. Translocation of inhaled ultrafine manganese oxide particles to the central nervous system. Environ Health Perspect 114(8):1172–1178, PMID: 16882521, 10.1289/ehp.9030.

8. Block ML, Elder A, Auten RL, Bilbo SD, Chen H, Chen J-C, et al. 2012. The outdoor air pollution and brain health workshop. Neurotoxicology 33(5):972–984, PMID: 22981845, 10.1016/j.neuro.2012.08.014.

9. Costa LG, Cole TB, Coburn J, Chang Y-C, Dao K, Roque P. 2014. Neurotoxicants are in the air: convergence of human, animal, and in vitro studies on the effects of air pollution on the brain. Biomed Res Int 2014:736385, PMID: 24524086, 10.1155/2014/736385.

10. Levesque S, Surace MJ, McDonald J, Block ML. 2011. Air pollution & the brain: subchronic diesel exhaust exposure causes neuroinflammation and elevates early markers of neurodegenerative disease. J Neuroinflammation 8:105, PMID: 21864400, 10.1186/1742-2094-8-105.

11. Lough GC, Schauer JJ, Park JS, Shafer MM, Deminter JT, Weinstein JP. 2005. Emissions of metals associated with motor vehicle roadways. Environ Sci Technol 39(3):826–836, PMID: 15757346, 10.1021/es048715f.

12. Sanders T, Liu Y, Buchner V, Tchounwou PB. 2009. Neurotoxic effects and biomarkers of lead exposure: a review. Rev Environ Health 24(1):15–45, PMID: 19476290, 10.1515/REVEH.2009.24.1.15.

13. U.S. EPA (U.S. Environmental Protection Agency). 2017. Health Effects of Exposure to Mercury. https://www.epa.gov/mercury/health-effects-exposures-mercury [accessed 20 September 2017].

14. Malek AM, Barchowsky A, Bowser R, Heiman-Patterson T, Lacomis D, Rana S, et al. 2015. Exposure to hazardous air pollutants and the risk of amyotrophic lateral sclerosis. Environ Pollut 197:181–186, PMID: 25544309, 10.1016/j.envpol.2014.12.010.

15. Kioumourtzoglou M-A, Schwartz JD, Weisskopf MG, Melly SJ, Wang Y, Dominici F, et al. 2016. Long-term PM2.5 exposure and neurological hospital admissions in the northeastern United States. Environ Health Perspect 124(1):23–29, PMID: 25978701, 10.1289/ehp.1408973.

16. Liu R, Young MT, Chen J-C, Kaufman JD, Chen H. 2016. Ambient air pollution exposures and risk of Parkinson disease. Environ Health Perspect 124(11):1759–1765, PMID: 27285422, 10.1289/EHP135.

17. Ritz B, Lee P-C, Hansen J, Lassen CF, Ketzel M, Sørensen M, et al. 2016. Traffic-related air pollution and Parkinson’s disease in Denmark: a case–control study. Environ Health Perspect 124(3):351–356, PMID: 26151951, 10.1289/ehp.1409313.

18. Ranft U, Schikowski T, Sugiri D, Krutmann J, Krämer U. 2009. Long-term exposure to traffic-related particulate matter impairs cognitive function in the elderly. Environ Res 109(8):1004–1011, PMID: 19733348, 10.1016/j.envres.2009.08.003.

DDT and Obesity in Humans: Exploring the Evidence in a New Way

PDF icon PDF Version (2.1 MB)

  • Published: 22 February 2018

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

Although conventional wisdom holds that overeating and a sedentary lifestyle are the main causes of obesity, increasing evidence indicates that additional risk may be conferred by exposure to obesogens, environmental chemicals suspected of influencing the development and maintenance of adipose (fat) tissue.1,2,3,4 The insecticide dichlorodiphenyltrichloroethane (DDT) and its breakdown products are among many such suspected obesogens.2,5 A systematic review and meta-analysis in Environmental Health Perspectives now concludes that the collective evidence supports the presumption that DDT is a human obesogen.5

The excess accumulation of body fat can cause adverse health effects including diabetes, cardiovascular disease, and cancer.2,6 Obesogens are thought to disrupt the molecular mechanisms controlling the development and maintenance of adipose tissue. This disruption has the potential to produce larger and more numerous fat cells, which could in turn lead to obesity and related complications.1,6 Obesogens can also alter programing of metabolic set points, appetite, and satiety.6

From the 1940s to the 1970s, DDT was used widely to control mosquitoes and the diseases they transmit.7,8 As a result of concerns about its adverse effects on wildlife and humans and its persistence in the environment, its use was largely banned,7,8 although it is still used in some countries to fight mosquito-borne diseases.9 Despite the relatively limited use today, DDT is highly persistent in the human body, and most people throughout the world carry at least traces of it and its metabolites in their bodies.5,8

Archival photograph of Marines applying pesticides to clothing during World War II
U.S. Marines dust their clothing with DDT shortly before landing on Iwo Jima in February 1945. For several decades DDT was widely used to fight mosquitoes. Although the insecticide has been largely banned around the world, it is still used in some Asian and African countries to control malaria. Image: U.S. National Archives.

The current review evaluated the body of research on DDT as an obesogen using what is known as the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach. The authors applied this approach with guidance from two sources: the Handbook for Conducting a Literature-Based Health Assessment published by the National Toxicology Program Office of Health Assessment and Translation (OHAT),10 and the Navigation Guide developed by a work group of nearly two dozen environmental health experts.11

The GRADE approach originated as a means of methodically and rigorously assessing human studies in clinical medicine and public health. The OHAT Handbook and Navigation Guide adapt this approach to integrate lines of evidence from epidemiological, animal, and in vitro studies. Investigators can then systematically assess the findings of various studies and conclude how strongly they collectively support a particular conclusion.

“This kind of hazard identification that integrates across experimental systems and human systems has not been done for an obesogen before,” says review coauthor Michele A. La Merrill, an assistant professor of environmental toxicology at the University of California, Davis. “I think it is important, because as the first study to really do this for obesogens, it adds legitimacy to this new and emerging field that not everyone is familiar with or necessarily believes has real scientific merit,” she says.

For the current review, La Merrill and colleagues searched for literature related to whether exposure to DDT may increase obesity in humans, following a protocol they developed prior to conducting this review. The researchers first identified studies pertaining to DDT and obesity, markers of obesity, or underlying mechanisms of obesity in human epidemiological research, animal experiments, and in vitro investigations. Human evidence and in vivo assessments of adiposity in animals made up the primary evidence for evaluating the DDT–obesity relationship. In vitro assessment of adipocyte development and in vivo studies of energy balance, lipids, and molecular markers were considered supporting evidence potentially informing the biological plausibility of findings in the primary evidence.

The authors graded each category of evidence based on factors such as risk of bias in study design, inconsistency among studies, and imprecision. They judged that both the relevant human epidemiology and the primary in vivo toxicology constituted a moderate level of evidence, which along with a moderate level of supporting evidence led to the overall conclusion that DDT can be presumed to be a hazard to humans, namely by increasing the risk of obesity.5

The researchers also identified epidemiological and experimental research needs for refining understanding of the DDT–obesity relationship and protecting public health. Gaps in current research, such as the limited assessment of dose–response relationships and a dearth of prospective epidemiological data, prevent more definitive conclusions.

“I think the authors made reasonable conclusions and evaluated the strengths and weaknesses of their study appropriately,” says Bruce Blumberg, a professor of developmental and cell biology at the University of California, Irvine, and a pioneer in the study of obesogens. “This paper is notable for the thoroughness of the analysis and the transparency of the methodology.” Blumberg was not involved in the review.

The study serves as a model for deriving conclusions from research on other obesogens, notes Blumberg. For him, the current paper is proof of principle for the impact that such analyses will have once additional prospective cohort studies have been conducted on known obesogens.


Julia R. Barrett, MS, ELS, a Madison, Wisconsin–based science writer and editor, is a member of the National Association of Science Writers and the Board of Editors in the Life Sciences.

References

1. Janesick AS, Blumberg B. 2016. Obesogens: an emerging threat to public health. Am J Obstet Gynecol 214(5):559–565, PMID: 26829510, 10.1016/j.ajog.2016.01.182.

2. Heindel JJ, Newbold R, Schug TT. 2015. Endocrine disruptors and obesity. Nat Rev Endocrinol 11(11):653–661, PMID: 26391979, 10.1038/nrendo.2015.163.

3. Brown RE, Sharma AM, Ardern CI, Mirdamadi P, Mirdamadi P, Kuk JL. Secular differences in the association between caloric intake, macronutrient intake, and physical activity with obesity. Obes Res Clin Pract 10(3):243–255, PMID: 26383959, 10.1016/j.orcp.2015.08.007.

4. Klimentidis YC, Beasley TM, Lin H-Y, Murati G, Glass GE, Guyton M, et al. 2011. Canaries in the coal mine: a cross-species analysis of the plurality of obesity epidemics. Proc Biol Sci 278(1712):1626–1632, PMID: 21106594, 10.1098/rspb.2010.1890.

5. Cano-Sancho G, Salmon AG, La Merrill MA. 2017. Association between exposure to p,pʹ-DDT and its metabolite p,pʹ-DDE with obesity: integrated systematic review and meta-analysis. Environ Health Perspect 125(9):096002, PMID: 28934091, 10.1289/EHP527.

6. Levian C, Ruiz E, Yang X. 2014. The pathogenesis of obesity from a genomic and systems biology perspective. Yale J Biol Med 87(2):113–126, PMID: 24910557.

7. Bouwman H, van den Berg H, Kylin H. 2011. DDT and malaria prevention: addressing the paradox. Environ Health Perspect 119(6):744–747, PMID: 21245017, 10.1289/ehp.1002127.

8. Eskenazi B, Chevrier J, Rosas LG, Anderson HA, Bornman MS, Bouwman H, et al. 2009. The Pine River statement: human health consequences of DDT use. Environ Health Perspect 117(9):1359–1367, PMID: 19750098, 10.1289/ehp.11748.

9. van den Berg H, Manuweera G, Konradsen F. 2017. Global trends in the production and use of DDT for control of malaria and other vector-borne diseases. Malaria J 16:401, PMID: 28982359, 10.1186/s12936-017-2050-2.

10. OHAT (Office of Health Assessment and Translation). 2015. Handbook for Conducting a Literature-Based Health Assessment Using OHAT Approach for Systematic Review and Evidence Integration. Washington, DC:National Toxicology Program. https://ntp.niehs.nih.gov/ntp/ohat/pubs/handbookjan2015_508.pdf [accessed 12 January 2018].

11. Woodruff TJ, Sutton P. Navigation Guide Work Group. 2011. An evidence-based medicine methodology to bridge the gap between clinical and environmental health sciences. Health Aff (Millwood) 30(5):931–937, PMID: 21555477, 10.1377/hlthaff.2010.1219.

Associations between Personal Care Product Use Patterns and Breast Cancer Risk among White and Black Women in the Sister Study

Author Affiliations open

1Office of Health Assessment and Translation, National Toxicology Program, National Institute of Environmental Health Sciences (NIEHS), National Institutes of Health (NIH), Department of Health and Human Services (DHHS), Research Triangle Park, North Carolina, USA

2Department of Epidemiology, University of North Carolina, Chapel Hill, North Carolina, USA

3Department of Biostatistics, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, USA

4Carolina Population Center, University of North Carolina, Chapel Hill, North Carolina, USA

5Epidemiology Branch, Division of Intramural Research, NIEHS, NIH, DHHS, Research Triangle Park, North Carolina, USA

PDF icon PDF Version (181 KB)

  • Background:
    Many personal care products include chemicals that might act as endocrine disruptors and thus increase the risk of breast cancer.
    Objective:
    We examined the association between usage patterns of beauty, hair, and skin-related personal care products and breast cancer incidence in the Sister Study, a national prospective cohort study (enrollment 2003–2009).
    Methods:
    Non-Hispanic black (4,452) and white women (n=42,453) were examined separately using latent class analysis (LCA) to identify groups of individuals with similar patterns of self-reported product use in three categories (beauty, skin, hair). Multivariable Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between product use and breast cancer incidence.
    Results:
    A total of 2,326 women developed breast cancer during follow-up (averagefollow-up=5.4 y). Among black women, none of the latent class hazard ratios was elevated, but there were <100 cases in any category, limiting power. Among white women, those classified as “moderate” and “frequent” users of beauty products had increased risk of breast cancer relative to “infrequent” users [HR=1.13 (95% CI: 1.00, 1.27) and HR=1.15 (95% CI: 1.02, 1.30), respectively]. Frequent users of skincare products also had increased risk of breast cancer relative to infrequent users [HR=1.13 (95% CI: 1.00, 1.29)]. None of the hair product classes was associated with increased breast cancer risk. The associations with beauty and skin products were stronger in postmenopausal women than in premenopausal women, but not significantly so.
    Conclusions:
    This work generates novel hypotheses about personal care product use and breast cancer risk. Whether these results are due to specific chemicals or to other correlated behaviors needs to be evaluated. https://doi.org/10.1289/EHP1480
  • Received: 9 December 2016
    Revised: 5 January 2018
    Accepted: 5 January 2018
    Published: 21 February 2018

    Address correspondence to K. W. Taylor, National Institute of Environmental Health Sciences, P.O. Box 12233, MD K2-04, Research Triangle Park, NC 27709 USA. Telephone: (984) 287-3194. Email: kyla.taylor@nih.gov

    *Current affiliations: Department of Statistical Science, Duke University, Duke University, Durham, NC, USA, and Duke Global Health Institute, Duke University, Durham, NC, USA.

    Supplemental Material is available online (https://doi.org/10.1289/EHP1480).

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (296 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (96 KB)

Introduction

There is concern that use of personal care products (e.g., cosmetics, lotions, and fragrances) may be associated with breast cancer risk (Brody et al. 2007). These products are a possible source of human exposure to endocrine-disrupting chemicals, such as phthalates, parabens, and phenols (Braun et al. 2014; Dodson et al. 2012; Meeker et al. 2013). Endocrine-disrupting chemicals have been hypothesized to mimic the carcinogenic effects of estrogenic exposures (Chen 2008; Morgan et al. 1998). For example, phthalates, an ingredient commonly used in personal care products, have been associated with risk of breast cancer (López-Carrillo et al. 2010; Shanle and Xu 2011). However, endocrine-disrupting chemicals have a much lower affinity to the estrogen receptor (ER) than does estradiol (Shanle and Xu 2011).

Women are the primary consumers of many personal care products and are disproportionately exposed to the chemicals within these products (CDC 2012). A national survey of >2,300 U.S. women reported that the average adult woman uses approximately 12 individual personal care products each day and that more than a quarter of all women use ≥15 products per day (EWG 2004). A cross-sectional analysis of women in northern Mexico reported that increased personal care product use was associated with higher urinary concentrations of monoethyl phthalate (MEP) (Romero-Franco et al. 2011), a metabolite of phthalates that are used in a range of personal care products (Koo and Lee 2004). The Environmental and Reproductive Health (EARTH) Study, a cohort study of women attending fertility clinics (18–45 y old), reported evidence of a monotonic dose–response relationship between the number of products used and urinary paraben and phthalate metabolite concentrations (Braun et al. 2014). Few studies have evaluated the association between individual personal care products, or components of products, and breast cancer risk. In addition, the studies focused on deodorant/antiperspirant and hair dye use have generally not supported an increase in risk of breast cancer (Fakri et al. 2006; López-Carrillo et al. 2010; McGrath 2003; Mirick et al. 2002; Rollison et al. 2006; Takkouche et al. 2005, 2009). One population based case–control study of women residing in northern Mexico, with 233 histopathologically confirmed breast cancer cases and 221 age-matched controls, did report that exposure to MEP may be associated with increased risk of breast cancer (López-Carrillo et al. 2010). However, the same study also reported that exposure to other phthalates used in personal care products [i.e., monobenzyl phthalate (MBzP) and mono (3-carboxypropyl) phthalate (MCPP)] was inversely associated with breast cancer (López-Carrillo et al. 2010).

A challenge facing epidemiologic studies of personal care products lies in the fact that individual chemical exposures or personal care product usage will not capture overarching patterns of use across multiple products. We have previously shown that latent class analysis could identify mutually exclusive groups of women with differing patterns of personal care product use among participants in the Sister Study (Taylor et al. 2017). We also found racial differences in population distribution across product classes. For example, we observed race-related patterns of hair product use consistent with findings from previous, smaller studies (James-Todd et al. 2011; Silva et al. 2004; Tiwary 1998; Tiwary and Ward 2003). Breast cancer incidence and mortality rates also vary by race. Research has shown that black women have higher breast cancer mortality than white women even though current mammography screening rates are similar, or even slightly higher, in black women than in white women (DeSantis et al. 2016). Therefore, in the present work, we stratified by race when we evaluated patterns of beauty-, hair-, and skin-related personal care product use in association with breast cancer risk.

Methods

The Sister Study is directed at identifying environmental and genetic risk factors for breast cancer in a cohort of 50,884 women in the continental United States and Puerto Rico, enrolled during 2003–2009. Women eligible for enrollment were 35–74 y of age and had at least one sister diagnosed with breast cancer but were cancer-free themselves. Enrollment activities included a computer-assisted telephone interview and self-administered questionnaires that elicited information about environmental and reproductive exposures. This analysis includes breast cancers diagnosed as of June 1, 2014 (data release 4.1, updated July 2014). The Sister Study was approved by the institutional review boards at the National Institute of Environmental Health Sciences and Copernicus Group. All study participants provided written informed consent.

Breast Cancer Ascertainment

Participants reported breast cancer diagnoses on annual and biennial health questionnaires or by calling the Sister Study helpline. Women who reported an incident breast cancer during follow-up were asked to authorize release of pertinent medical records. Response rates were >94% over follow-up (Nichols et al. 2013). Among participants in our sample population, 2,326 breast cancers were reported for 304,034 person-years (average follow-up ∼5.4 years). At the time of the present analysis, pathology reports or medical records had been obtained for >80% of these cases (n=1,923). Confirmation of self-reported breast cancer diagnoses by medical record was very high [positive predictivevalue (PPV)=99.5%] (NIEHS 2010). After medical record review, self-reported ER status information was confirmed for 99% of ER-positive cases and 85% of ER-negative cases. Because agreement between self-reported and medically abstracted data was high, we used self-reported tumor information when medical records were not available.

Personal Care Product Exposure

Information on the use of 48 personal care products was self-reported during the enrollment phase of the study (see Table S1) by inquiring about frequency of use (five-level response option) during the previous 12 months. The five response options in the questionnaire varied according to intended use of the product. For example, the response options for a product intended to be used regularly (e.g., hand lotion) included: a) did not use, b) used less than once a month, c) used 1–3 times per month, d) 1–5 times per week, e) >5 times per week. Response options for products that are used less often (e.g., hair dye) included: a) did not use, b) 1–2 times a year, c) every 3–4 months, d) every 5–8 weeks, e) once a month or more. Because PROC LCA requires the same number of response options for each item (i.e., each personal care product), and to ensure that each of the response options had an adequate sample size (≥10% of the sample population), the five original response options were condensed into three (rarely/never used, moderate use, and frequent use); this was done for each product individually (see Figure S1) and was based on the distribution of participants who fell into each frequency-of-use option. Products were then categorized as beauty (n=14), hair (n=15), or skincare (n=19) products, and separate latent classes were defined for each product category. As described in more detail in our previous paper (Taylor et al. 2017), we performed latent class analyses (LCAs) using PROC LCA (Lanza et al. 2015) and SAS statistical software (version 9.3; SAS Institute Inc.). PROC LCA is an SAS procedure for latent class analysis (LCA) developed by the Methodology Center at Penn State (https://methodology.psu.edu/downloads/proclcalta). It allows the user to preprocess data, fit a variety of latent class models, and postprocess the results, all within SAS.

Latent classes within each product category were defined by item-response probabilities (Dean and Raftery 2010) for the products driving each class (Lanza et al. 2007). To identify the classes, we fit a sequence of LCA models starting with two classes and increasing up to six for each model, and we used Akaike’s information criterion (AIC), the Bayesian information criterion (BIC), and entropy (Lanza et al. 2007) to select the optimum number of classes. We used a common classify-analyze approach (the maximum-probability assignment rule) to assign each participant to the class in which she had the highest posterior probability of membership (Bray et al. 2012). To reduce dimensionality and improve interpretability, classification, and precision, each product category included in our model was limited to the personal care products that were most useful for distinguishing between latent classes (i.e., ≥10% difference in posterior probabilities between classes), as described by Dean and Raftery (2010). Within each product category, latent classes were described and considered as exposure groups. Women with missing data for any of the individual products included in the corresponding product category (i.e., beauty products, hair products, and skincare products) were classified as missing for a latent class assignment in that product category and were not included in statistical analyses.

Statistical Analysis

The present analysis was limited to non-Hispanic white (n=42,447, 91%) and non-Hispanic black (n=4,450, 9%) women (Table 1). Multivariable Cox proportional hazards models were used to estimate adjusted hazard ratios (adjHRs) and 95% confidence intervals (CIs) for associations between the personal care product latent classes and breast cancer risk. Statistical models used age as the time scale, where participants entered the analysis at their enrollment age (left-truncation) and accrued person-time until they exited at their cancer diagnosis or were administratively censored at their age at last follow-up. Women who reported that they had undergone natural menopause, bilateral oophorectomy, irradiation to the ovaries, or otherwise reported cessation of menstruation were classified as postmenopausal; women who reported that they were still cycling were classified as premenopausal. In analyses investigating associations by menopausal status at the time of breast cancer diagnosis, the person-time of women who became postmenopausal during the follow-up period was counted as premenopausal time at risk up until menopause (after which it was censored for the premenopausal analysis), and subsequent person-time after menopause was counted as postmenopausal person-time at risk. The proportional hazards assumption was visually assessed using ln-ln survival plots; there was no suggestion of time-variant associations.

 

Table 1. Descriptive statistics of sample population by race and number of breast cancer cases.
Characteristic White women Black women
Total n (%) Cases n (%) Total n (%) Cases n (%)
Age (y)
 Total 42,447 2,146 4,450 180
 35–39 1,601 (4) 49 (2) 237 (5) 8 (4)
 40–44 3,536 (8) 142 (7) 480 (11) 15 (8)
 45–49 6,248 (15) 293 (14) 783 (18) 27 (15)
 50–54 8,022 (19) 384 (18) 993 (22) 30 (17)
 55–59 8,436 (20) 418 (19) 960 (22) 48 (27)
 60–64 6,631 (16) 384 (18) 596 (13) 27 (15)
 64–69 5,209 (12) 308 (14) 280 (6) 21 (12)
 70–74 2,764 (7) 168 (8) 121 (3) 4 (2)
Menopausal statusa
 Total 42,432 2,145 4,447 180
 Premenopausal 14,755 (35) 691 (32) 1,937 (44) 67 (37)
 Postmenopausal 27,677 (65) 1,454 (68) 2,510 (56) 113 (63)
 Missing or unknown 15 1 3 0
Education
 Total 42,444 2,146 4,449 180
 <High school 324 (1) 15 (1) 44 (1) 14 (8)
 ≥High school 42,120 (99) 2,131 (99) 4,405 (99) 166 (92)
 Missing 3 0 1 0
Geographic location
 Total 42,424 2,145 4,447 180
 Northeast 7,744 (18) 292 (18) 420 (9) 17 (9)
 Midwest 12,237 (29) 597 (28) 996 (22) 39 (22)
 South 12,885 (30) 672 (31) 2,685 (60) 106 (59)
 West 9,558 (23) 484 (23) 346 (8) 18 (10)
 Missing, don’t know, refused, or PR 23 1 3 0
Adult BMI kg/m2
 Total 42,447 2,146 4,450 180
 <25 2,253 (5) 833 (39) 737 (17) 32 (18)
 25 to <30 13,282 (31) 690 (32) 1,406 (32) 55 (31)
 ≥30 11,612 (27) 623 (29) 2,307 (52) 93 (52)
Oral contraceptive use
 Total 42,413 2,145 4,448 180
 Ever 6,538 (15) 363 (17) 621 (14) 23 (13)
 Never 35,875 (85) 1,782 (83) 3,827 (86) 157 (87)
 Missing 34 1 2 0
Hormone replacement therapy use
 Total 42,301 2,143 4,441 179
 Ever 20,645 (49) 989 (46) 2,895 (65) 110 (61)
 Never 21,655 (51) 1,154 (54) 1,546 (35) 69 (38)
 Missing 146 3 9 1
Age at menarche (y)
 Total 42,447 2,146 4,339 180
 <12 8,315 (20) 438 (20) 3,339 (75) 47 (26)
 ≥12 34,132 (80) 1,708 (80) 1,000 (25) 133 (74)
 Missing 0 0 111 0
Parity
 Total 42,427 2,146 4,436 180
 Nulliparous 7,782 (18) 419 (20) 831 (19) 19 (11)
 1–2 children 21,783 (51) 1,108 (52) 2,512 (56) 111 (62)
 ≥3 children 12,862 (30) 611 (28) 1,093 (25) 50 (28)
 Missing 20 0 14 0
Age at first live birth (y)
 Total 42,427 2,146 4,436 180
 Nulliparous 7,782 (18) 419 (20) 831 (19) 19 (11)
 <26 23,004 (54) 1,116 (52) 2,729 (61) 122 (68)
 ≥26 11,641 (27) 611 (28) 876 (20) 39 (22)
 Missing 20 0 14 0
Total months breastfeeding
 Total 42,447 2,146 4,450 180
 <12 31,550 (74) 1,643 (77) 3,881 (87) 154 (86)
 ≥12 10,897 (26) 503 (23) 569 (13) 26 (14)
Family history
 Total 41,333 2,102 4,089 170
 One sister with breast cancer 29,986 (71) 1,343 (63) 3,061 (69) 115 (64)
 ≥1 sister and/or a mother with breast cancer 11,347 (27) 759 (35) 1,028 (23) 55 (31)
 Missing 1,114 44 361 10
Alcohol consumption
 Total 42,392 2,143 4,434 180
 Never drinker 1,303 (3) 61 (3) 238 (5) 11 (6)
 Former drinker 5,855 (14) 278 (13) 1,101 (25) 49 (27)
 Currently drink <1 drink/d 28,905 (68) 1,458 (68) 2,861 (64) 112 (62)
 Currently drink ≥1 drink/d 6,329 (15) 346 (16) 234 (5) 8 (4)
 Missing 55 3 16 0
Smoking
 Total 42,391 2,144 4448 180
 Never smoker 22,180 (52) 1,068 (50) 2747 (62) 113 (63)
 Former smoker 16,774 (40) 914 (43) 1252 (28) 57 (32)
 Current smoker 3,437 (8) 162 (8) 449 (10) 10 (6)
 Missing 56 2 2 0
Note: Data are complete unless numbers of missing observations are shown. BMI, body mass index; PR, Puerto Rico.

aMenopausal status as assigned at baseline.

Models were stratified by race, menopausal status, or both at time of diagnosis or follow-up. The following covariates, measured at baseline, were included in adjusted models: menopausal status (premenopausal or postmenopausal), age at menarche (<12 y or ≥12 y), age at first birth (nulliparous, <26 y, or ≥26 y), parity (nulliparous, 1–2 children, or ≥3 children), duration of breastfeeding (<12 mo or ≥12 mo), oral contraceptive (OC) use (ever or never), hormone therapy (HT) use (ever or never), education (<high school or ≥high school), alcohol consumption (never drinker, former drinker, currently drink <1 drink/d, or currently drink ≥1 drink/d), adult body mass index (BMI) (<25 kg/m2, 25 to <30 kg/m2, or ≥30 kg/m2), family history (having one sister with breast cancer or ≥1 sister and/or a mother with breast cancer), smoking status (never smoker, former smoker, current smoker), and current region of residence (West, South, North, East). Interaction by menopausal status was tested by adding an interaction term for menopausal status and latent class into the Cox proportional hazard models stratified by race. In sensitivity analyses among postmenopausal white women only (the only subgroup with sufficient numbers for this analysis), models were stratified by ER status (ER positive or ER negative according to the clinical record) and breast cancer type (in situ or invasive).

For all analyses, results are presented only for the personal care product latent classes that included ≥20 exposed breast cancer cases.

Results

During the 304,034 person-years contributed by 46,897 black and white non-Hispanic women, 2,326 breast cancers were diagnosed (average follow-up ∼5.4 years). Characteristics of the women included in the present study population are provided in Table 1. As previously reported (Taylor et al. 2017), our final LCA model included nine beauty products (mascara, lipstick, foundation, nail polish, perfume, eye shadow, eyeliner, blush, and makeup remover), six hair products (pomade, hair straightener, conditioner, hair spray, hair gel, and shampoo), and nine skincare products (cleansing cream, antiaging cream, body lotion, hand lotion, face cream, foot cream, petroleum jelly, talcum powder applied under arms, and talcum powder applied elsewhere) (see Table S1).

Latent classes within each category of products were described (Table 2) based on the item response probabilities for frequency of use of the different products (see Figure S1). Although women within any given class are more likely than women in the other classes to have frequencies of product use as described in Table 2, all the individual responses may not fit those parameters. Our final model included three latent classes for beauty products, three for hair products, and four for skincare products, as described in detail in Table 2. Associations between product use latent classes and breast cancer are shown in Table 3. With only 165 incident breast cancer cases in black women, some latent classes lacked the 20 exposed cases we required for analysis, and those that were analyzed showed no evidence of association with breast cancer incidence.

 

Table 2. Latent class descriptions by product category and distribution of breast cancer events among white and black women at time of follow-up.
Product category/class Class descriptiona White women Black women
n (%) BC events [n (%)] n (%) BC events [n (%)]
Beauty product classes
 A. Infrequent users Likely to have infrequent use of eye shadow, eyeliner, mascara, foundation, makeup remover, nail polish, perfume, blush and lipstick. 9,208 (22) 448 (21) 1,276 (31) 53 (32)
 B. Moderate users Likely to have moderate use of eye shadow, eyeliner, mascara, foundation, makeup remover, nail polish, perfume, blush and lipstick. 15,967 (38) 816 (38) 2,153 (52) 86 (52)
 C. Frequent users Likely to have frequent use of eye shadow, eyeliner, mascara, foundation, makeup remover, nail polish, perfume, blush and lipstick. 16,720 (40) 859 (41) 744 (18) 26 (16)
 Total in analyses 41,895 2,123 4,173 165
 Missing 552 23 277 15
Hair product classes
 A. Infrequent users of hair spray Likely to have infrequent use of hair spray, hair gel, pomade and hair straightener); frequent use of shampoo, conditioner 20,896 (50) 1,059 (50) 1,006 (24) 42 (25)
 B. Moderate users of pomade, hair straightener, hair spray, and conditioner Likely to have infrequent use of shampoo, hair gel; moderate use of hair spray, conditioner, pomade, and hair straightener 1,178 (3) 61 (3) 2,984 (72) 118 (72)
 C. Frequent users of hair spray and hair gel Likely to have frequent use of hair spray, hair gel, shampoo, conditioner; infrequent use of pomade and hair straightener 19,609 (47) 997 (47) 172 (4) 5 (3)
 Total in analyses 41,683 2,117 4,162 165
 Missing 764 29 288 15
Skincare product classes
 A. Infrequent users Likely to have infrequent use of face cream, cleansing cream, antiaging cream, foot cream, body lotion and hand lotion, petroleum jelly, and talcum powderb 7,936 (19) 404 (19) 810 (19) 39 (24)
 B. Moderate users Likely to have moderate use of face cream, cleansing cream, antiaging cream, foot cream, body lotion and hand lotion; infrequent use of petroleum jelly and talcum powderb 18,572 (44) 930 (44) 2,186 (52) 88 (53)
 C. Frequent users Likely to have frequent use of face cream, cleansing cream, antiaging cream, foot cream body lotion, and hand lotion; infrequent use of petroleum jelly and talcum powderb 10,236 (25) 548 (26) 551 (13) 23 (14)
 D. Talcum powder users Likely to have frequent use of face cream, cleansing cream, antiaging cream, foot cream body lotion, and hand lotion; most frequent use of petroleum jelly, hand lotion, and talcum powderb 5,148 (12) 242 (11) 625 (15) 15 (9)
 Total in analyses 41,892 2,124 4,172 165
 Missing 555 22 278 15
Note: BC, breast cancer.

aClass labels and descriptions are based on likely item-response probabilities for each product, but all responses for individual women may not fit these parameters; each class is described relative to the other classes in each product category (Taylor et al. 2017).

bRefers to two different uses of talcum powder: talcum powder applied under arms and talcum powder applied elsewhere.

 

Table 3. Hazard ratios and 95% confidence intervals for the association between personal care product latent classes and overall breast cancer risk among white and black women.
Exposure White women Black women
Person-yearsa BC eventsa HR (95% CI)b adjHR (95% CI)c Person-yearsa BC eventsa HR (95% CI)b adjHR (95% CI)c
Beauty classes
 Infrequent user 60,484 448 1 1 7,066 53 1 1
 Moderate user 104,790 816 1.13 (1.01, 1.27) 1.13 (1.00, 1.27) 11,782 86 1.00 (0.71, 1.40) 0.95 (0.66, 1.36)
 Frequent user 111,191 859 1.12 (1.00, 1.26) 1.15 (1.02, 1.30) 4,171 26 0.85 (0.53, 1.36) 0.86 (0.53, 1.39)
 Totals 276,464 2,123 23,019 165
Hair classes
 Infrequent users of hair spray 137,672 1,059 1 1 5,615 42 1 1
 Moderate users of pomade, hair straightener, hair spray, and conditioner 7,686 61 0.91 (0.70, 1.18) 0.91 (0.70, 1.19) 16,333 118 0.93 (0.65, 1.32) 0.90 (0.63, 1.28)
 Frequent users of hair spray and hair gel 129,205 997 1.01 (0.93, 1.10) 1.02 (0.93, 1.11) 998 5
 Totals 274,563 2,117 22,946 165
Skincare classes
 Infrequent user 52,506 404 1 1 4,311 39 1 1
 Moderate user 123,010 930 1.00 (0.89, 1.12) 1.03 (0.91, 1.17) 12,124 88 0.76 (0.52, 1.11) 0.75 (0.51, 1.10)
 Frequent user 66,346 548 1.11(0.98, 1.27) 1.13 (1.00, 1.29) 3,032 23 0.82 (0.49, 1.38) 0.79 (0.47, 1.34)
 Talcum powder user 34,582 242 0.90 (0.77, 1.06) 0.92 (0.78, 1.08) 3,542 15
 Totals 276,444 2,124 23,009 165
Note: —, <20 cases; adjHR, adjusted hazard ratio; BC, breast cancer; CI, confidence interval; HR, hazard ratio.

aNumbers of person-years and BC events are for women with complete data for each product class only; after accounting for missing data, total numbers of person-years and BC events were 279,699 and 2,146, respectively, for white women and 24,336 and 180, respectively, for black women; results were not reported if <20 BC events.

bModels accounted for age by using age as the time scale, where participants entered the analysis at their enrollment age (left-truncation) and accrued person-time until they exited at their cancer diagnosis or were administratively censored at their age at last follow-up.

cIn addition to the adjustments described in b, models were adjusted for baseline menopausal status, parity, age at first live birth, duration of breastfeeding, adult body mass index, alcohol use, oral contraceptive use, hormone therapy use, education, family history, region of residence, age at menarche, and smoking status.

Among white women, “moderate” and “frequent” users of beauty products had increased risk of breast cancer relative to “infrequent” users [moderate users, adjHR=1.13 (95% CI: 1.00, 1.27) and frequent users, adjHR=1.15 (95% CI: 1.02, 1.30)] (Table 3). Similarly, among white women, frequent users of skincare products had increased risk of breast cancer relative to infrequent users [adjHR=1.13 (95% CI: 1.00, 1.29)]. Patterns of hair product use were not associated with breast cancer incidence.

In analyses stratified by menopausal status (conducted in white women only), HRs for breast cancer associated with the frequent (compared with the infrequent) users of beauty or skincare products were higher among postmenopausal women [adjHR=1.18 (95% CI: 1.14, 1.21) and 1.12 (95% CI: 1.09, 1.16), respectively] than premenopausal women [adjHR=1.01 (95% CI: 0.76, 1.33) and 1.06 (95% CI: 0.79, 1.42), respectively] but were not statistically different (p-interaction 0.33 and 0.65, respectively) (Table 4).

 

Table 4. Hazard ratios and 95% confidence intervals for the association between latent classes and breast cancer risk among postmenopausal and premenopausal white women at time of breast cancer diagnosis.
Exposure Premenopausal Postmenopausal Interaction p-Valued
Person-years BC eventsa HR (95% CI)b adjHR (95% CI)c Person-years BC eventsa HR (95% CI)b adjHR (95% CI)c
Beauty Classes
 Infrequent user 6,834 74 1 1 53,247 372 1 1
 Moderate user 19,586 177 0.99 (0.75, 1.30) 0.98 (0.74, 1.30) 84,289 631 1.13 (1.10, 1.16) 1.16 (1.12, 1.19) 0.26
 Frequent user 19,861 191 1.01 (0.77, 1.33) 1.01 (0.76, 1.33) 90,181 656 1.14 (1.10, 1.17) 1.18 (1.14, 1.21) 0.33
 Totalse 46,281 442 227,717 1,659
Hair Classes
 Infrequent users of hair spray 21,377 208 1 1 115,014 841 1 1
 Moderate users of hair spray, conditioner, pomade and hair straightener 443 4 7,190 56 0.76 (0.71, 0.81) 0.76 (0.71, 0.81) 0.46
 Frequent users of hair spray and hair gel 24,277 229 1.07 (0.89, 1.29) 1.04 (0.86, 1.26) 103,825 757 1.01 (0.99, 1.03) 1.03 (1.01, 1.06) 0.84
 Totalse 46,097 441 226,029 1,654
Skincare Classes
 Infrequent user 8,266 76 1 1 43,793 324 1 1
 Moderate user 20,031 200 1.09 (0.84, 1.42) 1.06 (0.81, 1.39) 101,807 721 1.05 (1.02, 1.08) 1.08 (1.05, 1.12) 0.51
 Frequent user 13,152 122 1.07 (0.80, 1.42) 1.06 (0.79, 1.42) 52,621 420 1.11 (1.07, 1.14) 1.12 (1.09, 1.16) 0.65
 Talcum powder user 4,853 44 0.88 (0.61, 1.29) 0.84 (0.57, 1.23) 29,453 195 0.99 (0.95, 1.03) 1.03 (0.99, 1.07) 0.91
 Totalse 46,302 442 227,674 1,660
Note: adjHR, adjusted hazard ratio; BC, breast cancer; CI, confidence interval; HR, hazard ratio.

aTotal numbers of BC events and person-years are for women with complete data for each product class; after accounting for missing data, among women with known menopausal status, total numbers of person-years and BC events were 47,166 and 445, respectively, for premenopausal white women and 230,023 and 1,679, respectively, for postmenopausal white women; results were not reported if <20 BC events.

bModels accounted for age by using age as the time scale, where participants entered the analysis at their enrollment age (left-truncation) and accrued person-time until they exited at their cancer diagnosis or were administratively censored at their age at last follow-up.

cIn addition to the adjustments described in b, models were adjusted for baseline menopausal status, parity, age at first live birth, duration of breastfeeding, adult body mass index, alcohol use, oral contraceptive use, hormone therapy use, education, family history, region of residence, age at menarche, and smoking status. Results were not reported if <20 BC events.

dp-Value for interaction by menopausal status, derived by adding interaction term to adjusted Cox proportional hazard model for pre- and postmenopausal white women.

eWomen with unknown or missing menopausal status at follow-up (n=346) were excluded; n=22 breast cancer cases were premenopausal at baseline and had an unknown menopausal status at follow-up.

In exploratory analyses, ER status was available for 85% (n=1,420) of white postmenopausal women with a breast cancer diagnosis (the only subgroup with sufficient numbers for this analysis). In this group (see Table S2), the association between breast cancer and moderate and frequent users (compared to infrequent users) of beauty products did not appear to differ substantially between ER+ [adjHR=1.05 (95% CI: 0.90, 1.23) and adjHR=1.10 (95% CI: 0.94, 1.28), respectively] and ER− [adjHR=1.03 (95% CI: 0.69, 1.54) and adjHR=0.72 (95% CI: 0.47, 1.10), respectively]. However, because the overall HRs for women with known ER status are not consistent with the stratum-specific hazard ratios, it is unlikely that women were missing ER status at random, so the HRs presented may be biased. When white postmenopausal women were stratified by breast cancer type, in situ (n=326) and invasive (n=1,091) (see Table S3), and compared with infrequent use, moderate and frequent use of beauty products were associated with higher risk of in situ breast cancer [adjHR=1.41 (95% CI: 1.06, 1.89) and adjHR=1.38 (95% CI: 1.03, 1.85), respectively] but not with invasive breast cancer [adjHR=1.09 (95% CI: 0.94, 1.27) and adjHR=1.13 (95% CI: 0.97, 1.32), respectively].

Discussion

Our findings from this large, prospective study with detailed self-report of personal care product use suggest that for non-Hispanic white women (the majority of the cohort), the risk of breast cancer was 10–15% higher among those classified as moderate and frequent users of beauty products than among women classified as infrequent users of beauty products. When stratified by menopausal status, associations with beauty product use appeared to be limited to postmenopausal white women because all HRs for premenopausal women were close to the null. Frequent users of beauty products can be broadly categorized as women who report using a combination of beauty products on a weekly basis (e.g., mascara, foundation, and lipstick). Moderate users were more likely to report using these same products at least monthly or up to several times a month. The relative risk of breast cancer among non-Hispanic white women classified as frequent users of skincare products (likely to use cleansing cream, antiaging cream, body lotion, hand lotion, face cream, and foot cream at least weekly, but unlikely to use talcum powder) was approximately 13% higher (95% CI: 0, 29%) than among women classified as infrequent users of skincare products.

The hypothesis that personal care products are associated with increased breast cancer risk is primarily based on animal and laboratory studies. In these settings, chemicals found in a wide variety of personal care products (e.g., parabens and phthalates) mimic estrogens (Davis et al. 1993), alter hormonal signaling, affect developing reproductive systems (Colborn et al. 1993), disrupt normal mammary development (Macon 2013), or provide a combination of any or all of these effects. The association between beauty latent classes and breast cancer risk appeared stronger among postmenopausal women than among premenopausal women. This finding is consistent with the hypothesis that weak estrogenic effects might have a greater impact during the postmenopausal period because women with lower endogenous estrogen levels are more susceptible to exogenous estrogenic exposures.

It is important to acknowledge that the frequency of use of different beauty products may not be a true risk factor for breast cancer but may instead be a proxy marker for other breast cancer risk factors. Similarly, the frequency of use of different personal care products might be associated with the likelihood of screenings and mammograms, and thus the likelihood that a woman would be diagnosed with a carcinoma in situ.

Personal care product exposure is difficult to characterize because each product is a complex mixture, and multiple products are often used in combination by one person. Co-occurring exposures may have additive or interacting effects or may result in confounding. For example, a chemical that does not show estrogenic activity could be a marker for other chemicals that are estrogenic. Products that include chemicals that can be estrogenic may show either estrogenic or antiestrogenic effects in specific tissues (Myers et al. 2015). We used LCA as an innovative method for characterizing exposure to mixtures. As mentioned in previous work (Taylor et al. 2017), relative to individual product use questions, latent class variables capture complex patterns of personal care product usage and have been used to capture complex exposures in a variety of research settings (Lanza et al. 2010; Lanza and Rhoades 2013). LCA also addresses many limitations related to mixtures. For example, latent classes can be used to describe the variability among multiple correlated and observed exposures. However, although this approach provides insights about patterns of use, it does not allow us to address the individual and combined effects of separate agents. Previous studies have examined correlation structure between specific personal care products (Biesterbos et al. 2013; Manová et al. 2013; Wu et al. 2010), but these studies did not evaluate associations between personal care product use and health outcomes.

Our study was not able to assess breast cancer risk associated with specific chemicals. The questionnaire did not capture information on specific brands of personal care products or on the individual components of these products, nor did it capture potential changes in brand preference over time. In a previous study, brand loyalty varied greatly by product type; there was less loyalty for antiaging products, antibacterial liquid soap, and hair mousse, and more loyalty for contact lens solution and lip balm (Wu et al. 2010). However, that study population was limited to 604 households in northern California and may not be generalizable to the U.S. population. Additionally, even if product brand information were available from our study population, manufacturers are not required to disclose all chemical ingredients in consumer products (EWG 2012). Chemical compositions of products change over time and across batches, and chemicals [e.g., bisphenol A (BPA)] can leach from containers into the product (Yang et al. 2011). In addition to the challenges involved in collecting brand information, exposure may vary depending on how products are used or applied. Thus, it was not feasible for us to ascertain exposures to individual chemicals from the questionnaire data.

Our study addresses the idea that combinations and patterns of exposure may be particularly important in relation to risk. The large sample size of white women, detailed self-report of personal care product use, prospective identification of breast cancer, multivariable analysis, and inclusion of both the aggregated and individual exposure data strengthen our study. Women provided information on personal care products used during the twelve months before enrollment. We do not know how constant women’s exposures are over time, or if product use at baseline would capture use during etiologically relevant time windows of exposure, which may have occurred years in the past. Latency periods of 8 to 15 y have been reported for breast cancer (Aschengrau et al. 1998; Brody et al. 2007; Lewis-Michl et al. 1996; Petralia et al. 1999), and empirical induction periods could be at least one to two decades (Brody et al. 2007). However, if the exposure acts primarily on tumor survival and growth, more recent exposures may be relevant. Additional research is needed to investigate the stability of personal care product use within different populations of women and to identify relevant windows of exposure.

Our study has other limitations. First, we had limited power to examine associations among black women. As previously reported (Taylor et al. 2017), we observed that latent class distribution differed by race. We observed race-related patterns of hair product use: The class that was characterized by moderate use of pomade, hair straightener, hair spray, and conditioner contained the majority of black women, but only 3% of white women (Aschengrau et al. 1998; Brody et al. 2007; Lewis-Michl et al. 1996; Petralia et al. 1999). The associations we report between breast cancer and personal care product use are modest in magnitude and do not show a dose–response relationship. Finally, although our study was motivated by the literature on endocrine disruptors, the results may have other interpretations. The finding that the associations tended to be stronger for in situ than for invasive breast cancer raises the concern that the latent classes might be confounded to some extent by cancer-screening behavior. Some in situ lesions never progress to invasive cancer (Kerlikowske et al. 2010), and if frequent beauty and skincare product users are screened more often, such lesions may be over-represented. If so, the observed associations may have little causal relationship. Even if screening frequency is not an issue, given the relatively modest hazard ratios and the opportunity for exposure misclassification, further epidemiologic and mechanistic studies would be needed to infer causality.

Conclusion

The results from this study generate novel hypotheses concerning the relationship between the use of personal care products and risk of breast cancer. Evidence that breast cancer may be associated with moderate or frequent beauty product use or with frequent skincare product use may indicate effects of chemicals used in the products, although noncausal associations resulting from confounding by correlated behaviors and conditions are also possible. Future work should also address duration of exposure and how product use patterns vary over time.

Acknowledgments

This research was supported in part by Intramural Research Programs of the Division of the National Toxicology Program and of the NIH, National Institute of Environmental Health Sciences (Z01 ES 044005), and the Center for Environmental Health and Susceptibility (P30-ES010126).

References

Aschengrau A, Paulu C, Ozonoff D. 1998. Tetrachloroethylene-contaminated drinking water and the risk of breast cancer. Environ Health Perspect 106 (suppl 4):947–953, PMID: 9703477, 10.1289/ehp.98106s4947.

Biesterbos JWH, Dudzina T, Delmaar CJE, Bakker MI, Russel FGM, von Goetz N, et al. 2013. Usage patterns of personal care products: important factors for exposure assessment. Food Chem Toxicol 55:8–17, PMID: 23174517, 10.1016/j.fct.2012.11.014.

Braun JM, Just AC, Williams PL, Smith KW, Calafat AM, Hauser R. 2014. Personal care product use and urinary phthalate metabolite and paraben concentrations during pregnancy among women from a fertility clinic. J Expo Sci Environ Epidemiol 24(5):459–466, PMID: 24149971, 10.1038/jes.2013.69.

Bray BC, Lanza ST, Tan Z. 2012. “An introduction to eliminating bias in classify-analyze approaches for latent class analysis.” Technical Report Series #12-118. State College, PA:The Methodology Center, Pennsylvania State University. https://methodology.psu.edu/media/techreports/12-118.pdf [accessed 19 December 2014].

Brody JG, Moysich KB, Humblet O, Attfield KR, Beehler GP, Rudel RA. 2007. Environmental pollutants and breast cancer: epidemiologic studies. Cancer 109(12 suppl):2667–2711, PMID: 17503436, 10.1002/cncr.22655.

CDC (Centers for Disease Control and Prevention). 2012. “Fourth national report on human exposure to environmental chemicals.” Atlanta, GA:Centers for Disease Control and Prevention, Department of Health and Human Services. http://www.cdc.gov/ExposureReport/pdf/FourthReport.pdf [accessed 11 December 2014].

Chen WY. 2008. Exogenous and endogenous hormones and breast cancer. Best Pract Res Clin Endocrinol Metab 22(4):573–585, PMID: 18971119, 10.1016/j.beem.2008.08.001.

Colborn T, vom Saal FS, Soto AM. 1993. Developmental effects of endocrine-disrupting chemicals in wildlife and humans. Environ Health Perspect 101(5):378–384, PMID: 8080506, 10.1289/ehp.93101378.

Davis D, Bradlow H, Wolff M, Woodruff T, Hoel D, Anton-Culver H. 1993. Medical hypothesis: xenoestrogens as preventable causes of breast cancer. Environ Health Perspect 101(5):372–377, PMID: 8119245, 10.1289/ehp.93101372.

Dean N, Raftery AE. 2010. Latent class analysis variable selection. Ann Inst Stat Math 62(1):11–35, PMID: 20827439, 10.1007/s10463-009-0258-9.

DeSantis CE, Fedewa SA, Goding Sauer A, Kramer JL, Smith RA, Jemal A. 2016. Breast cancer statistics, 2015: convergence of incidence rates between black and white women. CA Cancer J Clin 66(1):31–42, PMID: 26513636, 10.3322/caac.21320.

Dodson RE, Nishioka M, Standley LJ, Perovich LJ, Brody JG, Rudel RA. 2012. Endocrine disruptors and asthma-associated chemicals in consumer products. Environ Health Perspect 120(7):935–943, PMID: 22398195, 10.1289/ehp.1104052.

EWG (Environmental Working Group). 2004. Exposures Add Up—Survey Results. http://www.ewg.org/skindeep/2004/06/15/exposures-add-up-survey-results/.

EWG. 2012. EWG’s Skin Deep cosmetics database. http://www.ewg.org/skindeep/ [accessed 5 November 2012].

Fakri S, Al-Azzawi A, Al-Tawil N. 2006. Antiperspirant use as a risk factor for breast cancer in Iraq. East Mediterr Health J 12(3–4):478–482, PMID: 17037719.

James-Todd T, Terry MB, Rich-Edwards J, Deierlein A, Senie R. 2011. Childhood hair product use and earlier age at menarche in a racially diverse study population: a pilot study. Ann Epidemiol 21(6):461–465, PMID: 21421329, 10.1016/j.annepidem.2011.01.009.

Kerlikowske K, Molinaro AM, Gauthier ML, Berman HK, Waldman F, Bennington J, et al. 2010. Biomarker expression and risk of subsequent tumors after initial ductal carcinoma in situ diagnosis. J Natl Cancer Inst 102(9):627–637, PMID: 20427430, 10.1093/jnci/djq101.

Koo HJ, Lee BM. 2004. Estimated exposure to phthalates in cosmetics and risk assessment. J Toxicol Environ Health A 67:1901–1914, PMID: 15513891, 10.1080/15287390490513300.

Lanza ST, Collins LM, Lemmon DR, Schafer JL. 2007. Proc LCA: a SAS procedure for latent class analysis. Struct Equ Modeling 14(4):671–694, PMID: 19953201, 10.1080/10705510701575602.

Lanza ST, Dziak JJ, Huang L, Wagner A, Collins LM. 2015. PROC LCA & PROC LTA Users’ Guide (version 1.3.2). University Park, PA:The Methodology Center, Pennsylvania State University. https://methodology.psu.edu/sites/default/files/software/proclcalta/proc_lca_lta_1-3-2-1_users_guide.pdf [accessed 27 September 2015].

Lanza ST, Rhoades BL. 2013. Latent class analysis: an alternative perspective on subgroup analysis in prevention and treatment. Prev Sci 14(2):157–168, PMID: 21318625, 10.1007/s11121-011-0201-1.

Lanza ST, Savage JS, Birch LL. 2010. Identification and prediction of latent classes of weight-loss strategies among women. Obesity (Silver Spring) 18(4):833–840, PMID: 19696754, 10.1038/oby.2009.275.

Lewis-Michl EL, Melius JM, Kallenbach LR, Ju CL, Talbot TO, Orr MF, et al. 1996. Breast cancer risk and residence near industry or traffic in Nassau and Suffolk Counties, Long Island, New York. Arch Environ Health 51(4):255, PMID: 8757405, 10.1080/00039896.1996.9936024.

López-Carrillo L, Hernández-Ramírez RU, Calafat AM, Torres-Sánchez L, Galván-Portillo M, Needham LL, et al. 2010. Exposure to phthalates and breast cancer risk in northern Mexico. Environ Health Perspect 118(4):539–544, PMID: 20368132, 10.1289/ehp.0901091.

Macon MB. 2013. Endocrine disruptors and the breast: Early life effects and later life disease. J Mammary Gland Biol Neoplasia. 18(1):43–61, PMID: 23417729, 10.1007/s10911-013-9275-7.

Manová E, von Goetz N, Hauri U, Bogdal C, Hungerbühler K. 2013. Organic UV filters in personal care products in Switzerland: a survey of occurrence and concentrations. Int J Hyg Environ Health 216(4):508–514, PMID: 23026542, 10.1016/j.ijheh.2012.08.003.

McGrath KG. 2003. An earlier age of breast cancer diagnosis related to more frequent use of antiperspirants/deodorants and underarm shaving. Eur J Cancer Prev 12(6):479–485, PMID: 14639125, 10.1097/01.cej.0000103462.62592.c6.

Meeker JD, Cantonwine DE, Rivera-González LO, Ferguson KK, Mukherjee B, Calafat AM, et al. 2013. Distribution, variability, and predictors of urinary concentrations of phenols and parabens among pregnant women in Puerto Rico. Environ Sci Technol 47(7):3439–3447, PMID: 23469879, 10.1021/es400510g.

Mirick DK, Davis S, Thomas DB. 2002. Antiperspirant use and the risk of breast cancer. J Natl Cancer Inst 94(20):1578–1580, PMID: 12381712, 10.1093/jnci/94.20.1578.

Morgan JW, Gladson JE, Rau KS. 1998. Position paper of the American council on science and health on risk factors for breast cancer: established, speculated, and unsupported. Breast Journal 4(3):177–197, 10.1046/j.1524-4741.1998.430177.x.

Myers SL, Yang CZ, Bittner GD, Witt KL, Tice RR, Baird DD. 2015. Estrogenic and anti-estrogenic activity of off-the-shelf hair and skin care products. J Expo Sci Environ Epidemiol 25(3):271–277, PMID: 24849798, 10.1038/jes.2014.32.

Nichols HB, Baird DD, DeRoo LA, Kissling GE, Sandler DP. 2013. Tubal ligation in relation to menopausal symptoms and breast cancer risk. Br J Cancer 109(5):1291–1295, PMID: 23922107, 10.1038/bjc.2013.433.

NIEHS (National Institute of Environmental Health Sciences). 2010. The Sister Study. https://sisterstudy.niehs.nih.gov/English/index1.htm [accessed 1 January 2013].

Petralia SA, Vena JE, Freudenheim JL, Dosemeci M, Michalek A, Goldberg MS, et al. 1999. Risk of premenopausal breast cancer in association with occupational exposure to polycyclic aromatic hydrocarbons and benzene. Scand J Work Environ Health 25(3):215–221, PMID: 10450771, 10.5271/sjweh.426.

Rollison DE, Helzlsouer KJ, Pinney SM. 2006. Personal hair dye use and cancer: a systematic literature review and evaluation of exposure assessment in studies published since 1992. J Toxicol Environ Health B Crit Rev 9(5):413–439, PMID: 17492526, 10.1080/10937400600681455.

Romero-Franco M, Hernández-Ramírez RU, Calafat AM, Cebrián ME, Needham LL, Teitelbaum S, et al. 2011. Personal care product use and urinary levels of phthalate metabolites in Mexican women. Environ Int 37(5):867–871, PMID: 21429583, 10.1016/j.envint.2011.02.014.

Shanle EK, Xu W. 2011. Endocrine disrupting chemicals targeting estrogen receptor signaling: identification and mechanisms of action. Chem Res Toxicol 24(1):6–19, PMID: 21053929, 10.1021/tx100231n.

Silva MJ, Barr DB, Reidy JA, Malek NA, Hodge CC, Caudill SP, et al. 2004. Urinary levels of seven phthalate metabolites in the U.S. Population from the national health and nutrition examination survey (NHANES) 1999-2000. Environ Health Perspect 112(5):331–338, PMID: 14998749, 10.1289/ehp.6723.

Takkouche B, Etminan M, Montes-Martínez A. 2005. Personal use of hair dyes and risk of cancer: a meta-analysis. JAMA 293(20):2516–2525, PMID: 15914752, 10.1001/jama.293.20.2516.

Takkouche B, Regueira-Méndez C, Montes-Martínez A. 2009. Risk of cancer among hairdressers and related workers: a meta-analysis. Int J Epidemiol 38(6):1512–1531, PMID: 19755396, 10.1093/ije/dyp283.

Taylor KW, Baird DD, Herring AH, Engel LS, Nichols HB, Sandler DP, et al. 2017. Associations among personal care product use patterns and exogenous hormone use in the NIEHS Sister Study. J Expo Sci Environ Epidemiol 27(5):458–464, PMID: 28120835, 10.1038/jes.2016.82.

Tiwary CM. 1998. Premature sexual development in children following the use of estrogen- or placenta-containing hair products. Clin Pediatr (Phila) 37(12):733–739, PMID: 9864648, 10.1177/000992289803701204.

Tiwary CM, Ward JA. 2003. Use of hair products containing hormone or placenta by US military personnel. J Pediatr Endocrinol Metab 16(7):1025–1032, PMID: 14513880, 10.1515/JPEM.2003.16.7.1025.

Wu X, Bennett DH, Ritz B, Cassady DL, Lee K, Hertz-Picciotto I. 2010. Usage pattern of personal care products in California households. Food Chem Toxicol 48(11):3109–3119, PMID: 20696198, 10.1016/j.fct.2010.08.004.

Yang CZ, Yaniger SI, Jordan VC, Klein DJ, Bittner GD. 2011. Most plastic products release estrogenic chemicals: a potential health problem that can be solved. Environ Health Perspect 119(7):989–996, PMID: 21367689, 10.1289/ehp.1003220.

Erratum: “Estimated Changes in Life Expectancy and Adult Mortality Resulting from Declining PM2.5 Exposures in the Contiguous United States: 1980–2010”

PDF icon PDF Version (668 KB)

  • Received: 29 January 2018
    Accepted: 29 January 2018
    Published: 20 February 2018

    Address correspondence to N. Fann, U.S. Environmental Protection Agency, 109 T.W. Alexander Dr., Mail Drop C539-07; Research Triangle Park, NC 27711 USA. Telephone: (919) 541-0209. Email: Fann.neal@epa.gov

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

In Figure 1, the key to the color gradations (“Percentage of deaths attributable to PM2.5”) was omitted. It is included in the figure reproduced here.

Photograph of a woman replacing a dust collection filter

Figure 1 The fraction of total all-cause deaths attributable to PM2.5 in U.S. counties in the years 1980, 1990, 2000, and 2010 among adults ages 30 and older (calculated using risk coefficient from Krewski et al. 2009). State and county boundaries for each year drawn according to Census Topologically Integrated Geographic Encoding and Referencing (TIGER)/Line files as reported by the Minnesota Population Center National Historical Geographic Information System (NHGIS Database) (http://www.nhgis.org).

EHP regrets the error.

What’s in the Mix? Improving Risk Assessment of Food Contact Materials

PDF icon PDF Version (611 KB)

  • Published: 20 February 2018

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

The steps involved in bringing food to grocery store shelves present a ubiquitous yet poorly understood route of exposure to chemicals of which most people are unaware.1 Packaging, storage tanks, machinery, transportation containers—these and other critical components of food production have the potential to leach chemicals into the food itself. Researchers have limited knowledge of the breadth and relative safety of all materials that may come into contact with food during processing. This lack of information impairs their ability to assess risk and to inform public policy, according to the authors of a commentary in Environmental Health Perspectives.2

Photograph of tortellini on a food production line
Sometimes there’s more to food than just, well, the food itself—materials that come into contact with food as it is being processed, packaged, and transported have the potential to leach chemicals into the product. Image: © RomboStudio/Shutterstock.

In addition to their constituent chemicals, food contact materials may also contain complex mixtures of nonintentionally added substances (NIAS), such as reaction by-products and impurities.3,4,5 In one 2007 study, chemical analysis of plastic samples from food contact materials turned up unidentified compounds that could not have been predicted from the known composition of the samples.5 According to the commentary authors, led by Martin Scheringer of the Research Centre for Toxic Compounds in the Environment at Masaryk University, Czech Republic, and Jane Muncke, manager of the nonprofit Food Packaging Forum, these findings indicate that “comprehensive qualitative and quantitative chemical analyses of plastic [articles that come into contact with food] are currently impossible.”2

There is evidence that plastic products such as food wraps, bags, clamshell containers, and baby and water bottles leach chemicals with endocrine-active properties that potentially pose health risks even at very low levels.6 However, this consideration has not made its way into risk assessment as far as food contact materials are concerned.7,8

To improve risk assessment, the authors recommend evaluating potential low-dose effects (common with endocrine disrupters) for all of what they dub food contact chemicals, or any chemical that is either used in the manufacture of food contact materials or otherwise present in finished food contact materials. Notably, they also recommend performing toxicological assessments of these finished materials—in other words, the complete mixture of substances used to produce the material as well as any NIAS that may be present.

The latter recommendation would be a significant change to current practices in the United States and Europe for two reasons, Muncke says. First, it takes into account the idea that mixtures, not just single substances, can migrate into food. Second, it shifts the chemical risk assessment from the beginning of the manufacturing process to the final stage—“better reflecting the reality of what comes into contact with food and what people are exposed to,” she says.

Mark Maier, formerly a staff toxicologist for Valspar Corporation, led that company’s efforts to identify a nonestrogenic replacement for bisphenol A (BPA)9 in its coatings for food and beverage cans. A January 2017 study coauthored by Maier, who continues to consult for Valspar and was not affiliated with the new commentary, found no evidence of estrogenic activity by the company’s replacement epoxy monomer, tetramethyl bisphenol F (TMBPF).10

“The way I look at it, Valspar is trying so hard to do the right thing,” he says. “They’re trying to get the data that’s called for in this paper. As far as their willingness to make data transparent and make data available, I think Valspar stands alone in that regard. But gosh, I hope that can change.”

However, Maier disputes the authors’ recommendation that researchers and regulators ensure adequate toxicological assessment of all food contact chemicals. “I would frame the recommendation differently,” Maier says. “If you [attempt to] test everything for every possible problem, you have to evaluate every packaging chemical and mixture to the nth degree. That makes no sense for chemicals with such low exposures.” Instead, he says, researchers should focus on plausible effects for relevant classes of chemicals at relevant exposure levels. He also believes there is no reason to keep retesting materials such as polyesters and acrylics “just to show what we already know.” He explains, “You can’t test everything; you have to test for things that make sense. You will never get all of them.”

That said, the U.S. Food and Drug Administration’s “generally recognized as safe” (GRAS) designation is a loophole that has allowed many unknown and potentially unsafe chemicals into foods over the last 60 years, according to a 2014 report11 from the nonprofit Natural Resources Defense Council (NRDC). A general lack of transparency pervades the regulatory environment for food contact materials, says commentary coauthor Maricel Maffini, an independent consultant on issues related to environmental health, chemical safety, and science policy.

“In many cases, some of those chemicals were approved decades ago with likely almost no toxicity data,” says Maffini, who also coauthored the NRDC and TBMPF studies. Even today, she says, “if a company claims that exposure will be below a certain level, the company may not be required to provide any toxicity info. The amount of data that’s available is very limited.”

Nate Seltenrich covers science and the environment from Petaluma, California. His work has appeared in High Country News, Sierra, Yale Environment 360, Earth Island Journal, and other regional and national publications.

References

1. Seltenrich N. 2015. A hard nut to crack: reducing chemical migration in food-contact materials. Environ Health Perspect 123(7):A174–A179, PMID: 26133041, 10.1289/ehp.123-A174.

2. Muncke J, Backhaus T, Geueke B, Maffini MV, Martin OV, Myers JP, et al. 2017. Scientific challenges in the risk assessment of food contact materials. Environ Health Perspect 125(9):095001, PMID: 28893723, 10.1289/EHP644.

3. Hoppe M, de Voogt P, Franz R. 2016. Identification and quantification of oligomers as potential migrants in plastics food contact materials with a focus in polycondensates – a review. Trends Food Sci Technol 50:118–130, 10.1016/j.tifs.2016.01.018.

4. Nerin C, Alfaro P, Aznar M, Domeño C. 2013. The challenge of identifying non-intentionally added substances from food packaging materials: a review. Anal Chim Acta 775:14–24, PMID: 23601971, 10.1016/j.aca.2013.02.028.

5. Bradley E, Coulier L. 2007. An Investigation into the Reaction and Breakdown Products from Starting Substances Used to Produce Food Contact Plastics. London, UK:Food Standards Agency Central Science Laboratory.

6. Yang CZ, Yaniger SI, Jordan VC, Klein DJ, Bittner GD. 2011. Most plastic products release estrogenic chemicals: a potential health problem that can be solved. Environ Health Perspect 119(7):989–996, PMID: 21367689, 10.1289/ehp.1003220.

7. Muncke J. 2009. Exposure to endocrine disrupting compounds via the food chain: is packaging a relevant source? Sci Total Environ 407(16):4549–4559, PMID: 19482336, 10.1016/j.scitotenv.2009.05.006.

8. Geueke B, Wagner CC, Muncke J. 2014. Food contact substances and chemicals of concern: a comparison of inventories. Food Addit Contam Part A Chem Anal Control Expo Risk Assess 31(8):1438–1450, PMID: 24999917, 10.1080/19440049.2014.931600.

9. FDA (U.S. Food and Drug Administration). 2016. Bisphenol A (BPA): Use in Food Contact Application. https://www.fda.gov/Food/IngredientsPackagingLabeling/FoodAdditivesIngredients/ucm064437.htm [accessed 28 July 2017].

10. Soto AM, Schaeberle C, Maier MS, Sonnenschein C, Maffini MV. 2017. Evidence of absence: estrogenicity assessment of a new food-contact coating and the bisphenol used in its synthesis. Environ Sci Technol 51(3):1718–1726, PMID: 28098991, 10.1021/acs.est.6b04704.

11. Neltner T, Maffini M. 2014. “Generally Recognized as Secret: Chemicals added to food in the United States.” R:14-03-A. New York, NY:Natural Resources Defense Council. https://www.nrdc.org/sites/default/files/safety-loophole-for-chemicals-in-food-report.pdf [accessed 28 July 2017].

Erratum: “Traditional Ecological Knowledge: A Different Perspective on Environmental Health”

PDF icon PDF Version (37 KB)

  • Received: 25 January 2018
    Accepted: 26 January 2018
    Published: 16 February 2018

    Address correspondence to S.M. Booker, Environmental Health Perspectives, P.O. Box 12233, Research Triangle Park, NC 27709. Email: booker@niehs.nih.gov

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

Dorothy Castille was incorrectly identified as an investigator at the National Institute on Minority Health and Health Disparities (NIMHD). Her correct title at NIMHD is health scientist administrator.

EHP regrets the error.

The Association of Long-Term Exposure to Particulate Matter Air Pollution with Brain MRI Findings: The ARIC Study

Author Affiliations open

1Milken Institute School of Public Health, George Washington University, Washington, DC, USA

2Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, USA

3RTI International, Research Triangle Park, North Carolina, USA

4School of Medicine, Pennsylvania State University, Hershey, Pennsylvania, USA

5School of Public Health, Texas A&M Health Science Center, College Station, Texas, USA

6Mayo Clinic, Rochester, Minnesota, USA

7Johns Hopkins University, Baltimore, Maryland, USA

8University of Mississippi Medical Center, Jackson, Mississippi, USA

9Carolina Population Center, University of North Carolina, Chapel Hill, North Carolina, USA

10School of Medicine, University of North Carolina, Chapel Hill, North Carolina, USA

PDF icon PDF Version (165 KB)

  • Background:
    Increasing evidence links higher particulate matter (PM) air pollution exposure to late-life cognitive impairment. However, few studies have considered associations between direct estimates of long-term past exposures and brain MRI findings indicative of neurodegeneration or cerebrovascular disease.
    Objective:
    Our objective was to quantify the association between brain MRI findings and PM exposures approximately 5 to 20 y prior to MRI in the Atherosclerosis Risk in Communities (ARIC) study.
    Methods:
    ARIC is based in four U.S. sites: Washington County, Maryland; Minneapolis suburbs, Minnesota; Forsyth County, North Carolina; and Jackson, Mississippi. A subset of ARIC participants underwent 3T brain MRI in 2011–2013 (n=1,753). We estimated mean exposures to PM with an aerodynamic diameter less than 10 or 2.5 μm (PM10 and PM2.5) in 1990–1998, 1999–2007, and 1990–2007 at the residential addresses of eligible participants with MRI data. We estimated site-specific associations between PM and brain MRI findings and used random-effect, inverse variance–weighted meta-analysis to combine them.
    Results:
    In pooled analyses, higher mean PM2.5 and PM10 exposure in all time periods were associated with smaller deep-gray brain volumes, but not other MRI markers. Higher PM2.5 exposures were consistently associated with smaller total and regional brain volumes in Minnesota, but not elsewhere.
    Conclusions:
    Long-term past PM exposure in was not associated with markers of cerebrovascular disease. Higher long-term past PM exposures were associated with smaller deep-gray volumes overall, and higher PM2.5 exposures were associated with smaller brain volumes in the Minnesota site. Further work is needed to understand the sources of heterogeneity across sites. https://doi.org/10.1289/EHP2152
  • Received: 5 May 2017
    Revised: 9 January 2018
    Accepted: 10 January 2018
    Published: 16 February 2018

    Address correspondence to M.C. Power, Department of Epidemiology and Biostatistics, George Washington University, Milken Institute School of Public Health, 950 New Hampshire Avenue NW, 5th Floor, Washington, DC 20052 USA. Telephone: (202) 994-7778. Email: melindacpower@gmail.com

    Supplemental Material is available online (https://doi.org/10.1289/EHP2152).

    R.F.G. is Associate Editor for Neurology. C.R.J. serves on a scientific advisory board for Eli Lilly & Company and receives research support from the NIH/NIA (R01-AG011378, U01-HL096917, U01-AG024904, RO1 AG041851, R01 AG37551, R01AG043392, U01-AG06786) and the Alexander Family Alzheimer’s Disease Research Professorship of the Mayo Foundation. All other authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (271 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (26 KB)



Introduction

Common environmental pollutants may promote cognitive decline, cognitive impairment, and dementia. In particular, recent epidemiologic studies have reported that higher exposure to particulate air pollution is associated with increased risk of cognitive decline, cognitive impairment, and dementia (Power et al. 2016a; Tzivian et al. 2016; Xu et al. 2016). Although this body of work is highly suggestive, work linking air pollution to MRI markers of brain injury may provide mechanistic insight and would allay concerns about residual confounding by sociodemographic and socioeconomic characteristics that are common to studies of air pollution and cognition (Casanova et al. 2016; Chen et al. 2015; Wilker et al. 2015; Wilker et al. 2016). However, relatively little work has been done to examine the link between particulate air pollution and available markers of brain injury, and prior studies exclusively report on associations between recent air pollution exposures and markers of brain injury (Chen et al. 2015; Power et al. 2016a; Wilker et al. 2015). However, current brain health is a result of cumulative causes of brain injury that likely accumulate over decades, including aggregating proteins, ischemic injury, inflammation and oxidative stress, or exposure to toxins. As such, it is reasonable to expect that air pollution exposures over the prior years to decades may significantly contribute to current brain health. In addition, prior studies on air pollution and markers of brain injury are limited by lack of understanding of the selection process by which persons were selected for neuroimaging, which may lead to bias (Weuve et al. 2015).

To address these limitations, we conducted a study to quantify the association of long-term past exposure to particulate matter air pollution with MRI markers of neurodegeneration and subclinical cerebrovascular disease in older adults from the Atherosclerosis Risk in Communities Neurocognitive Study (ARIC-NCS). We hypothesized that long-term past exposure to particulate matter (PM) air pollution, specifically PM <2.5 μm in aerodynamic diameter (PM2.5), would be associated with smaller total brain volumes, as atrophy is an etiologically nonspecific indicator of cumulative brain damage, and increased risk of subclinical cerebrovascular disease. We also considered associations with regional brain volumes, given focal atrophy may suggest that PM exposures contribute to the pathogenesis of specific neurodegenerative processes.

Methods

Study Population

In 1987–1989 (Visit 1), the ARIC Study recruited 15,792 participants from four U.S. communities: Minneapolis, Minnesota suburbs; Jackson, Mississippi; Washington County, Maryland; and Forsyth County, North Carolina. Participants have since been invited to complete four additional study visits: Visit 2, 1990–1992; Visit 3, 1993–1995; Visit 4, 1996–1998; and Visit 5, 2011–2013. A sample of participants who attended Visit 5 were invited to undergo brain MRI as part of the ARIC-NCS (Knopman et al. 2015). Briefly, at each site, excluding those with contraindications to MRI, all persons who had any indication of cognitive impairment at Visit 5, all persons who had previously completed brain MRI as part of an ARIC substudy, and a stratified random sample of the remaining participants (stratified by age) were invited to complete a brain MRI. Of those who completed brain MRI (n=1,978), we excluded those with a history of surgery or radiation to the head, multiple sclerosis, or brain tumor (n=15), all nonblack or nonwhite participants from any study site and all black individuals from Minnesota or Maryland (n=15), those with an implausible estimated intracranial volume (eTIV) (n=2), and those for whom we were unable to estimate historical air pollution exposures (n=193). This study was approved by the institutional review boards of all participating institutions. All subjects provided written informed consent to participate at each study visit.

Particulate Matter Air Pollution Exposures

Based on each participant’s residential address, which was updated at each ARIC study visit, we estimated monthly exposures to PM2.5 and PM10 (PM with an aerodynamic diameter <10 μm) using validated spatiotemporal statistical models (Paciorek et al. 2009; Yanosky et al. 2014; Yanosky et al. 2008; Yanosky et al. 2009). These models used PM monitoring, and geographic and meteorological covariates, in conjunction with spatial smoothing, to describe monthly PM2.5 and PM10 levels with high spatial resolution. Given national monitoring data were available for PM2.5 only for 1999 onward, separate spatiotemporal models for PM2.5 were fit for the 1988–1998 and 1999–2007 time periods. The PM2.5 model for the earlier time period (1988–1998) relied on PM10 model predictions and had a simpler space–time structure. The PM2.5 models for both time periods had high predictive accuracy [cross-validation (CV) R2=0.77 for both 1988–1998 and 1999–2007]. The predictive ability of the PM10 model was slightly lower (CVR2=0.58 for both 1988–1998 and 1999–2007). Models generally performed well in both urban and rural areas and across seasons, though predictive performance varied somewhat by region [CVR2=0.81, 0.81, 0.83, 0.72, 0.69, 0.50, and 0.60 for the Northeast, Midwest, Southeast, Southcentral, Southwest, Northwest, and Central Plains regions, respectively, for PM2.5 from 1999–2007 (Yanosky et al. 2014)]. As our study sites are located in the Northeast, Midwest, and Southeast, predictive performance is expected to be similar across study sites.

Input data were available from 1988 onward; we generated PM estimates at the residential address of each participant from 1990–2007, given lower confidence in PM estimates in the first few years covered by the model and our goal to quantify associations with long-term past exposures. We did not use moving averages to avoid issues of bias due to secular trends in air pollution coupled with differences in brain health for those who underwent MRI early or late in the study period.

Specifically, among those participants with complete air pollution exposure estimates, we created three exposure summaries for use in our analyses. First, we considered average exposures from 1990–2007, which represents the period approximately 22 to 5 y prior to neuroimaging. We hypothesized that these long-term cumulative exposures would be most relevant to current brain health. Structural brain changes detectable on MRI considered here are expected to represent the culmination of years of brain injury; thus, long-term cumulative average exposure would be expected to be relevant to the severity of brain injury detectable on MRI. In addition, we also separately considered average exposures from 1990–1998 (approximately 14 to 22 y prior to neuroimaging) and from 1999–2007 (approximately 14 to 5 y prior to neuroimaging), to explore whether changes to exposure model before and after 1999 impacted our findings. However, we recognize that, if observed, differences in association across averaging periods could also be attributable to true differences in the impact of exposure based on the timing of exposure relative to outcome assessment.

Neuroimaging Measures

At each study site, participants completed 3T MRI scans according to a standardized protocol. Pulse sequences included a sagittal T1-weighted 3-D volumetric magnetization-prepared gradient echo (MPRAGE) pulse sequence, axial T2 fluid-attenuated inversion recovery and axial T2* weighted gradient echo. The ARIC MRI reading center (Mayo Clinic, Minnesota) analyzed all images.

Regional gray-matter volumes were quantified with FreeSurfer (version 5.1; Laboratory for Computational Neuroimaging at the Athinoula A. Martinos Center for Biomedical Imaging), and total brain and intracranial volumes were estimated using in-house algorithms (Jack et al. 2014). In our analyses, we consider gray-matter volumes of the total brain, the four lobes (frontal, parietal, temporal, occipital), the hippocampus, the deep-gray structures (thalamus, caudate, putamen, and pallidum), and total volume of multiple gray-matter regions known to atrophy preferentially in Alzheimer’s disease (parahippocampal, entorhinal, and inferior parietal lobules, hippocampus, precuneus, and cuneus), which we refer to as the AD signature region (Dickerson et al. 2011).

White matter hyperintensity (WMH) volumes were measured using an in-house algorithm. (Raz et al. 2013) As WMH volumes were not normally distributed, we created a dichotomous severe WMH variable defined as present if WMH volume is >5% of total white matter volume. Brain infarcts and microbleeds were identified, counted, and measured by a trained imaging technician and confirmed by a radiologist (Knopman et al. 2015). Lacunar infarcts were subsequently identified based on location and size (315 mm in diameter) (Wardlaw et al. 2013). Microbleeds were subsequently classified as lobar or subcortical based on location. In our analyses, we characterized infarcts, lacunar infarcts, microbleeds, lobar microbleeds, and subcortical microbleeds as present or absent.

Covariates

We used data collected at Visits 1 and 4 to define participant age, gender (male/female), education (highschool, >highschool), body mass index (BMI; normal/overweight/obese), and smoking status (current/former/never). BMI was defined as measured weight (kg) divided by the square of measured height (m), while all other covariates were defined via self-report. We also considered two measures of area-level socioeconomic status (SES), the proportion of the residential census tract population below the U.S. poverty line, and a summary measure of neighborhood wealth/income, education, and occupation combining U.S. Census tract–level characteristics denoted the Neighborhood SES score (Diez Roux et al. 2001). Each measure of SES was categorized into three levels (bottom quintile, middle three quintiles, top quintile) using center-specific cutoffs.

Statistical Analysis

We initially conducted all analyses stratified by study site. Brain volumes were z-transformed prior to use in analyses based on the mean and standard deviation (SD) of volumes in those individuals who met eligibility criteria for inclusion in our study. We used weighted linear or logistic regression to quantify the site-specific association between a 1μg/m3-higher PM exposure measure and each of our neuroimaging features. The weights accounted for the stratified random sampling used to select participants from each site from ARIC Visit 5 into the ARIC MRI sample; thus, our site-specific analyses can be interpreted as the association that would be observed in the full Visit 5 ARIC sample at each site. All models were adjusted for age, gender, race, education, and eTIV. Associations with 1990–1998 and 1990–2007 exposure summaries were adjusted for covariate values at the time of Visit 1 (1987–1989), while associations with 1999–2007 exposure summaries were adjusted for covariate values at the time of Visit 4 (1996–1998). To provide a summary estimate combining data from all four sites, we combined site-specific estimates using random effects meta-analysis (DerSimonian and Laird 1986). Use of random effects meta-analysis was chosen given potential heterogeneity in association due to differences in PM composition or other factors across study sites. It also allowed for formal evaluation of the evidence for heterogeneity across estimates using the I2 test. Moreover, this method has the benefit of allowing us to derive a summary measure of association despite evidence of intractable confounding by site; exposure and confounder distributions across sites did not always overlap.

In sensitivity analyses, we reestimated our site-specific and combined estimates of association a) additionally adjusting for BMI, smoking status, and our two measures of area-level SES; b) excluding persons with documented stroke before MRI; c) restricting to persons who did not move during follow-up; d) considering white participants only (there were too few black participants in the North Carolina site to allow a site-specific estimate among blacks or a pooled estimate combining the North Carolina and Mississippi site estimates); e) incorporating inverse probability of attrition weighting (Hernán et al. 2000; Power et al. 2016b) to account for potentially informative attrition from Visit 1 to Visit 5; f) excluding potential outliers in our exposure estimates through application of the generalized extreme studentized derivative test (Rosner 1983); g) using log-transformed WMH volumes as an outcome in linear regression models; and h) in models omitting weighting. We also reestimated our site-specific estimates of association using a 1-SD unit increase in site-specific exposure as the exposure contrast. All analyses were completed using SAS (version 9.4; SAS Institute Inc.) or Stata (version 14.0; StataCorp).

Results

In total, 1,753 persons met our eligibility criteria and were included in the analyses. At the time of MRI, participants were on average 76 y old, 40% were male, and 45% had greater than a high school education. Table 1 provides demographic and clinical characteristics of the study participants, as well as information on our MRI outcomes by study site. Overall, the Minnesota site was the most affluent of the four sites, followed in order by North Carolina, Maryland, and Mississippi.

Table 1. Selected characteristics for eligible ARIC-NCS participants by study site.
Characteristic MN (n=419)% or mean±SD MD (n=443)% or mean±SD NC (n=446)% or mean±SD MS (n=441)% or mean±SD p-Valuea
Age at baseline, y 53±5 53±5 54±5 52±5 0.0004
Age at MRI, y 76±5 77±5 77±5 75±5
Male 48 37 43 33 <0.0001
Black 0 0 6 100 <0.0001
>HS education 55 30 53 45 <0.0001
Smoking at baseline <0.0001
 Current 15 13 17 19
 Former 40 31 32 26
 Never 45 56 52 55
BMI at baseline, kg/m2 27±4 27±5 25±4 29±5 <0.0001
Neighborhood SES score at baseline 4.4±3.1 0.4±2.8 3.4±5.1 4.6±4.6 <0.0001
Proportion of residential census track below U.S. poverty line at baseline 0.05±0.03 0.08±0.05 0.07±0.06 0.31±0.14 <0.0001
Estimated intracranial volume, cm3 1436±154 1378±150 1406±159 1308±134 <0.0001
Total brain volume, cm3 1048±109 1009±104 1022±108 967±98 0.05
Frontal lobe volume, cm3 155±15 149±15 153±17 143±14 <0.0001
Parietal lobe volume, cm3 111±12 106±12 107±12 99±11 <0.0001
Occipital lobe volume, cm3 43±5 41±5 41±5 37±5 <0.0001
Temporal lobe volume, cm3 105±12 101±11 102±12 98±11 <0.0001
Deep-gray volume, cm3 30±3 30±3 30±3 29±3 0.004
Hippocampal volume, cm3 7.0±0.9 6.7±0.9 6.9±1.0 6.8±1.0 0.002
AD signature region volume, cm3 62±7 59±7 59±7 56±6 <0.0001
Severe WMHb 22 26 25 28 <0.0001
Infarcts present 24 25 27 27 0.01
Lacunes present 17 18 18 19 0.07
Microbleeds present 24 20 27 28 <0.0001
Subcortical microbleeds present 20 17 22 23 0.0005
Lobar microbleeds present 10 7 10 9 0.0003

Note: ARIC-NCS, Atherosclerosis Risk in Communities Neurocognitive Study; BMI, body mass index; HS, high school; MD, Maryland; MN, Minnesota; MS, Mississippi; NC, North Carolina; SD, standard deviation; SES, socioeconomic status; WMH, white matter hyperintensities.

aChi-square of F-test p-value for comparison of characteristics by site, after weighting; p-values for brain volumes are additionally adjusted for estimated intracranial volume.

bSevere WMH defined as WMH volume >5% of white matter volume.

Of the four sites, Minnesota and Mississippi had the lowest PM exposures, while Maryland and North Carolina had the highest (Table 2). Variation in exposure to PM10 was generally larger than variation in exposure to PM2.5. Site-specific coefficients of variation for our exposure estimates ranged from 0.03 to 0.11 μg/m3 for PM10 and 0.02 to 0.10 μg/m3 for PM2.5.

Table 2. Distribution of exposure by site and exposure averaging time period for eligible ARIC-NCS participants.
Exposure Site Time period Mean SD Minimum 25th Percentile 75th Percentile Maximum
PM2.5 MN 1990–1998 9.4 0.4 7.7 9.2 9.6 11.5
PM2.5 MN 1999–2007 13.1 0.7 9.3 12.9 13.4 16.7
PM2.5 MN 1990–2007 11.2 0.5 9.0 11.1 11.5 13.9
PM2.5 MD 1990–1998 15.1 1.0 11.8 14.6 15.9 18.2
PM2.5 MD 1999–2007 19.1 1.8 9.9 18.5 20.1 22.9
PM2.5 MD 1990–2007 17.1 1.3 11.4 16.5 17.9 20.5
PM2.5 NC 1990–1998 15.7 0.5 13.8 15.4 16.0 17.7
PM2.5 NC 1999–2007 11.4 0.7 8.7 11.1 11.7 18.9
PM2.5 NC 1990–2007 13.6 0.5 11.6 13.4 13.8 16.7
PM2.5 MS 1990–1998 12.4 0.3 11.6 12.2 12.5 13.3
PM2.5 MS 1999–2007 10.2 0.3 8.7 10.1 10.4 11.3
PM2.5 MS 1990–2007 11.3 0.2 10.4 11.2 11.4 12.3
PM10 MN 1990–1998 17.0 1.2 12.1 16.6 17.6 20.0
PM10 MN 1999–2007 16.6 1.8 10.5 16.2 17.5 21.6
PM10 MN 1990–2007 16.8 1.4 11.5 16.3 17.5 20.6
PM10 MD 1990–1998 23.3 2.3 16.2 22.0 25.1 30.2
PM10 MD 1999–2007 19.4 2.1 13.5 18.0 20.9 25.5
PM10 MD 1990–2007 21.4 2.1 15.5 20.1 22.9 27.8
PM10 NC 1990–1998 21.9 0.9 18.6 21.3 22.3 24.8
PM10 NC 1999–2007 18.2 0.8 14.4 17.7 18.5 20.8
PM10 NC 1990–2007 20.0 0.8 16.9 19.5 20.4 22.7
PM10 MS 1990–1998 18.7 0.5 17.1 18.3 18.9 20.0
PM10 MS 1999–2007 17.4 0.5 15.9 17.1 17.6 18.9
PM10 MS 1990–2007 18.0 0.5 16.6 17.7 18.3 19.4

Note: ARIC-NCS, Atherosclerosis Risk in Communities Neurocognitive Study; MD, Maryland; MN, Minnesota; MS, Mississippi; NC, North Carolina; PM, particulate matter; SD, standard deviation.

As there was evidence of moderate to high heterogeneity (I2>40%) across sites when considering analyses of PM2.5 and brain volumes, we discuss both the site-specific and pooled analyses (Table 3). In the Minnesota site, higher PM2.5 exposures were generally associated with smaller total and regional brain volumes, with slightly stronger associations observed when considering the 1990–1998 exposure period compared to the 1999–2007 exposure period. This pattern was not observed in the other three sites. Results from the Maryland and North Carolina sites were consistently null. In the Mississippi site, there was some evidence to support a protective association between higher PM exposures and larger AD signature region, temporal lobe, and occipital lobe volumes, regardless of exposure period; associations with other regions were typically null. When site-specific associations were pooled via meta-analysis, the resulting effect estimates were generally null. However, consistently adverse associations between PM2.5 exposure from 1999–2007 and frontal lobe volumes across sites resulted in a small, marginally significant pooled association [beta: 0.02 SD units per 1μg/m3 higher exposure; 95% confidence interval (CI): 0.04, 0.00] Similarly, consistently adverse associations between higher PM2.5 exposures in all three time periods and smaller deep-gray volumes across the Minnesota, Maryland, and North Carolina sites resulted in small, marginally significant pooled associations (e.g., for mean PM2.5 from 1990–2007, beta: 0.03 SD units per 1μg/m3 higher exposure; 95% CI: 0.08, 0.00). The overall pattern of site-specific and combined results was similar across our sensitivity analyses, including analyses implementing inverse probability weighting (Tables S1 and S2) and those omitting use of sampling weights (Table S3).

Table 3. Adjusted association between a 1μg/m3-higher past PM2.5 exposure and SD-unit brain volumes in 2011–2013 in the ARIC-NCS study.
Outcome and site n PM2.5 1990–1998 PM2.5 1999–2007 PM2.5 1990–2007
Beta (95% CI) p-Value I2/p-Value for heterogeneity Beta (95% CI) p-Value I2/p-Value for heterogeneity Beta (95% CI) p-Value I2/p-Value for heterogeneity
Total brain
 MN 419 0.09 (0.16, 0.01) 0.02 0.07 (0.11, 0.03) <0.01 0.1 (0.16, 0.05) <0.01
 MD 443 0.01 (0.03, 0.05) 0.58 0 (0.02, 0.02) 0.84 0 (0.03, 0.03) 0.96
 NC 446 0.03 (0.05, 0.11) 0.47 0.02 (0.09, 0.05) 0.54 0 (0.09, 0.08) 0.93
 MS 441 0.07 (0.09, 0.23) 0.38 0.04 (0.08, 0.16) 0.52 0.07 (0.09, 0.22) 0.39
 Combined 0 (0.06, 0.05) 0.87 58.1/0.07 0.02 (0.07, 0.02) 0.31 72.9/0.01 0.02 (0.09, 0.04) 0.5 75.8/0.01
Frontal lobe
 MN 417 0.1 (0.2, 0) 0.04 0.05 (0.11, 0.01) 0.08 0.09 (0.17, 0) 0.04
 MD 442 0.01 (0.06, 0.04) 0.76 0.01 (0.04, 0.01) 0.27 0.02 (0.05, 0.02) 0.36
 NC 446 0.02 (0.09, 0.13) 0.69 0.01 (0.09, 0.06) 0.75 0 (0.11, 0.11) 1
 MS 440 0.07 (0.08, 0.23) 0.36 0.02 (0.15, 0.11) 0.72 0.02 (0.14, 0.18) 0.81
 Combined 0.01 (0.07, 0.04) 0.63 35.8/0.20 0.02 (0.04, 0) 0.08 0/0.66 0.02 (0.05, 0.01) 0.13 0/0.41
Occipital lobe
 MN 417 0.09 (0.2, 0.03) 0.13 0.06 (0.13, 0) 0.06 0.1 (0.19, 0) 0.05
 MD 442 0 (0.06, 0.06) 0.97 0 (0.03, 0.02) 0.82 0 (0.04, 0.04) 0.91
 NC 446 0.01 (0.13, 0.14) 0.92 0.07 (0.16, 0.03) 0.17 0.06 (0.2, 0.08) 0.43
 MS 440 0.23 (0.02, 0.47) 0.07 0.15 (0.01, 0.31) 0.06 0.23 (0.02, 0.45) 0.04
 Combined 0 (0.08, 0.08) 1 46.5/0.13 0.02 (0.08, 0.04) 0.59 63.5/0.04 0.01 (0.10.1, 0.08) 0.79 64.0/0.04
Parietal lobe
 MN 417 0.1 (0.2, 0.01) 0.04 0.07 (0.14, 0.01) 0.08 0.1 (0.2, 0) 0.05
 MD 442 0.01 (0.04, 0.06) 0.65 0 (0.03, 0.02) 0.83 0 (0.04, 0.04) 0.99
 NC 446 0.01 (0.08, 0.1) 0.82 0.03 (0.09, 0.03) 0.34 0.02 (0.11, 0.07) 0.63
 MS 440 0.09 (0.1, 0.29) 0.36 0.05 (0.1, 0.19) 0.54 0.08 (0.11, 0.27) 0.39
 Combined 0.01 (0.07, 0.05) 0.77 44.4/0.15 0.01 (0.04, 0.01) 0.33 15.3/0.15 0.02 (0.07, 0.03) 0.48 34.4/0.21
Temporal Lobe
 MN 417 0.08 (0.17, 0.01) 0.08 0.05 (0.1, 0.01) 0.09 0.07 (0.15, 0) 0.05
 MD 442 0.01 (0.07, 0.04) 0.59 0 (0.03, 0.02) 0.82 0.01 (0.04, 0.03) 0.75
 NC 446 0.03 (0.07, 0.13) 0.54 0 (0.08, 0.08) 0.91 0.02 (0.09, 0.13) 0.71
 MS 440 0.24 (0.03, 0.45) 0.02 0.11 (0.04, 0.26) 0.14 0.21 (0.01, 0.41) 0.04
 Combined 0.01 (0.07, 0.09) 0.85 65.5/0.03 0.01 (0.04, 0.03) 0.68 35.7/0.20 0 (0.07, 0.07) 0.99 62.5/0.05
Deep gray
 MN 417 0.08 (0.21, 0.04) 0.19 0.06 (0.13, 0.01) 0.07 0.09 (0.2, 0.01) 0.07
 MD 442 0.04 (0.1, 0.02) 0.15 0.01 (0.04, 0.01) 0.29 0.03 (0.06, 0.01) 0.18
 NC 446 0.04 (0.18, 0.1) 0.55 0.04 (0.13, 0.05) 0.35 0.06 (0.19, 0.07) 0.38
 MS 440 0.11 (0.1, 0.32) 0.3 0.04 (0.12, 0.21) 0.61 0.09 (0.12, 0.29) 0.39
 Combined 0.04 (0.09, 0.01) 0.1 0/0.48 0.02 (0.04, 0) 0.09 0/0.46 0.03 (0.07, 0) 0.07 1.5/0.38
Hippocampus
 MN 416 0.01 (0.14, 0.12) 0.87 0.05 (0.12, 0.02) 0.15 0.06 (0.17, 0.05) 0.28
 MD 442 0.02 (0.1, 0.05) 0.53 0 (0.03, 0.04) 0.83 0 (0.05, 0.05) 1
 NC 443 0.07 (0.05, 0.2) 0.27 0.04 (0.17, 0.1) 0.6 0 (0.15, 0.16) 0.99
 MS 437 0.07 (0.32, 0.45) 0.74 0.12 (0.16, 0.41) 0.39 0.13 (0.25, 0.51) 0.5
 Combined 0 (0.06, 0.06) 0.99 0/0.63 0.01 (0.04, 0.02) 0.64 0/0.40 0.01 (0.05, 0.04) 0.71 0/0.68
AD signature
 MN 417 0.09 (0.19, 0.02) 0.11 0.03 (0.1, 0.04) 0.39 0.06 (0.15, 0.03) 0.2
 MD 442 0.01 (0.05, 0.06) 0.8 0.01 (0.03, 0.02) 0.7 0 (0.04, 0.03) 0.86
 NC 446 0.01 (0.09, 0.11) 0.85 0.02 (0.09, 0.05) 0.57 0.01 (0.12, 0.09) 0.8
 MS 440 0.2 (0.02, 0.42) 0.07 0.12 (0.03, 0.27) 0.12 0.2 (0.01, 0.4) 0.06
 Combined 0 (0.07, 0.07) 0.97 48.8/0.12 0.01 (0.03, 0.02) 0.63 9.2/0.35 0 (0.06, 0.05) 0.86 41.0/0.17

Note: Adjusted for age, gender, race, education and estimated intracranial volume. —, no data; ARIC-NCS, Atherosclerosis Risk in Communities Neurocognitive Study; CI, confidence interval; MD, Maryland; MN, Minnesota; MS, Mississippi; NC, North Carolina; PM, particulate matter.

Similarly, there was some evidence to suggest heterogeneity of association across sites when considering analyses of PM10 and brain volumes (Table 4). Site-specific analyses suggested adverse associations between higher mean PM10 over 1999–2007 or 1990–2007 and smaller total brain volumes, occipital lobe volumes, and deep-gray volumes in Minnesota (Table 4). As with the PM2.5 analyses, we also observed protective associations between higher long-term PM10 exposure and larger occipital lobe, temporal lobe, and AD signature region volumes in Mississippi. As with PM2.5, there was little evidence of an association between long-term PM10 exposure and total or regional brain volumes in pooled analyses, with the exception of an adverse association between higher mean PM10 in all three time periods and smaller deep-gray-region volumes (e.g., the PM10 1990–2007 time period, beta: 0.02; 95% CI: 0.04, 0.00). As above, the overall pattern of site-specific and combined results was similar across sensitivity analyses, including analyses implementing inverse probability weighting (Tables S4 and S5) and those omitting use of sampling weights (Table S6).

Table 4. Adjusted association between a 1μg/m3-higher past PM10 exposure and SD-unit brain volumes in 2011–2013 in the ARIC-NCS study.
Outcome and site n PM2.5 1990–1998 PM2.5 1999–2007 PM2.5 1990–2007
Beta (95% CI) p-Value I2/p-Value for heterogeneity Beta (95% CI) p-Value I2/p-Value for heterogeneity Beta (95% CI) p-Value I2/p-Value for heterogeneity
Total brain
 MN 419 0.01 (0.03, 0.01) 0.56 0.02 (0.05, 0) 0.06 0.02 (0.04, 0.01) 0.16
 MD 443 0 (0.01, 0.02) 0.75 0.01 (0.01, 0.02) 0.51 0 (0.01, 0.02) 0.64
 NC 446 0.01 (0.05, 0.07) 0.85 0.01 (0.06, 0.07) 0.85 0.01 (0.06, 0.07) 0.83
 MS 441 0.03 (0.05, 0.11) 0.42 0.03 (0.03, 0.09) 0.27 0.04 (0.04, 0.11) 0.31
 Combined 0 (0.01, 0.01) 0.96 0/0.77 0 (0.02, 0.02) 0.9 41.8/0.16 0 (0.02, 0.01) 0.86 7.9/0.35
Frontal lobe
 MN 417 0.01 (0.02, 0.04) 0.49 0.01 (0.04, 0.02) 0.52 0 (0.03, 0.03) 0.87
 MD 442 0 (0.03, 0.02) 0.69 0 (0.03, 0.02) 0.84 0 (0.03, 0.02) 0.75
 NC 446 0 (0.08, 0.08) 0.99 0.03 (0.05, 0.11) 0.51 0.01 (0.07, 0.1) 0.73
 MS 440 0.02 (0.05, 0.1) 0.54 0 (0.06, 0.06) 1 0.01 (0.06, 0.09) 0.74
 Combined 0 (0.01, 0.02) 0.81 0/0.82 0 (0.02, 0.01) 0.7 0/0.86 0 (0.02, 0.02) 0.86 0/0.96
Occipital lobe
 MN 417 0.02 (0.05, 0.01) 0.24 0.03 (0.05, 0) 0.05 0.03 (0.06, 0) 0.09
 MD 442 0 (0.03, 0.03) 0.87 0.01 (0.02, 0.04) 0.64 0 (0.03, 0.03) 0.9
 NC 446 0.01 (0.09, 0.12) 0.82 0.03 (0.08, 0.13) 0.62 0.02 (0.09, 0.13) 0.72
 MS 440 0.1 (0.02, 0.22) 0.1 0.1 (0.01, 0.19) 0.02 0.11 (0, 0.22) 0.05
 Combined 0 (0.03, 0.02) 0.77 21.6/0.28 0.01 (0.03, 0.05) 0.6 66.8/0.03 0 (0.03, 0.04) 0.86 54.3/0.09
Parietal lobe
 MN 417 0.01 (0.02, 0.04) 0.42 0.02 (0.05, 0.01) 0.29 0.01 (0.03, 0.02) 0.65
 MD 442 0 (0.02, 0.03) 0.72 0 (0.02, 0.03) 0.84 0 (0.02, 0.03) 0.78
 NC 446 0.02 (0.05, 0.09) 0.6 0.04 (0.02, 0.1) 0.22 0.03 (0.04, 0.1) 0.38
 MS 440 0.04 (0.05, 0.14) 0.38 0.05 (0.02, 0.13) 0.18 0.06 (0.04, 0.15) 0.24
 Combined 0.01 (0.01, 0.02) 0.31 0/0.85 0 (0.02, 0.03) 0.71 31.8/0.22 0 (0.01, 0.02) 0.75 0/0.51
Temporal lobe
 MN 417 0.01 (0.02, 0.03) 0.65 0.01 (0.04, 0.02) 0.57 0 (0.04, 0.03) 0.84
 MD 442 0.01 (0.03, 0.01) 0.47 0.01 (0.03, 0.02) 0.58 0.01 (0.03, 0.02) 0.5
 NC 446 0.02 (0.05, 0.09) 0.66 0.04 (0.02, 0.11) 0.2 0.03 (0.04, 0.1) 0.39
 MS 440 0.12 (0.02, 0.23) 0.02 0.1 (0.02, 0.17) 0.01 0.12 (0.03, 0.22) 0.01
 Combined 0.01 (0.02, 0.04) 0.52 51.1/0.11 0.02 (0.02, 0.05) 0.37 65.2/0.03 0.01 (0.02, 0.05) 0.43 61.0/0.05
Deep gray
 MN 417 0.01 (0.04, 0.01) 0.27 0.03 (0.05, 0) 0.04 0.02 (0.05, 0) 0.04
 MD 442 0.02 (0.05, 0.01) 0.12 0.02 (0.05, 0.01) 0.15 0.02 (0.05, 0.01) 0.12
 NC 446 0.04 (0.14, 0.06) 0.44 0.01 (0.11, 0.09) 0.78 0.03 (0.14, 0.08) 0.59
 MS 440 0.05 (0.05, 0.15) 0.33 0.04 (0.04, 0.12) 0.3 0.05 (0.04, 0.14) 0.28
 Combined 0.02 (0.03, 0) 0.07 0/0.57 0.02 (0.04, 0) 0.03 0/0.45 0.02 (0.04, 0) 0.02 0/0.49
Hippocampus
 MN 416 0.01 (0.04, 0.03) 0.69 0.02 (0.04, 0.01) 0.13 0.02 (0.04, 0.01) 0.24
 MD 442 0.02 (0.05, 0.02) 0.35 0.01 (0.04, 0.03) 0.75 0.01 (0.05, 0.02) 0.5
 NC 443 0.06 (0.03, 0.16) 0.2 0.08 (0.02, 0.19) 0.1 0.08 (0.03, 0.19) 0.14
 MS 437 0.05 (0.14, 0.24) 0.62 0.08 (0.05, 0.22) 0.23 0.08 (0.1, 0.25) 0.38
 Combined 0.01 (0.03, 0.02) 0.58 0/0.46 0 (0.03, 0.04) 0.87 47.9/0.12 0.01 (0.03, 0.02) 0.7 25.1/0.26
AD signature
 MN 417 0 (0.02, 0.03) 0.88 0.01 (0.04, 0.01) 0.28 0.01 (0.03, 0.02) 0.49
 MD 442 0 (0.02, 0.02) 0.99 0 (0.03, 0.02) 0.93 0 (0.03, 0.02) 0.96
 NC 446 0.02 (0.06, 0.1) 0.59 0.05 (0.03, 0.14) 0.2 0.04 (0.05, 0.13) 0.36
 MS 440 0.1 (0.01, 0.21) 0.07 0.12 (0.03, 0.2) 0.01 0.12 (0.02, 0.23) 0.02
 Combined 0 (0.01, 0.02) 0.6 12.0/0.33 0.02 (0.02, 0.06) 0.35 70.1/0.02 0.01 (0.02, 0.04) 0.54 53.5/0.09

Note: Adjusted for age, gender, race, education, and estimated intracranial volume. —, no data; ARIC-NCS, Atherosclerosis Risk in Communities Neurocognitive Study; MD, Maryland; MN, Minnesota; MS, Mississippi; NC, North Carolina; PM, particulate matter.

When considering the relation between PM2.5 or PM10 and the presence of MRI markers of cerebrovascular disease, there was little statistical evidence of heterogeneity of association across the four sites; thus, we focused on the analyses pooling estimates from all four sites via meta-analysis. Overall, there was little conclusive evidence to support an association between higher exposure to PM2.5 or PM10 in any time period and the presence of MRI markers of cerebrovascular disease in pooled analyses combining all four sites (Tables S7 and S8). However, the odds ratios (ORs) for the pooled associations between a 1μg/m3-higher mean PM2.5 exposure and either lacunes or subcortical microbleeds were consistently in the range of 1.04 to 1.10, although these associations were not statistically significant. Similarly, although the ORs for the pooled associations between a 1μg/m3-higher mean PM10 exposure and microbleeds were consistently in the range of 1.04 to 1.05; these associations were also not statistically significant. Results from our sensitivity analyses were broadly consistent with our primary analysis findings, including analyses implementing inverse probability weighting or omitting use of sampling weights (Tables S9 to S14).

Discussion

In pooled analyses combining all four sites, higher mean PM2.5 and PM10 exposures in the 5 to 20 y prior were associated with smaller deep-gray regional brain volumes and higher PM2.5 exposures 5–14 y prior were marginally associated with smaller frontal lobe volumes. We found little evidence in support of an association between higher long-term exposure to PM2.5 or PM10 over our three time periods of exposure and other brain volume measures or markers of cerebrovascular and small vessel disease in pooled analyses. However, there was evidence of significant heterogeneity in associations between PM and brain volumes by study site. When considering site-specific associations, we consistently observed smaller total and regional brain volumes with greater long-term exposure to PM2.5 in the Minnesota site, but not the other three sites. Throughout, where there was evidence of an association between PM exposure and brain volumes, the magnitude of these associations was similar to that seen in prior analyses in this sample, considering the association between midlife blood pressure and brain volumes. (Power et al. 2016b) For reference, the 0.05 to 0.1 SD unit effect size observed in the Minnesota site can be interpreted as loss of approximately 0.5% to 1% of regional brain volume.

Strengths of this study include the relatively large number of participants with MRI, our ability to use weighting to account for selection into the MRI subcohort in primary analyses and attrition from the baseline ARIC visit in sensitivity analyses, and consideration of long-term, cumulative past exposures. Our coefficients of variation for air pollution exposure estimates are similar to those calculated from other studies based in geographically constrained locations (albeit typically using shorter averaging periods) (Power et al. 2016a), suggesting that the variation in exposure at our four sites is similar to that found in other locations. However, the relatively small number of persons in each center limits our power to detect small true effects, systematically evaluate the potential for nonlinear associations, or assess effect modification by age or other personal factors. In addition, mild brain atrophy may have several root causes. Heterogeneity in the causes of neurodegeneration in our sample may contribute to the muted dose–response, especially if only a subset of the potential causes of neurodegeneration, including both neurodegenerative diseases and other sources of brain injury, are related to air pollution exposure. Similarly, our findings do not preclude the possibility of neurotoxic effects on the brain that are not captured by the considered MRI markers of brain injury; studies considering alternate markers (e.g., cortical thickness) may be useful. We did not consider associations with more recent exposures, or with cumulative exposures that include recent exposures. As such, we cannot comment on the relative importance of recent versus past exposures or whether recent exposures are an acceptable surrogate for long-term cumulative exposures. As with many recent studies of the health effects of air pollution, we used modeled exposure measures using residential address rather than personal exposure metrics, and we were unable to address the issue of indoor air pollution. Moreover, we cannot discount the possibility that regional variation in predictive accuracy of our model may complicate or invalidate comparison of site-specific effect estimates. Finally, we cannot exclude the possibility of chance findings.

There are a small number of reports considering the association between PM exposures and MRI-based measurements of brain structure or subclinical cerebrovascular disease. Collectively, including the current study, this body of literature fails to identify a consistent pattern of associations, as results are frequently null, with the few positive findings differing across studies. In a study nested within the Women’s Health Initiative Memory Study (WHIMS), PM2.5 air pollution exposures in 1999–2006 were not associated with gray-matter brain volumes assessed in 2005–2006 (Chen et al. 2015). However, higher PM2.5 exposures were associated with smaller, normal-appearing white matter brain volumes, with magnitudes of association of roughly 0.01-SD units volume per interquartile range increase in exposure. Additional analyses in WHIMS using a voxel-based approach found PM2.5 exposures in the 3 y prior to MRI were associated with areas of smaller cortical gray-matter and subcortical white-matter volumes (Casanova et al. 2016). Notably, the authors also report significant clusters of association whereby higher PM2.5 was associated with larger deep gray–matter nuclei volumes in WHIMS participants (Casanova et al. 2016), opposite to our own observations of associations between higher PM and smaller regional gray-matter volumes in ARIC participants. In participants from the Framingham Offspring Study who lived in the New England region, higher past-year PM2.5 exposure was associated with smaller total cerebral brain volumes and greater risk of covert brain infarcts, but not with WMH volumes, age-adjusted extensive WMH volumes, or hippocampal volumes (Wilker et al. 2015). Finally, in a study of participants from the Massachusetts Alzheimer’s Disease Research Center Longitudinal Cohort, there was no association between higher PM2.5 exposures in 2003 and either brain parenchymal fraction (a measure of brain atrophy) or the presence of microbleeds at an MRI between 2004 and 2010, while there was a protective association between higher PM2.5 exposure and smaller WMH volumes (Wilker et al. 2016).

Interestingly, studies of the relationship between air pollution and cognitive or related outcomes (e.g., MRI markers or neuropathology) that consider geographically localized samples are more likely to report null associations. In contrast, studies considering participants spread over larger geographic regions have been more likely to report adverse associations (Power et al. 2016a). We suggest several potential explanations. First, studies in geographically constrained locations are typically small, and the range of exposures tends to be smaller. Thus, such studies are likely underpowered to detect small effects. A meta-analytic approach, such as demonstrated here, for combining information about multiple small, geographically constrained studies in different locations can overcome this limitation without inducing concerns about strong or intractable confounding that may arise in pooled analyses. Second, studies with wider geographic distribution may be more susceptible to confounding by characteristics that vary regionally. As we have previously demonstrated elsewhere (Power et al. 2016a), it appears unlikely that residual confounding may fully account for the adverse findings in more geographically dispersed settings, given the characteristics such a confounder would have to have in order to fully account for previously observed associations. However, this possibility cannot be fully discounted, especially given evidence in this study that exposure and confounder distributions across sites do not always overlap. Thus, residual confounding may still lead to a biased estimate of the true association in more expansive settings when spatial confounding is strong and meta-analysis of site-specific associations is not used. Finally, it is possible that a focus on quantifying exposure based on particulate mass is contributing to this heterogeneity of findings. If specific PM species or other physical characteristics such as surface area confer the relevant toxic effect, geographically constrained studies may be studying the impact of less toxic exposures, while geographically broad studies may be capturing mixtures of these effects due to their larger study area. Our finding of heterogeneity in association across sites would support this hypothesis, and the finding of adverse associations in Minnesota but not the other three sites may be attributable not to chance, but to the relative toxicity of exposures. Future work will be needed to understand the drivers of the divide in findings between these two study types in order to establish a causal effect of air pollution on late-life brain health.

Another potential explanation for the finding of adverse associations between PM2.5 and brain volumes in the Minnesota site, but not the others, lies in the potential nonlinearity of the association. Minnesota had the lowest air pollution levels of the four sites, and previous studies have suggested a nonlinear relationship between PM and both total cerebral brain volume (Wilker et al. 2015) and cognitive function (Ailshire and Crimmins 2014; Oudin et al. 2015; Power et al. 2011), whereby the strongest associations were observed at the lowest levels of exposure. Given relatively small samples per site, we were not able to assess nonlinearity of exposure within site, but hope others may be able to follow up on this possibility in the future.

Conclusions

In conclusion, we found no associations between cumulative past PM exposure and MRI-based markers of cerebrovascular disease. Combining data across sites, higher past PM exposures were associated with smaller deep-gray volumes across sites, and higher PM2.5 in 1999–2007 was marginally associated with smaller frontal lobe volumes. When considering individual sites, higher PM2.5 exposures were associated with smaller brain volumes in the Minnesota site. Further work will be needed to replicate these findings and understand the sources of heterogeneity across sites, and will require consideration of a broader number of sites.

Acknowledgments

The ARIC study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (HHSN268201100005C, HHSN268201100006C, HHSN268201100007C, HHSN268201100008C, HHSN268201100009C, HHSN268201100010C, HHSN268201100011C, and HHSN268201100012C). Neurocognitive data is collected by U01 HL096812, HL096814, HL096899, HL096902, and HL096917, with previous brain MRI examinations funded by R01-HL70825. The sponsors had no role in the design and conduct of the study; collection management, analysis, and interpretation of the data; preparation review; or approval of the manuscript. The authors thank the staff and participants of the ARIC study for their important contributions.

References

Ailshire JA, Crimmins EM. 2014. Fine particulate matter air pollution and cognitive function among older US adults. Am J Epidemiol 180(4):359–366, PMID: 24966214, 10.1093/aje/kwu155.

Casanova R, Wang X, Reyes J, Akita Y, Serre ML, Vizuete W. 2016. A voxel-based morphometry study reveals local brain structural alterations associated with ambient fine particles in older women. Front Hum Neurosci 10:495, PMID: 27790103, 10.3389/fnhum.2016.00495.

Chen JC, Wang X, Wellenius GA, Serre ML, Driscoll I, Casanova R, et al. 2015. Ambient air pollution and neurotoxicity on brain structure: evidence from women’s health initiative memory study. Ann Neurol 78(3):466–476, PMID: 26075655, 10.1002/ana.24460.

DerSimonian R, Laird N. 1986. Meta-analysis in clinical trials. Control Clin Trials 7(3):177–188, PMID: 3802833, 10.1016/0197-2456(86)90046-2.

Dickerson BC, Stoub TR, Shah RC, Sperling RA, Killiany RJ, Albert MS, et al. 2011. Alzheimer-signature MRI biomarker predicts AD dementia in cognitively normal adults. Neurology 76(16):1395–1402, PMID: 21490323, 10.1212/WNL.0b013e3182166e96.

Diez Roux AV, Merkin SS, Arnett D, Chambless L, Massing M, Nieto FJ, et al. 2001. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med 345(2):99–106, PMID: 11450679, 10.1056/NEJM200107123450205.

Hernán MA, Brumback B, Robins JM. 2000. Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology 11(5):561–570, PMID: 10955409, 10.1198/016214501753168154.

Jack CR, Jr., Wiste HJ, Weigand SD, Rocca WA, Knopman DS, Mielke MM, et al. 2014. Age-specific population frequencies of cerebral beta-amyloidosis and neurodegeneration among people with normal cognitive function aged 50-89 years: a cross-sectional study. Lancet Neurol 13(10):997–1005, PMID: 25201514, 10.1016/S1474-4422(14)70194-2.

Knopman DS, Griswold ME, Lirette ST, Gottesman RF, Kantarci K, Sharrett AR, et al. 2015. Vascular imaging abnormalities and cognition: mediation by cortical volume in nondemented individuals: atherosclerosis risk in communities-neurocognitive study. Stroke 46(2):433–440, PMID: 25563642, 10.1161/STROKEAHA.114.007847.

Oudin A, Forsberg B, Adolfsson AN, Lind N, Modig L, Nordin M, et al. 2015. Traffic-related air pollution and dementia incidence in northern Sweden: a longitudinal study. Environ Health Perspect 124(3):306–312, PMID: 26305859, 10.1289/ehp.1408322.

Paciorek CJ, Yanosky JD, Puett RC, Laden F, Suh HH. 2009. Practical large-scale spatio-temporal modeling of particulate matter concentrations. Ann Appl Stat 3(1):370–397, 10.1214/08-AOAS204.

Power MC, Adar SD, Yanosky JD, Weuve J. 2016a. Exposure to air pollution as a potential contributor to cognitive function, cognitive decline, brain imaging, and dementia: a systematic review of epidemiologic research. Neurotoxicology 56:235–253, PMID: 27328897, 10.1016/j.neuro.2016.06.004.

Power MC, Schneider AL, Wruck L, Griswold M, Coker LH, Alonso A, 2016b. Life-course blood pressure in relation to brain volumes. Alzheimers Dement 12(8):890–899, PMID: 27139841, 10.1016/j.jalz.2016.03.012.

Power MC, Weisskopf MG, Alexeeff SE, Coull BA, Spiro A III, Schwartz J. 2011. Traffic-related air pollution and cognitive function in a cohort of older men. Environ Health Perspect 119(5):682–687, PMID: 21172758, 10.1289/ehp.1002767.

Raz L, Jayachandran M, Tosakulwong N, Lesnick TG, Wille SM, Murphy MC, et al. 2013. Thrombogenic microvesicles and white matter hyperintensities in postmenopausal women. Neurology 80(10):911–918, PMID: 23408873, 10.1212/WNL.0b013e3182840c9f.

Rosner B. 1983. Percentage points for a generalized ESD many-oulier procedure. Technometrics 25(2):165–172, 10.1080/00401706.1983.10487848.

Tzivian L, Dlugaj M, Winkler A, Weinmayr G, Hennig F, Fuks KB, et al. 2016. Long-term air pollution and traffic noise exposures and mild cognitive impairment in older adults: a cross-sectional analysis of the Heinz Nixdorf Recall Study. Environ Health Perspect 124(9):1361–1368, PMID: 26863687, 10.1289/ehp.1509824.

Wardlaw JM, Smith EE, Biessels GJ, Cordonnier C, Fazekas F, Frayne R, et al. 2013. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol 12(8):822–838, PMID: 23867200, 10.1016/S1474-4422(13)70124-8.

Weuve J, Proust-Lima C, Power MC, Gross AL, Hofer SM, Thiebaut R, et al. 2015. Guidelines for reporting methodological challenges and evaluating potential bias in dementia research. Alzheimers Dement 11(9):1098–1109, PMID: 26397878, 10.1016/j.jalz.2015.06.1885.

Wilker EH, Martinez-Ramirez S, Kloog I, Schwartz J, Mostofsky E, Koutrakis P, et al. 2016. Fine particulate matter, residential proximity to major roads, and markers of small vessel disease in a memory study population. J Alzheimers Dis 53(4):1315–1323, PMID: 27372639, 10.3233/JAD-151143.

Wilker EH, Preis SR, Beiser AS, Wolf PA, Au R, Kloog I, et al. 2015. Long-term exposure to fine particulate matter, residential proximity to major roads and measures of brain structure. Stroke 46(5):1161–1166, PMID: 25908455, 10.1161/STROKEAHA.114.008348.

Xu X, Ha SU, Basnet R. 2016. A review of epidemiological research on adverse neurological effects of exposure to ambient air pollution. Front Public Health 4:157, PMID: 27547751, 10.3389/fpubh.2016.00157.

Yanosky J, Paciorek C, Laden F, Hart J, Puett RC, Liao D, et al. 2014. Spatio-temporal modeling of particulate air pollution in the conterminous United States using geographic and meteorological predictors. Environ Health 13:63, PMID: 25097007, 10.1186/1476-069X-13-63.

Yanosky JD, Paciorek CJ, Schwartz J, Laden F, Puett R, Suh HH. 2008. Spatio-temporal modeling of chronic PM10 exposure for the Nurses’ Health Study. Atmos Environ (1994) 42(18):4047–4062, PMID: 19584946, 10.1016/j.atmosenv.2008.01.044.

Yanosky JD, Paciorek CJ, Suh HH. 2009. Predicting chronic fine and coarse particulate exposures using spatiotemporal models for the Northeastern and Midwestern United States. Environ Health Perspect 117(4):522–529, PMID: 19440489, 10.1289/ehp.11692.

Ultrafine and Fine Particle Number and Surface Area Concentrations and Daily Cause-Specific Mortality in the Ruhr Area, Germany, 2009–2014

Author Affiliations open

1Institute of Occupational, Social and Environmental Medicine, Center for Health and Society, Heinrich-Heine-University of Düsseldorf, Düsseldorf, Germany

2Institute of Energy and Environmental Technology e.V., Duisburg, Germany

3Federal Institute of Occupational Safety and Health, Dortmund, Germany

4Center for Nanointegration Duisburg-Essen (CENIDE), University Duisburg-Essen, Duisburg and Essen, Germany

5Department of Epidemiology, Lazio Region Health Service, Rome, Italy

6Institute of Environmental Medicine, Karolinska Institute, Stockholm, Sweden

PDF icon PDF Version (1.3 MB)

  • Background:
    Although epidemiologic studies have shown associations between particle mass and daily mortality, evidence on other particle metrics is weak.
    Objectives:
    We investigated associations of size-specific particle number concentration (PNC) and lung-deposited particle surface area concentration (PSC) with cause-specific daily mortality in contrast to PM10.
    Methods:
    We used time-series data (March 2009–December 2014) on daily natural, cardiovascular, and respiratory mortality (NM, CVM, RM) of three adjacent cities in the Ruhr Area, Germany. Size-specific PNC (electric mobility diameter of 13.3750nm), PSC, and PM10 were measured at an urban background monitoring site. In single- and multipollutant Poisson regression models, we estimated percentage change (95% confidence interval) [% (95% CI)] in mortality per interquartile range (IQR) in exposure at single-day (0–7) and aggregated lags (0–1, 2–3, 4–7), accounting for time trend, temperature, humidity, day of week, holidays, period of seasonal population decrease, and influenza.
    Results:
    PNC100750 and PSC were highly correlated and had similar immediate (lag0–1) and delayed (lag4–7) associations with NM and CVM, for example, 1.12% (95% CI: 0.09, 2.33) and 1.56% (95% CI: 0.22, 2.92) higher NM with IQR increases in PNC100750 at lag0–1 and lag4–7, respectfully, which were slightly stronger then associations with IQR increases in PM10. Positive associations between PNC and NM were strongest for accumulation mode particles (PNC 100500nm), and for larger UFPs (PNC 50100nm). Associations between NM and PNC<100 changed little after adjustment for O3 or PM10, but were more sensitive to adjustment for NO2.
    Conclusion:
    Size-specific PNC (50500nm) and lung-deposited PSC were associated with natural and cardiovascular mortality in the Ruhr Area. Although associations were similar to those estimated for an IQR increase in PM10, particle number size distributions can be linked to emission sources, and thus may be more informative for potential public health interventions. Moreover, PSC could be used as an alternative metric that integrates particle size distribution as well as deposition efficiency. https://doi.org/10.1289/EHP2054
  • Received: 18 April 2017
    Revised: 19 December 2017
    Accepted: 21 December 2017
    Published: 15 February 2018

    Address correspondence to F. Hennig, Universitätsklinikum Düsseldorf, AG Umweltepidemiologie, Postfach 101007, 40001 Düsseldorf, Germany. Telephone: 49 211 586729111. Email: Frauke.hennig@uni-duesseldorf.de

    Supplemental Material is available online (https://doi.org/10.1289/EHP2054).

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (278 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (77 KB)

Introduction

Increases of daily fine particulate matter [PM ≤2.5 μm and ≤10 μm, respectively, in aerodynamic diameter (PM2.5 and PM10)] have been shown to be associated with natural mortality (NM) in several North American and European cities (HEI 2010; Katsouyanni and Samet 2009; Samoli et al. 2008). Epidemiological studies have further shown that PM is associated with adverse health effects, such as short- and long-term cardiovascular morbidity and mortality, diseases of the central nervous system, respiratory morbidity, and lung cancer (WHO 2013). Toxicological studies suggest that inhaled ultrafine particles (UFPs) might be particularly harmful, because they can pass the lung epithelium more easily and translocate into the blood to be transported to other organs (Oberdörster et al. 2005). However, epidemiological evidence on pathogenic health effects of UFPs is still limited and inconclusive (HEI Review Panel on Ultrafine Particles 2013; WHO 2013), mainly due to the lack of routinely monitored UFP data and few dedicated measurement campaigns in the framework of specific research projects. UFPs are commonly measured as particle number concentration (PNC), representing more than 85% of the total PM2.5 particle number (Hinds 1999) while contributing little to the PM concentration. The latter is usually the only regulated ambient air particle metric worldwide. Although PM is a mixture composited by different particle sizes and numbers, particles of different size and number concentration are usually generated by different sources (Morawska et al. 1999) such that size and number distribution may provide a better understanding to identify sources as a potential basis for an intervention measure. The commonly used UFPs, defined as particles with an electric diameter <100 nm, for example, combine nucleation and Aitken mode particles (<30 nm and 30–100 nm respectively), whereas combustion-generated particles (from vehicle emissions) range from 30 nm to 500 nm (Vu et al. 2015). UFP concentration alone therefore does not inform about the different sources of the particles.

Another potentially important metric is the integrated measure of lung-deposited surface area concentration of airborne particles (PSC), which takes into account the surface area as well as the size-dependent deposition efficiency of respective particles in the respiratory system. This metric thus constitutes a proxy of the particle’s reactivity, which is related to surface area, as well as its capacity to carry adsorbed chemical species, both possibly promoting oxidative stress, a precursor of inflammatory effects (Hussain et al. 2009). Besides PM, PNC in different size fractions and particle surface area may hence provide a better measurement regarding the toxicity of PM exposure (Noël et al. 2016; Oberdörster 2000) as well as the identification of sources (Morawska et al. 1999).

In a European multicenter analysis on health effects of UFP number on natural and cardiorespiratory mortality including Finland, Sweden, Denmark, Germany, Italy, Spain, and Greece (Stafoggia et al. 2017), a weak delayed effect of UFP was estimated (>lag5). However, this multicenter study was limited by the heterogeneity of its exposure assessment methodology such as instrumentation capturing slightly different size ranges of particles or placement of monitors (background vs. traffic location) as well as by different measurement periods (time and duration) (Stafoggia et al. 2017). A slight difference in the size ranges of measured UFPs due to the use of different instruments has a great impact on the measured overall PNC because the number concentration of particles increases remarkably in the smallest size fractions. Moreover, the location of the monitoring equipment (height and placement of monitors with respect to local sources and the location of the study population) also substantially influence the representativeness of the exposure measurements and might introduce bias as a consequence of differential exposure measurement error (Stafoggia et al. 2017).

In this study we tried to overcome the aforementioned limitations by focusing on one large single study, located in the densely populated German Ruhr Area (Essen, Mülheim, and Oberhausen). Being part of the German Ultrafine Aerosol Network (GUAN) (Birmili et al. 2016), this time-series study benefits from a comparatively long measurement period of almost 6 y and an in-depth characterization of ultrafine (electric mobility diameter <100 nm, including the size ranges 13.3–30, 30–50, and 50–100 nm) and fine (electric mobility diameter 100–750 nm, including the size ranges 100–250, 250–500, and 500–750 nm) particles, including number concentration and lung-deposited PSC, a metric that has rarely been investigated in epidemiological studies to date. In addition, the measurement site is co-located with a central urban background monitoring station of the regional air quality network (Mülheim Styrum), enabling us to also make use of monitored PM10, nitrogen dioxide (NO2), and ozone (O3), which potentially confound or modify ultrafine or fine particle effects on health. In a recent meta-analysis, NO2 and PM2.5 both were associated with mortality in multipollutant analyses (Faustini et al. 2014), whereas O3 is highly correlated with temperature and sunlight, and hence might be an additional or even independent risk factor (Levy et al. 2012). Considering that multiple air pollutants originate from common sources, multiple air pollutants may interact with or confound each other. Results of the European study indicated that associations between UFPs and mortality were confounded by NO2, PM2.5, and PM2.5–10; whereas adjusting for PM10 or O3 had little influence on effect estimates (Stafoggia et al. 2017).

The objective of this study was to investigate the associations of size-specific PNC as well as lung-deposited PSC on natural, cardiovascular, and respiratory mortality in the Ruhr Area, based on a time-series study from March 2009 to December 2014. In addition to the toxicologically important novel particle metrics, we investigated the role of copollutants such as PM10, NO2, and O3.

Methods

Mortality Data

We collected daily mortality counts based on the primary cause of death, defined as natural [International Classification of Diseases, 10th Revision (ICD-10) A00–R99], cardiovascular (ICD-10 I00–I99) and respiratory (ICD-10 J00–J99) mortality in the three adjacent cities of Essen, Mülheim, and Oberhausen between January 2008 and December 2014 from the central statistical and IT services provider of North Rhine-Westphalia. The three adjacent cities (in an area of ∼379 km2) with a total of nearly 1 million inhabitants [Essen, ∼580,000 (210 km2); Mülheim, ∼170,000 (91 km2); and Oberhausen, ∼211,000 (77 km2)] are located in the western part of the metropolitan Ruhr Area. As respective outcomes, we used the sum of city-specific natural and cause-specific deaths per day. The primary cause of deaths was assigned based on the underlying disease instead of the immediate cause of death.

Air Pollution Data

Exposure data was collected at the project-specific measurement site (i.e., GUAN) provided by the Institute of Energy and Environmental Technology (IUTA), co-located to an urban background monitoring site of the regional air quality network (code “STYR”) operated by the North Rhine-Westphalia State Agency for Nature, Environment and Consumer Protection [Landesamt für Natur, Umwelt und Verbraucherschutz (LANUV) North Rhine-Westphalia (NRW), Essen, Germany] from March 2009 until December 2014. A detailed description of the measurement site and respective measurement techniques can be found elsewhere (Birmili et al. 2016). The measurement site is located close to the administrative border of the cities of Mülheim and Oberhausen. Within a 1-km buffer, the site is surrounded by highways with channel-like cross-sections (∼10 m below site level) and traffic of approximately 50,000 vehicles/day (∼250 m north), a railway yard (south/southwest), and a medium-trafficked street and its junction with a highway exit (west/northwest). Main wind directions are south/southwest and northeast. The mixed residential-, industrial-, and traffic-influenced character of the site is typical for many urban areas in the Ruhr Area and hence believed to be representative for a large part of the population living in the adjacent cities of Mülheim and Oberhausen, including their eastern neighbor Essen.

Measured particle characteristics included size-specific PNC of ultrafine, fine, and coarse particles (as well as their PSC) that deposit in the alveolar or tracheobronchial region of the lung (short: lung-deposited PSC). PNCs [number per cubic centimeter (n/cm3)] were measured with a scanning mobility particle sizer (TSI Inc.) in the size ranges of 13.3–750 nm electrical mobility diameter (Wang and Flagan 1990). In an effort to understand the health effects of different particle sizes, potentially generated by different emission sources and reaction processes, we investigated six particle size fractions, including particles size ranges of 13.3–30 nm (reflecting the nucleation mode: <30 nm), 30–50 nm, 50–100 nm (reflecting the Aitken mode: 30–100 nm), 100–250 nm, 250–500 nm, and 500–750 nm (reflecting the accumulation mode: 100–1,000 nm). The PSC of lung-deposited particles with a diameter of 20–1,000 nm was measured in micrometers squared per cubic centimeter every second using a nanoparticle surface area monitor (NSAM; model 3550, TSI Inc.) (Asbach et al. 2009). The NSAM uses an opposed flow unipolar diffusion charger followed by an ion trap to remove excess ions. Particles >1 μm are withheld by means of an impactor located at the NSAM entrance. The voltage in the ion trap can be adjusted to manipulate the particle size distribution and therefore the response function; that is, if the ion trap voltage is set to 200 V, the NSAM delivers the surface area deposited in the alveolar region, whereas it delivers the surface area of particles deposited in the tracheobronchial region when the voltage is set to 100 V. In our study, alveolar-deposited particles were monitored. The accuracy of surface determination decreases substantially for particle diameters below 20 nm and above 400 nm (Asbach et al. 2009). However, typical outdoor aerosol particles <20 nm in diameter and >400 nm in aerodynamic diameter contribute little to the total surface area.

Routinely monitored air pollutants at the central state-run (LANUV) monitoring site (STYR) included PM10 (β-attenuation), NO2 (chemiluminescence method), and O3 (ultraviolet absorption).

Covariates

Daily temperature [in degrees Celsius (°C), daily mean] and relative humidity were measured according to standardized protocols (VDI-guidelines 3786, parts 3 and 4; Verein Deutscher Ingenieure 2009, 2012) at a state-run monitoring site (Duisburg-Walsum), located 11 km northeasterly from the study site. External information on periods of influenza was collected from the central statistical and information technology services provider of North Rhine-Westphalia. In addition, we defined an indicator for population decrease during summer, following the definition in Stafoggia et al. (2017): namely a three-level variable assuming value “1” for the time of school holidays in North Rhine-Westphalia (6 wk within July and September; e.g., 9 July until 21 August in 2012 or 22 July until 3 September in 2013), and “2” in the 4-wk period around the school holidays; all other days stood for reference days and were assigned to “0”). Further variables included day of week (six indicator variables, with Sundays as the reference category), holiday (an indicator variable identifying the main bank holidays in North Rhine-Westphalia), and season (fall=September–November; spring=March–May, summer=June–August; and winter=December–February).

Statistical Analysis

The basic description of particle metrics, mortality, and meteorological data included visualizations of the time series, median [interquartile range (IQR)], and Spearman’s correlation coefficients between respective exposure variables.

To estimate associations between exposures and daily cause-specific mortality, we used Poisson regression models allowing for overdispersion. Regression models included penalized regression splines as a smoothing function for time trend. We further included potential confounders based on a review of current literature (Stafoggia et al. 2017). Adjusted models included mean air temperature [day of death (lag0) and a moving average of 1–3 d prior to the observed death (lag1–3)], relative humidity, and indicator variables for day of the week, holidays, influenza epidemics, and the presence of a population decrease in the respective cities during the summer vacation period. Air temperature was modeled by fitting a natural cubic regression spline to allow for nonlinear confounder adjustment.

We investigated single-lags from the same day of death (lag0) up to 7 d prior to death (lag7). Moreover, we investigated aggregated lags, representing immediate effects (0–1 d prior to the death; lag0–1), medium-term effects (lag2–3), and delayed effects (lag4–7). We chose single-lag models as well as aggregated 2- to 4-d lags over distributed lag-models because of multiple missing data in the PNC series and the respective loss of power, especially in the underlying small study population. By ending up with 11 models per investigated pollutant, we aimed to look for a general pattern of associations rather than identifying adverse health effects based on single-day lags that could be observed in such a multiple testing situation.

The main exposure metrics of interest were size-specific PNC, aggregated as ultrafine (PNC<100) and fine particles (PNC100–750), as well as PSC and PM10. In addition, we also investigated size-specific PNC in finer resolved size fractions (PNC13.3–30, PNC30–50, PNC50–100, PNC100–250, PNC250–500, and PNC500–750). All health effect estimates are presented as mean percentage increase [95% confidence interval (CI)] [% (95% CI)] in mortality per IQR of the respective exposure.

We calculated two-pollutant models in order to investigate whether results for UFPs (PNC<100) were independent of other pollutants or metrics: a) PNC<100 and PM10, b) PNC<100 and NO2, c) PNC<100 and O3, d) PNC<100 and PNC100–750, and e) PNC<100 and PSC. In addition we investigated two-pollutant models including a) PNC100–750 and PM10, b) PNC100–750 and NO2, c) PNC100–750 and O3, and d) PNC100–750 and PSC.

Furthermore, we investigated effect modification of UFPs and particles (PNC100–750) by cold and warm periods of the year (October–March vs. April–September), and by high or low concentration of PM10, O3, NO2 and PSC by including interaction terms between the potential effect modifier and the exposure of interest. High levels of PM10, O3, NO2, and PSC referred to concentrations above the 75th percentile of the respective distribution. Effect modification was checked based on a 5% significance level regarding the coefficient of the respective interaction term.

Results

Because particle metrics (PNC and PSC) were only measured beginning in March 2009, our analysis was based on the time period from March 2009 until December 2014 (2,132 d). We observed different missing patterns among exposures ranging from 266 missing days for PNC, 125 d for PSC, 110 d for O3, and 91 d for NO2 to 29 d for PM10. The majority of missing exposure data for PNC resulted from a sampling pump failure of the scanning mobility particle sizer during specific time windows (data not shown) and hence was assumed to be missing at random. Because of different missing patterns, the number of observations slightly changed between the analysis for each metric and lag.

Medians (IQRs) of daily cause-specific mortality per approximately 946,000 inhabitants in Essen, Mülheim, and Oberhausen were 32 (8) death/day for natural, 12 (5) for cardiovascular (corresponding to 37.5% of the overall deaths), and 3 (2) for respiratory mortality (corresponding to 9.4% of the overall deaths) (Table 1 and Figure 1). The city of Essen contributed most to the observed mortality (approximately 60%). Median (IQR) PNC of UFPs (PNC<100) was 9,871 n/cm3 (4,900), with the smallest size fraction (PNC13.3–30) contributing the most to PNC (4,623 n/cm3; 2,438). Median PSC was 36.1 μm2/cm3 (21.7) and PM10 was 20.2 μg/m3 (13.3), which is well below the European annual limit value of 40 μg/m3. In total, the PM10 24-h limit (50 μg/m3; EU 2008) was exceeded on 108 d (Figure 1). The median for NO2 was 29.2 μg/m3 (16.2), which was also below the annual limit value of 40 μg/m3. The median temperature was 11.9°C (9.9), and relative humidity 78.8% (18.5).

 

Table 1. Median (IQR) daily mortality, particle metrics, and meteorology in the Ruhr Area (Essen, Mülheim, and Oberhausen) between March 2009 and December 2014 (2,132 days).
Variable Median (IQR) Days (n)a
Mortality
 Naturalb 32.0 (8.0) 2,132
 Cardiovascularc 12.0 (5.0) 2,132
 Respiratoryd 3.0 (2.0) 2,132
Exposure PNC (n/cm3)
 PNC13.3–30 4,623.1 (2438.2) 1,866
 PNC30–50 2,673.1 (1492.5) 1,866
 PNC50–100 2,368.7 (1608.7) 1,866
 PNC<100 (UFP) 9,870.6 (4900.2) 1,866
 PNC100–250 1,209.7 (903.2) 1,866
 PNC250–500 195.8 (180.8) 1,866
 PNC500–750 9.0 (14.0) 1,866
 PNC100–750 (FP) 1,437.3 (1060.9) 1,866
PSC (μm2/cm3) 36.1 (21.7) 2,007
 PM10 (μg/m3) 20.2 (13.3) 2,103
 NO2 (μg/m3) 29.2 (16.2) 2,041
 O3 (μg/m3) 54.0 (37.0) 2,022
Meteorology
 Temperature (°C) 11.9 (9.9) 2,124
 Relative humidity 78.8 (18.5) 2,124

aThe number of days differs because of inconsistencies in measurements.

bEssen: 19.0 (6.0); Oberhausen 7.0 (4.0); Mülheim: 5.0 (3.0).

cEssen: 7.0 (4.0); Oberhausen 2.0 (3.0); Mülheim: 2.0 (2.0).

dEssen: 2.0 (2.0); Oberhausen 0.0 (1.0); Mülheim: 0.0 (1.0).

Figure 1 shows eight time series from March 2009 until December 2014.
Figure 1. Time series of daily cause-specific mortality (top left panel: natural mortality is shown in black, cardiovascular mortality is shown in gray, and respiratory mortality is shown in dark gray), PNC<100, PNC100–750, PSC, PM10 (top right panel: the dashed horizontal line indicates the 24-h limit of 50 μg/m3), NO2, O3, and temperature in the Ruhr Area. Note: NO2, nitrogen dioxide; O3, ozone; PM10, particulate matter ≤10 μm in aerodynamic diameter; PNC<100, size-specific particle number concentration of particles <100 nm electrical mobility diameter; PNC100–750, PNC of particles with 100–750 nm electrical mobility diameter; PSC, particle surface area concentration.

Spearman correlation (r) between air pollutants ranged from −0.39 (for NO2 and O3) to 0.99 (PNC100–250 and PNC100–750) (Table 2; based on data for 1,669 d with complete measurement data for all exposure metrics and pollutants.). PNC<100 (UFPs) generally correlated moderately with PSC and NO2 (r=0.63 and r=0.42), and correlated considerably more weakly with PM10 and O3 (r=0.26 and r=0.14). The smallest size fraction (PNC13.3–30) correlated weakly with other particle metrics and pollutants (0.00≤r≤0.28). PNC100–750 revealed overall high correlations with the particle metrics PSC (r=0.94) and PM10 (r=0.74). PNC100–750 correlated slightly weaker with NO2 than PNC<100 (r=0.65), whereas no correlation was observed between PNC100–750 and O3.

 

Table 2. Correlation coefficients (Spearman r) between exposure metrics and pollutants (n=1,669) in the Ruhr Area between March 2009 and December 2014 based on daily-based complete case data for all exposures, n=1,669.
PNC<100 PNC100–750 PNC13.3–30 PNC30–50 PNC50–100 PNC100–250 PNC250–500 PNC500–750 PSC PM10 NO2
PNC<100 1
PNC100–750 0.56 1
PNC13.3–30 0.86 0.25 1
PNC30–50 0.92 0.53 0.67 1
PNC50–100 0.79 0.85 0.43 0.82 1
PNC100–250 0.60 0.99 0.28 0.57 0.87 1
PNC250–500 0.27 0.82 0.03* 0.26 0.56 0.74 1
PNC500–750 0.21 0.71 −0.01* 0.21 0.49 0.64 0.9 1
PSC 0.63 0.94 0.28 0.66 0.90 0.93 0.76 0.68 1
PM10 0.26 0.74 0.00* 0.28 0.55 0.69 0.82 0.81 0.73 1
NO2 0.42 0.65 0.22 0.41 0.57 0.63 0.59 0.58 0.70 0.63 1
O3 0.14 −0.01* 0.15 0.13 0.06 0.03* −0.19 −0.25 −0.04* −0.12 −0.39

*p>0.05, all other p≤0.05.

Main Effects

Estimated associations of exposure with mortality showed different patterns for the different particle metrics and causes of mortality (Figure 2). Overall patterns of PNC100–750 and PSC were similar and comparable to those of PM10, showing immediate (lag0–1) and delayed (lag4–7) associations for NM and cardiovascular mortality (CVM) (Figure 2). Point estimates for immediate associations (lag0–1) of PNC100–750 were 1.12% (95% CI: −0.09, 2.33) for NM and 1.63% (95% CI: −0.40, 3.71) for CVM (see Table S1), and for more delayed associations (lag4–7) 1.56% (95% CI: 0.22, 2.92) for NM and 0.89% (95% CI: −0.43, 3.27) for CVM (see Table S1). These effect estimates were slightly stronger than those of PM10 on an IQR basis with an immediate (lag0–1) increase in NM and CVM of 0.67% (95% CI: −0.29, 1.64) and 0.99% (95% CI: −0.63, 2.65) or a more delayed (lag4–7) increase in NM and CVM of 0.97% (95% CI: −0.13, 2.09) and 0.75 (95% CI: −1.13, 2.67) (Figure 2; see also Table S1). We did not observe clear associations between PNC<100 (UFP) and NM or CVM, although the observed pattern suggested a more delayed association (lag4–7) with a slightly higher point estimate of 2.01% (95% CI: −1.41, 5.55), yet estimated less precisely (Figure 2; see also Table S1). For respiratory mortality (RM) we observed comparatively strong single-day associations at lag2 and lag6 with PNC<100 of 3.50% (95% CI: −0.77, 7.95) and 4.51% (95% CI: 0.37, 8.81), respectively. However, there were no conclusive patterns linking RM with aggregated lag-exposures of the considered pollutants.

Figure 2 shows twelve plots showing the percentage difference (95 percent confidence interval) of natural mortality, cardiovascular mortality, and respiratory mortality (y-axis) across single and aggregated lags (x-axis) estimated for particle metrics, namely, PNC subscript less than 100, PNC subscript 100 to 750, PSC, and PM subscript 10.
Figure 2. Short-term associations per IQR increase of air pollutant concentration and daily natural and cause-specific mortality in the Ruhr Area between March 2009 and December 2014, estimated for different particle metrics (PNC<100, PNC100–750, PSC, and PM10) at single-day lags (lag0–lag7) and for aggregated lags (lag0–1, lag2–3, lag4–7) in Poisson regression models, adjusted for time trend, temperature, humidity, day of week, holidays, period of seasonal population decrease, and influenza. (Corresponding numeric data are provided in Table S1.) Note: IQR, interquartile range; NO2, nitrogen dioxide; PM10, particulate matter ≤10 μm in aerodynamic diameter; PNC<100, size-specific particle number concentration of particles <100 nm electrical mobility diameter; PNC100–750, PNC of particles with 100–750 nm electrical mobility diameter; PSC, particle surface area concentration.

When looking at size-specific associations in more detail (Figure 3; see also Table S2), we observed immediate inverse associations of PNC13.3–30 with NM and CVM (−1.81% (95% CI: −3.30, −0.30) and −1.63% (95% CI: −4.16, 0.97), respectively; whereas for lag4–7, the estimate for NM moved close to the null and that for CVM was positive (95% CI: −0.55% (−2.40, 1.34) and 1.43% (95% CI: −1.86, 4.83) respectively). In contrast, patterns for PNC with an electric diameter >50–500 nm pointed to positive immediate (lag0–1) and delayed (lag4–7) associations with NM and CVM, similar to associations of PNC100–750, PSC and PM10. Clearest associations were observed for particles of 100–250 and 250–500 nm size and NM. For RM, patterns were less conclusive, yet somewhat different from NM and CVM, indicating only delayed associations with larger particles (electric diameter >250 nm).

Figure 3 shows eighteen plots showing the percentage difference (95 percent confidence interval) of natural mortality, cardiovascular mortality, and respiratory mortality (y-axis) across lags (x-axis) estimated for particle metrics, namely, PNC 13.3 to 30, PNC 30 to 50, PNC 50 to 100, PNC 100 to 250, PNC 250 to 500, and PNC 500 to 750.
Figure 3. Short-term associations (lag0–1, lag2–3, lag4–7) per IQR increase of size-specific particle number concentrations and daily natural and cause-specific mortality in the Ruhr Area between March 2009 and December 2014, estimated in Poisson regression models, adjusted for time trend, temperature, humidity, day of week, holidays, period of seasonal population decrease, and influenza and presented as percentage differences (95% confidence interval) [% (95% CI)] in mortality. (Corresponding numeric data are provided in Table S2.) IQR, interquartile range.

Adjustment for Copollutants

Effect estimates for NM and CVM in association with PNC<100 and PNC100–750 were similar after adjustment for O3 (Figure 4). In general, effect estimates were mostly robust towards adjustment for PM10, though associations between lag 4–7 PNC<100 and NM became more negative. Adjustment for NO2 on the other hand showed a slightly different pattern: Although effect estimates for UFP on CVM were unaffected by NO2 adjustment, effect estimates for PNC<100 and NM became more negative over all considered lags. Effect estimates for PNC100–750 on both NM and CVM were essentially unchanged after NO2 adjustment. After adjustment for PSC or PNC100–750, associations for PNC<100 and NM or CVM were similar to those adjusted for NO2. Associations between PNC100–750 and both outcomes at lag0–1 became more positive with adjustment for PSC, whereas the association between PNC100–750 and CVM at lag4–7 became negative, although confidence intervals were wide.

Figure 4A shows six plots for PNC subscript less than 100 showing the percentage difference (95 percent confidence interval) of natural mortality and cardiovascular mortality (y-axis) across main exposure metrics, PM subscript 10, O subscript 3, PNC subscript greater than 100, and PSC (x-axis) for lag 0 to 1, lag 2 to 3, and lag 4 to 7. Figure 4B shows the same data for PNC subscript 100 to 750.
Figure 4. Effect estimates for percentage differences (95% confidence interval) [% (95% CI)] in natural and cardiovascular-specific mortality in the Ruhr Area between March 2009 and December 2014 per IQR increase in (A) ultrafine particles (PNC<100) and (B) fine particles (PNC100–750, short: PNC>100) for averaged lags (lag0–1, lag2–3, lag4–7), estimated in Poisson regression models, adjusted for time trend, temperature, humidity, day of week, holidays, period of seasonal population decrease, and influenza with additional adjustment for PM10, NO2, O3, PNC>100 (PNC<100), and PSC. Note: IQR, interquartile range; NO2, nitrogen dioxide; O3, ozone; PM10, particulate matter ≤10 μm in aerodynamic diameter; PNC<100, size-specific particle number concentration of particles <100 nm electrical mobility diameter; PNC100–750, PNC of particles with 100–750 nm electrical mobility diameter; PSC, particle surface area concentration.

Associations between PNC13.3–30 and mortality remained unchanged after adjustment for other metrics (see Figure S1), consistent with expectations given the weak correlations with other pollutants (Table 2).

Effect Modification

Effect modification of associations between fine or ultrafine PNCs and natural or CV mortality were significant only for NM in association with O3 and PNC<100 at lag4–7 (interaction p=0.03), where PNC<100 was positively associated with NM when O3 was below the 75th percentile (1.31%; 95% CI: −0.46, 3.11), and negatively associated with NM when O3 was high (−1.94%; 95% CI: −4.63, 0.83) (Figure 5; see also Table S3). A similar pattern was observed for CVM in association with high or low O3 and PNC<100 at lag0–1 (interaction p=0.03). We did not observe significant (defined as interaction p<0.05) effect modification by season or higher levels of co-exposure (PM10, NO2, or PSC) regarding associations between fine or ultrafine PNCs and NM or CVM. However, at lag4–7, point estimates for PNC<100 were positive among those with lower levels of PM10, NO2, and PSC, but closer to the null among those with higher levels of co-exposure (interaction p: 0.17–0.67). Similarly, for NM and CVM, associations with PNC100–750 at lag0–1 were stronger for those with higher versus lower levels of PM10, NO2, and PSC co-exposure (interaction p=0.15–0.72). The effect estimate between lag2–3 PNC<100 and CVM was positive during the warmer season (April–September, 2.30%; 95% CI: −1.28, 6.06) but negative during colder months (October–March, −2.07%; 95% CI: −5.44, 1.43; interaction p=0.08).

Figure 5A shows six plots for PNC subscript less than 100 showing the percentage difference (95 percent confidence interval) of natural mortality and cardiovascular mortality (y-axis) across season, PM subscript 10, N O subscript 2, O subscript 3, and PSC (x-axis) for lag 0 to 1, lag 2 to 3, and lag 4 to 7. Figure 5B shows the same data for PNC subscript 100 to 750. The data are shown for warm (April to September) (high; greater than 75th percentile) and cold (October to March) (low; less than 75th percentile) periods of the year.
Figure 5. Estimated effect modification by season and copollutants for short-term (lag0–1, lag2–3, lag4–7) percentage differences in natural and cardiovascular mortality based on an IQR increase in the ultrafine particle concentration (PNC<100) in the Ruhr Area between March 2009 and December 2014 using Poisson regression models, adjusted for time trend, temperature, humidity, day of week, holidays, period of seasonal population decrease, and influenza. (Corresponding numeric data are provided in Table S3.) Note: IQR, interquartile range; PNC<100, size-specific particle number concentration of particles <100 nm electrical mobility diameter; PNC100–750, PNC of particles with an electrical mobility diameter between 100 and 750 nm.

Discussion

Our findings suggest that short-term exposures to lung-deposited PSC and PNC in the ultrafine (electric mobility diameter <100 nm) and fine (100–750 nm) particle size ranges (especially PNC 50–500 nm), are associated with small increases in daily NM and CVM. Associations suggested immediate (lag0–1) and slightly delayed (lag4–7) effects, and effect estimates were more precise for all NM than for the smaller subset of deaths due to cardiovascular disease. Associations of size-specific PNC were mostly robust to the adjustment for PM10 and O3, and slightly changed when adjusted for NO2. Effect estimates for PNC100–750 and PSC were similar to those observed for PM10, suggesting immediate as well as delayed effects on NM and CVM. Based on an IQR increase in respective exposure concentration, positive associations for PNC in the 50–500 nm range were stronger than positive associations for PM10.

In this study, we were able to investigate size-dependent PNC, including three size fractions in the UFP size range (13.3–30, 30–50, 50–500 nm) and three size fractions in the fine range (100–250, 250–500, 500–750 nm), aiming to identify the most pathogenic size fraction. We observed that the PNC of the smallest size ranges (13.5–50 nm) was inversely associated with natural and cause-specific mortality. This immediate inverse association of UFPs with natural and cause-specific mortality has been observed before in a German time-series study, showing inverse associations at lag1 and lag2, mainly driven by the smallest particle size, yet less pronounced than shown in our results (Stölzel et al. 2007). In contrast to the inverse association of the smallest size fraction, we observed positive immediate and delayed associations between UFP with an electric mobility diameter of 50–100 nm and daily mortality, which were similar to associations of other fine particle metrics (PNC100–750, PSC, and PM10). Among the fine to submicrometer particle size fractions (PNC100–750), particles with an electric mobility diameter from 100 to 250 and 250 to 500 nm revealed the clearest health effect estimates. Moreover, and in contrast to the inverse immediate associations, UFPs indicated delayed associations with CVM, as has been reported by others (Lanzinger et al. 2016; Stafoggia et al. 2017).

Typically, specific size ranges are related to major emission sources. Particles in the nucleation mode (<30 nm) reflect mainly new particles formed by gas-to-particle conversion, including particles originating from gaseous precursors in vehicle exhaust such as NO2 (Vu et al. 2015). Particles in the Aitken (30–100 nm) and accumulation (100 nm–1 μm) mode with an electric mobility diameter of 30–500 nm contain soot particles from combustion processes, including coal burning power plants, oil combustion, and combustion-engine powered vehicles (Vu et al. 2015). The modal size of vehicle-generated soot particles is in the size range of 100–250 nm (Harrison et al. 2010). Moreover, the particle size fraction 50–250 nm contains diesel exhaust particles, which have been shown to be specifically pathogenic in experimental settings (Mills et al. 2007). Particles from gasoline-powered engines, on the other hand, are typically smaller than diesel soot and mainly form particles <80 nm (Vu et al. 2015). Particles from mechanical abrasion processes such as brake, tire, and road wear are larger and can be found in the accumulation and coarse (>1 μm) mode (Vu et al. 2015). Moreover, accumulation mode particles encompass mostly long-range transported aerosols, whereas nucleation and Aitken mode particles usually have short lifetimes. From a biologic point of view, particles below 50 nm have the highest deposition efficiency, whereas Aitken and specifically accumulation mode particles deposit less efficiently (Kreyling et al. 2006). Moreover, particles below 50 nm contain a higher amount of soluble constituents.

Based on our findings, which show the largest associations for particles sized 50–500 nm, we concluded that primary combustion-generated soot particles might be more harmful than secondary particles formed via nucleation and condensation. This poses the question of whether the PNC in the size range from 50 to 500 nm might actually be a more important metric than the commonly used UFPs, which are defined as particles with a diameter <100 nm.

The repeatedly observed inverse associations for UFP (PNC<100) in temperature- and humidity-adjusted models seemed to be driven by the smallest particle size fraction (13.5–30 nm) and remained striking. From a biologic point of view, it seems implausible that the particles contained in the nucleation mode have a true protective effect on mortality. Associations with PNC<100 at lag0–1 remained inverse when additionally adjusted for NO2 and O3 in separate models, and they could not be explained through any investigated effect modification. In fact, point estimates became even more negative when O3 was below the 75th percentile.

Most time-series studies on short-term mortality effects of UFPs today have conducted single pollutant analyses. The important question remains, whether the observed effects of ultrafine or any other specific particle size fraction act independently of other pollutants considering that they are sharing potential sources. The answer to this question is of great interest with regard to the regulation of exposure and prevention of adverse health effects. In our study, inverse associations between UFP and natural and cause-specific mortality were robust to adjustment for O3 or PM10, but tended to move further from the null (i.e., became more negative) with adjustment for NO2, PNC100–750, or PSC. Similar patterns of for UFP-associations have been observed after adjustment for NO2, and also for PM2.5 before (Stafoggia et al. 2017), whereas others reported associations between prolonged exposure to UFP independent of particle mass exposures (Lanzinger et al. 2016). These contrary findings probably reflect important differences across studies caused by the different mixture of particles and sources due to the region of interest.

The rarely investigated lung-deposited PSC showed similar results as PNC100–750 or PM10, namely immediate and delayed associations with NM and CVM. Moreover, PSC correlated highly (>0.7) with PNC of particles sized 50–500 nm, which were the size-classes revealing the most clearly observed (immediate and delayed) health effect estimates.

Despite a strong correlation between PNC100–750, PM10 and PSC, PSC constitutes an integrated measure of reactive particle surface and deposition efficiency, which serve as a better marker understanding effect mechanisms between the inhalation of particles and health outcomes than solely mass-based or number-based metrics. It has been discussed that particle area surface plays a greater role in oxidative stress and pro-inflammatory effects than particle mass or particle number because the surface is the relevant location for oxidative processes (Hussain et al. 2009). Within this study, however, we were not able to disentangle biological effects of the mass, the number, and the surface of particles.

Season did not affect effect estimates of UFPs in the Ruhr Area consistently in terms of lag-time and cause of mortality, although season clearly affected effect estimates of UFPs on natural and cause-specific mortality and hospital admissions in other European regions (Samoli et al. 2016; Stafoggia et al. 2017). However, in comparison with the Mediterranean climate, the Ruhr Area has a more temperate climate with cool summers and mild and rainy winters, not displaying the strong seasonal pattern observed in Italy or Greece. Overall, we did not observe a consistent pattern among selected effect modifiers regarding associations between fine or ultrafine PNCs and natural or CV mortality.

Overall, our results are in line with results of other time-series studies, showing immediate (lag 0–1) and delayed effects (≥lag 4) of fine particles, while observing more delayed effects of UFPs on natural and cause-specific mortality (Breitner et al. 2009; Ibald-Mulli et al. 2002; Lanzinger et al. 2016; Stafoggia et al. 2017; Stölzel et al. 2007; Wichmann and Peters 2000). One of the first studies on UFPs reported the largest associations between UFPs and nonaccidental mortality for delayed (lag4) exposures in Erfurt, Germany (Wichmann et al. 2000). These results were confirmed in a reanalysis of an extended data base (Breitner et al. 2009; Stölzel et al. 2007). A European study including five cities (Augsburg, Chernovtsy, Dresden, Ljubljana, and Prague) reported an increase in respiratory mortality after 6 d (lag0–5) (Lanzinger et al. 2016). Another European study including eight cities (Helsinki, Stockholm, Copenhagen, Ruhr Area, Augsburg, Rome, Barcelona, and Athens) observed weak delayed associations (lag5–7) with NM and cardiovascular and respiratory mortality (Stafoggia et al. 2017). In contrast, several large multicenter time-series studies on fine particle mass showed primarily immediate effects on daily mortality (HEI 2010; Katsouyanni and Samet 2009; Samoli et al. 2008). Possible biological explanations for these different temporal patterns between size-specific particles could be local inflammation induced by fine particles in the bronchi and lung tissue, which may lead to immediate effects on mortality. In contrast, smaller particles such as UFPs may partly escape pulmonary clearing mechanisms, translocate across biologic membranes, and gain access to the vasculature and systemic circulation, stimulating systemic inflammatory mechanisms. This process can lead to an increased risk for cardiovascular events after several days. The overall reported delayed associations of UFPs and cardiovascular health seem plausible from this biological perspective. Supporting our findings, Stölzel et al. (2007) reported slightly higher delayed effect estimates with CVM than with NM for the UFPs.

Several limitations should be acknowledged in our study. The most obvious one is the small number of mortality events, limiting the statistical power of our results, especially regarding cause-specific mortality. Moreover, we have fitted several models to estimate adverse health effects of multiple pollutants regarding multiple lags and time windows, yielding a higher possibility of rejecting a null effect. However, in this study we aimed to identify a temporal pattern of different sized particles on the different causes of death instead of focusing on associations of single-day lags. In addition, this study used only one monitor as the reference exposure for three adjacent cities. Although PM10 and PM2.5 tend to be more homogeneously distributed over wider spatial regions with daily changes primarily dependent on meteorology, daily UFP concentration changes might differ considerably depending on location and local sources, especially in proximity to major roads or highways (Cyrys et al. 2008; Pekkanen and Kulmala 2004). For our study we assumed that the central monitor, placed at an urban background station, properly captured the day-to-day variability relevant for the surrounding population, as was assumed by others as well (Cyrys et al. 2008). Moreover, the high correlation of several exposure metrics limited our power to disentangle individual metric effects. Another limitation includes the lack of daily measurements of PM2.5, which has been shown to confound health effects of UFPs (Stafoggia et al. 2017).

The main strength of this study is the consistent exposure assessment throughout the study period of approximately 6 y. Furthermore, the study benefits from an in-depth characterization of particles, with the aim to specifically capture toxicologically important particle characteristics, including size-specific PNC and total lung-deposited PSC, a metric that has rarely been investigated in epidemiological studies to date. Moreover, the measurement site was located next to a routine monitoring site, enabling us to make use of monitored copollutants such as PM10, NO2, or O3, which can potentially confound or modify UFP effects on health.

Conclusions

Size-specific PNC (50–500 nm) and lung-deposited PSC indicated an association with NM and CVM in the Ruhr Area, showing immediate (lag0–1) and delayed (lag4–7) effect estimates revealing slightly higher point estimates than these of PM10 based on an IQR increase of exposure concentration. Although results from PM, PNC, and PSC could not be disentangled, it might be beneficial to investigate particle number size distributions, which can be linked to emission sources, in addition to the particle mixture captured by the measurement of PM10 only. Moreover, PSC could be used as an alternative metric that integrates particle size distribution as well as deposition efficiency. Further investigations are needed to establish the different temporal patterns among different particles sizes and surfaces.

Acknowledgments

We thank the Central statistical and IT services provider of North Rhine-Westphalia (Information und Technik NRW, Düsseldorf, Germany) and the North Rhine-Westphalia State Agency for Nature, Environment and Consumer Protection (Landesamt für Natur, Umwelt und Verbraucherschutz (LANUV) NRW, Essen, Germany) for providing, respectively, mortality and exposure data for the three cities of the Ruhr Area. We also thank D. Sugiri for the data management.

References

Asbach C, Fissan H, Stahlmecke B, Kuhlbusch TAJ, Pui DYH. 2009. Conceptual limitations and extensions of lung-deposited Nanoparticle Surface Area Monitor (NSAM). J Nanopart Res 11(1):101–109, 10.1007/s11051-008-9479-8.

Birmili W, Weinhold K, Rasch F, Sonntag A, Sun J, Merkel M. 2016. Long-term observations of tropospheric particle number size distributions and equivalent black carbon mass concentrations in the German Ultrafine Aerosol Network (GUAN). Earth Syst Sci Data 8(2):355–382, 10.5194/essd-8-355-2016.

Breitner S, Stölzel M, Cyrys J, Pitz M, Wölke G, Kreyling W. 2009. Short-term mortality rates during a decade of improved air quality in Erfurt, Germany. Environ Health Perspect 117(3):448–454, PMID: 19337521, 10.1289/ehp.11711.

Cyrys J, Pitz M, Heinrich J, Wichmann HE, Peters A. 2008. Spatial and temporal variation of particle number concentration in Augsburg, Germany. Sci Total Environ 401(1–3):168–175, PMID: 18511107, 10.1016/j.scitotenv.2008.03.043.

EU (European Commission). 2008. Directive 2008/50/EC of the European Parliament and of the Council of 21 May 2008 on ambient air quality and cleaner air for Europe. Off J Eur Union L 152/1.

Faustini A, Rapp R, Forastiere F. 2014. Nitrogen dioxide and mortality: review and meta-analysis of long-term studies. Eur Respir J 44(3):744–753, PMID: 24558178, 10.1183/09031936.00114713.

Harrison RM, Giorio C, Beddows DCS, Dall’Osto M. 2010. Size distribution of airborne particles controls outcome of epidemiological studies. Sci Total Environ 409(2):289–293, PMID: 21109288, 10.1016/j.scitotenv.2010.09.043.

HEI (Health Effects Institute). 2010. HEI Panel on the Health Effect of Traffic-Related Air Pollution: A Critical Review of the Literature on Emissions, Exposure, and Health Effects. HEI Spec Rep 17. Boston, MA:Health Effects Institute.

HEI Review Panel on Ultrafine Particles. 2013. HEI Perspectives 3: Understanding the Health Effects of Ambient Ultrafine Particles. Boston, MA:Health Effects Institute.

Hinds WC. 1999. Aerosol Technology. New York, NY:John Wiley & Sons.

Hussain S, Boland S, Baeza-Squiban A, Hamel R, Thomassen LCJ, Martens JA, et al. 2009. Oxidative stress and proinflammatory effects of carbon black and titanium dioxide nanoparticles: role of particle surface area and internalized amount. Toxicology 260(1-3):142–149, PMID: 19464580, 10.1016/j.tox.2009.04.001.

Ibald-Mulli A, Wichmann HE, Kreyling W, Peters A. 2002. Epidemiological evidence on health effects of ultrafine particles. J Aerosol Med 15(2):189–201, PMID: 12184869, 10.1089/089426802320282310.

Katsouyanni K, Samet JM. 2009. Air Pollution and Health: A European and North American Approach (APHENA). Health Effects Institute Research Report 142. Boston, MA:Health Effects Institute.

Kreyling WG, Semmler-behnke M, Möller W. 2006. Health implications of nanoparticles. J Nanopart Res 8(5):543–562, 10.1007/s11051-005-9068-z.

Lanzinger S, Schneider A, Breitner S, Stafoggia M, Erzen I, Dostal M, et al. 2016. Associations between ultrafine and fine particles and mortality in five central European cities – Results from the UFIREG study. Environ Int 88:44–52, PMID: 26708280, 10.1016/j.envint.2015.12.006.

Levy JI, Diez D, Dou Y, Barr CD, Dominici F. 2012. A meta-analysis and multisite time-series analysis of the differential toxicity of major fine particulate matter constituents. Am J Epidemiol 175(11):1091–1099, PMID: 22510275, 10.1093/aje/kwr457.

Mills NL, Törnqvist H, Gonzalez MC, Vink E, Robinson SD, Söderberg S. 2007. Ischemic and thrombotic effects of dilute diesel—exhaust inhalation in men with coronary heart disease. N Engl J Med 357(11):1075–1082, PMID: 17855668, 10.1056/NEJMoa066314.

Morawska L, Thomas S, Jamriska M, Johnson G. 1999. The modality of particle size distributions of environmental aerosols. Atmos Environ 33(27):4401–4411, 10.1016/S1352-2310(99)00217-4.

Noël A, Truchon G, Cloutier Y, Charbonneau M, Maghni K, Tardif R. 2016. Mass or total surface area with aerosol size distribution as exposure metrics for inflammatory, cytotoxic and oxidative lung responses in rats exposed to titanium dioxide nanoparticles. Toxicol Ind Health, PMID: 27256293, 10.1177/0748233716651560.

Oberdörster G. 2000. Toxicology of ultrafine particles: in vivo studies. Philos Trans A Math Phys Eng Sci 358(1775):2719–2740, 10.1098/rsta.2000.0680.

Oberdörster G, Oberdörster E, Oberdörster J. 2005. Nanotoxicology: an emerging discipline evolving from studies of ultrafine particles. Environ Health Perspect 113(7):823–839, PMID: 16002369, 10.1289/ehp.7339.

Pekkanen J, Kulmala M. 2004. Exposure assessment of ultrafine particles in epidemiologic time-series studies. Scand J Work Environ Health 30(suppl 2):9–18, PMID: 15487681.

Samoli E, Andersen ZJ, Katsouyanni K, Hennig F, Kuhlbusch TAJ, Bellander T. 2016. Exposure to ultrafine particles and respiratory hospitalisations in five European cities. Eur Respir J ERJ-02108-2015, PMID: 27338189, 10.1183/13993003.02108-2015.

Samoli E, Peng R, Ramsay T, Pipikou M, Touloumi G, Dominici F, et al. 2008. Acute effects of ambient particulate matter on mortality in Europe and North America: results from the APHENA study. Environ Health Perspect 116(11):1480–1486, PMID: 19057700, 10.1289/ehp.11345.

Stafoggia M, Schneider A, Cyrys J, Samoli E, Andersen ZJ, Bedada GB, et al. 2017. Association between short-term exposure to ultrafine particles and mortality in eight European urban areas. Epidemiology 28(2):172–180, PMID: 27922535, 10.1097/EDE.0000000000000599.

Stölzel M, Breitner S, Cyrys J, Pitz M, Wölke G, Kreyling W, et al. 2007. Daily mortality and particulate matter in different size classes in Erfurt, Germany. J Expo Sci Environ Epidemiol 17(5):458–467, PMID: 17108895, 10.1038/sj.jes.7500538.

Verein Deutscher Ingenieure. 2009. Environmental Meteorology – Meteorological Measurements – Air Humidity. Verein Deutscher Ingenieure-Richtlinien 1–14.

Verein Deutscher Ingenieure. 2012. Environmental Meteorology Meteorological Measurements Air Humidity. Verein Deutscher Ingenieure-Richtlinien 1–4.

Vu TV, Delgado-saborit JM, Harrison RM. 2015. Review: particle number size distributions from seven major sources and implications for source apportionment studies. Atmos Environ 122:114–132, 10.1016/j.atmosenv.2015.09.027.

Wang SC, Flagan RC. 1990. Scanning electrical mobility spectrometer. Aerosol Sci Technol 13(2):230–240, 10.1080/02786829008959441.

WHO (World Health Organization Regional Office for Europe). 2013. Review of Evidence on Health Aspects of Air Pollution – REVIHAAP Project, http://www.euro.who.int/__data/assets/pdf_file/0004/193108/REVIHAAP-Final-technical-report-final-version.pdf?ua=1 [accessed 5 February 2018].

Wichmann H, Spix C, Tuch T, Wölke G, Peters A, Heinrich J, et al. 2000. Daily mortality and fine and ultrafine particles in Erfurt, Germany Part I: role of particle number and particle mass. Res Rep Health Eff Inst 87–94, PMID: 11918089.

Wichmann HE, Peters A. 2000. Epidemiological evidence of the effects of ultrafine particle exposure. R Soc 358(1775):2751–2769, 10.1098/rsta.2000.0682.

Estimating Acute Cardiovascular Effects of Ambient PM2.5 Metals

Author Affiliations open

1Department of Environmental Health, Rollins School of Public Health, Emory University, Atlanta, Georgia, USA

2Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, Georgia, USA

3School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA

4Atmospheric Research & Analysis, Inc., Cary, North Carolina, USA

5Department of Biostatistics and Bioinformatics, Rollins School of Public Health, Emory University, Atlanta, Georgia, USA

PDF icon PDF Version (2.0 MB)

  • Background:
    Few epidemiologic studies have investigated health effects of water-soluble fractions of PM2.5 metals, the more biologically accessible fractions of metals, in their attempt to identify health-relevant components of ambient PM2.5.
    Objectives:
    In this study, we estimated acute cardiovascular effects of PM2.5 components in an urban population, including a suite of water-soluble metals that are not routinely measured at the ambient level.
    Methods:
    Ambient concentrations of criteria gases, PM2.5, and PM2.5 components were measured at a central monitor in Atlanta, Georgia, during 1998–2013, with some PM2.5 components only measured during 2008–2013. In a time-series framework using Poisson regression, we estimated associations between these pollutants and daily counts of emergency department (ED) visits for cardiovascular diseases in the five-county Atlanta area.
    Results:
    Among the PM2.5 components we examined during 1998–2013, water-soluble iron had the strongest estimated effect on cardiovascular outcomes [R͡R=1.012 (95% CI: 1.005, 1.019), per interquartile range increase (20.46 ng/m3)]. The associations for PM2.5 and other PM2.5 components were consistent with the null when controlling for water-soluble iron. Among PM2.5 components that were only measured during 2008–2013, water-soluble vanadium was associated with cardiovascular ED visits [R͡R=1.012 (95% CI: 1.000, 1.025), per interquartile range increase (0.19 ng/m3)].
    Conclusions:
    Our study suggests cardiovascular effects of certain water-soluble metals, particularly water-soluble iron. The observed associations with water-soluble iron may also point to certain aspects of traffic pollution, when processed by acidifying sulfate, as a mixture harmful for cardiovascular health. https://doi.org/10.1289/EHP2182
  • Received: 10 May 2017
    Revised: 5 October 2017
    Accepted: 8 December 2017
    Published: 15 February 2018

    Address correspondence to D. Ye, 1518 Clifton Rd. NE, CNR 2036, Atlanta, GA 30322 USA. Telephone: (203) 848-0225. Email: dongni.ye@emory.edu

    Supplemental Material is available online (https://doi.org/10.1289/EHP2182).

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (1.2 MB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (25 KB)

Introduction

Epidemiologic studies have indicated acute cardiovascular effects of fine particulate matter (PM2.5; particulate matter with aerodynamic diameter 2.5 μm) (Brook 2008; Dominici et al. 2006; Pope and Dockery 2006; Stafoggia et al. 2013). Because PM2.5 is a complex mixture of various chemical species, there is an ongoing effort to identify its health-relevant components. Nationwide multisite studies in the United States have examined whether the associations between PM2.5 and cardiovascular morbidity and mortality are modified by PM2.5 chemical composition (Bell et al. 2009; Franklin et al. 2008; Zanobetti et al. 2009). Other time-series studies have estimated associations between cardiovascular morbidity and mortality and individual PM2.5 components directly (Atkinson et al. 2015; Basagaña et al. 2015; Bell et al. 2014; Ito et&nfbsp;al. 2011; Levy et al. 2012; Lippmann et al. 2013; Ostro et al. 2006; Peng et al. 2009; Sarnat et al. 2015; Suh et al. 2011). Although the specific components that are associated with health outcomes vary across studies, there is growing evidence on the acute cardiovascular effects of metals/metalloids and carbonaceous components of PM2.5 (Kelly and Fussell 2012; Lippmann 2014; Rohr and Wyzga 2012).

Metals/metalloids exist in PM2.5 in different forms, with some forms being more water soluble and thus more biologically accessible than others (Allen et al. 2001; Birmili et al. 2006; Fang et al. 2015a; Heal et al. 2005). However, most ambient air pollution monitoring networks only measure these components in total elemental concentrations, and not in water-soluble concentrations. As a result, few epidemiologic studies have estimated health associations with PM2.5 water-soluble metals in their attempts to identify health-relevant components of PM2.5 (Heal et al. 2009; Huang et al. 2003).

To advance our understanding of acute cardiovascular effects of PM2.5 and its components, we conducted a time-series study in Atlanta, Georgia, to estimate the associations between daily counts of emergency department (ED) visits for cardiovascular diseases and daily concentrations of PM2.5 components, including a suite of PM2.5 water-soluble metals/metalloids that are not routinely measured at the ambient level. This analysis utilized up to 15 y of data on ambient air pollution and ED visits obtained as part of our ongoing Study of Particles and Health in Atlanta (SOPHIA) (Metzger et al. 2004; Sarnat et al. 2008; Tolbert et al. 2000; Ye et al. 2017).

Methods

Air Pollution Data

Ambient concentrations of criteria gases, PM2.5, and PM2.5 components were measured at the Jefferson Street ambient monitoring site during the period of 14 August 1998–15 December 2013 as part of the South Eastern Aerosol Research and Characterization (SEARCH) network and the Aerosol Research and Inhalation Epidemiology Study (ARIES) in Atlanta (Hansen et al. 2006). Criteria gases were measured daily, including 1-h maximum carbon monoxide (CO), 1-h maximum nitrogen dioxide (NO2), 1-h maximum sulfur dioxide (SO2), and 8-h maximum ozone (O3). PM2.5 and its major components—including organic carbon (OC), elemental carbon (EC), ammonium (NH4), nitrate (NO3), and sulfate (SO4)—were measured daily using filter-based 24-h integrated Federal Reference Methods. Total elemental concentrations of PM2.5 metals and metalloids (henceforth all referred to as metals), including titanium (Ti), manganese (Mn), iron (Fe), copper (Cu), zinc (Zn), aluminum (Al), lead (Pb), silicon (Si), calcium (Ca), sodium (Na), and potassium (K), were analyzed from the daily PM2.5 filters using X-ray fluorescence. X-ray fluorescence analyses were conducted by Desert Research Institute (Reno, NV) on filters collected through 22 March 2008, and by Atmospheric Research & Analysis, Inc. (Cary, NC) on filters collected after 23 March 2008; different limits of detection (LOD) were reported before and after the laboratory change for each species. Water-soluble concentrations of PM2.5 metals, including water-soluble vanadium (WS V), water-soluble chromium (WS Cr), water-soluble manganese (WS Mn), water-soluble iron (WS Fe), water-soluble nickel (WS Ni), and water-soluble copper (WS Cu), were analyzed using inductively coupled plasma optical emission spectrometry (ICP-OES) during 14 August 1998–6 April 2008. Starting on 7 April 2008, these water-soluble fractions were analyzed using inductively coupled plasma mass spectrometry (ICP-MS); again, different LODs were reported before and after the analytical change for these species. Additional water-soluble species—including water-soluble zinc (WS Zn), water-soluble cadmium (WS Cd), water-soluble lead (WS Pb), water-soluble selenium (WS Se), water-soluble arsenic (WS As), water-soluble barium (WS Ba), and water-soluble lanthanum (WS La)—were reported starting on 7 April 2008 from ICP-MS analyses. All water-soluble measures were available daily before 2009 and on one-in-three days after 2009.

The LODs of all PM2.5 metals are listed in Table S1. We calculated the percentage of samples below the LOD over the full time period, and over the time periods before and after measurement/laboratory changes separately. For this analysis, we included PM2.5 metals whose concentrations were above the LOD on at least 85% of days.

Ultimately, six PM2.5 metals (Si, K, Ca, Fe, Zn, WS Fe) were included in the analysis over the full time period (14 August 1998–15 December 2013), along with criteria gases (CO, NO2, SO2, and O3), PM2.5 mass, and PM2.5 major components (OC, EC, NO3, and SO4). We did not include NH4 in epidemiologic analyses because this component mainly exists as NH4NO3 or NH4SO4 Fifteen additional PM2.5 metals were included in the analysis over the later time period (7 April 2008–15 December 2013): Al, Na, Cu, Ti, WS Cr, WS Cu, WS Mn, WS Ni, WS V, WS As, WS Ba, WS Se, WS Zn, WS Cd, and WS Pb. For species included in the analysis, any observations below the LOD were assigned a value of the LOD divided by 2.

Emergency Department Visits

We obtained daily counts of cardiovascular ED visits for patients living within the five-county Atlanta area (Clayton, Cobb, DeKalb, Fulton, and Gwinnett) during 14 August 1998–15 December 2013. Daily ED visit counts were aggregated from individual-level billing records from metropolitan Atlanta hospitals as part of SOPHIA (Metzger et al. 2004; Winquist et al. 2016). We identified cardiovascular ED visits as those billing records with primary International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes for ischemic heart disease (ICD-9 410–414), cardiac dysrhythmias (ICD-9 427), congestive heart failure (ICD-9 428), or peripheral vascular and cerebrovascular disease (ICD-9 433–437, 440, 443–445, 451–453).

Analytic Approach

In a time-series framework, we estimated the associations between daily levels of air pollutants and daily counts of cardiovascular ED visits using Poisson regression accounting for over-dispersion. Based on our previous research of ambient air pollution and cardiovascular ED visits in Atlanta, we used the same-day (lag 0) pollution level (Metzger et al. 2004; Sarnat et al. 2013; Tolbert et al. 2000; Ye et al. 2017).

All models included the same covariate control for temporal trends and meteorology: time splines with monthly knots, cubic function of same-day maximum temperature, cubic function of lag 1-2–d moving average minimum temperature, cubic function of lag 0-1-2–d moving average mean dew point temperature, day of week, indicators for holidays, seasons, season–maximum temperature interaction, season–day of week interaction, indicators for hospital participation periods, and indicator for changes in air pollution measurement. The estimated associations were reported as rate ratios (RR) with 95% confidence intervals (CI) per interquartile range (IQR) increase in pollutant concentrations.

Primary Analysis

We included criteria gases (CO, NO2, SO2, and O3), PM2.5 mass, PM2.5 major components (OC, EC, NO3, and SO4), and PM2.5 metals (Si, K, Ca, Fe, Zn, WS Fe) in the analysis over the full time period (14 August 1998–15 December 2013). Fifteen additional PM2.5 metals (Al, Na, Cu, Ti, WS Cr, WS Cu, WS Mn, WS Ni, WS V, WS As, WS Ba, WS Se, WS Zn, WS Cd, and WS Pb) were included in the analysis over the later time period (7 April 2008–15 December 2013).

We first estimated the associations between these pollutants and cardiovascular ED visits using single-pollutant models. Based on the results, we applied multipollutant models to assess copollutant confounding. Because previous studies have reported differing effects of particulate matter on cardiovascular outcomes in cold versus warm days (Ito et al. 2011; Lippmann et al. 2013), we performed analyses in the warm and cold seasons separately for pollutants available over the full time period to see if the patterns of associations across pollutants were similar. We defined the warm season as May to October and the cold season as November to April in Atlanta.

For comparability, we restricted the analyses in each time period to days on which all pollutants were available. Thus, over the full time period, year-round analyses included 3,303 d; warm-season analyses included 1,737 d; and cold-season analyses included 1,566 d. Over the later time period, year-round analyses included 628 d.

Sensitivity Analyses

We evaluated model misspecification by estimating the associations between tomorrow’s pollutant levels and today’s ED visits, controlling for today’s (lag 0) pollutant and covariate levels. Tomorrow’s pollutant levels should not be associated with today’s ED visits in the absence of confounding, measurement error, or other model misspecification, because cause must precede effect (Flanders et al. 2011). To accommodate pollutants with one-in-three–day measurements, we defined “tomorrow” as the third day after today (lag3).

We restricted the primary analysis to days on which all pollutants were available so that the health associations of different pollutants were estimated on the same set of days (n=3,303 for the 1998–2013 year-round analysis; n=628 for the 1998–2013 year-round analysis). However, this led to reduced statistical power. As a sensitivity analysis, we performed the same set of analyses without this restriction by using all available days to see if the estimated associations were similar to those in the primary analysis.

Results

We calculated descriptive statistics of the pollutants over all seasons (Table 1), in the warm season (see Table S2a), and in the cold season (see Table S2b). OC, EC, NH4, SO4, and NO3 together contributed about 80% of the PM2.5 mass, whereas the concentrations of metals were much lower. Among metals, Si and Fe were most abundant. Water-soluble Fe had the highest average concentration among water-soluble species [as commonly seen in other studies (Allen et al. 2001; Birmili et al. 2006; Duan et al. 2014; Fang et al. 2015a; Lough et al. 2005)]. Secondary pollutants such as O3 and SO4 had higher concentrations in the warm than in the cold season, whereas primary pollutant such as CO had higher concentrations in the cold than in the warm season. The concentrations of metals were generally similar in the warm and cold seasons, whereas water-soluble Fe was higher in the warm than in the cold season.

Table 1. Summary statistics of ambient air pollutants measured at the Atlanta Jefferson Street monitoring site.
Pollutants Unit n Mean±SD 50th (25th, 75th) percentiles Interquartile range
14 August 1998–15 December 2013
 Criteria gases
  CO ppm 5,458 0.86±0.83 0.56 (0.36, 1.02) 0.66
  NO2 ppb 5,321 37.2±15.2 35.9 (26.4, 46.3) 20.0
  SO2 ppb 5,465 13.4±14.7 8.1 (3.2, 18.7) 15.5
  O3 ppb 5,490 42.1±19.9 39.6 (27.2, 54.9) 27.7
PM2.5
  PM2.5 mass μg/m3 5,588 14.46±7.69 12.81 (8.93, 18.21) 9.28
  OC μg/m3 5,546 3.67±2.08 3.22 (2.31, 4.47) 2.16
  EC μg/m3 5,515 1.26±0.98 0.98 (0.63, 1.58) 0.95
  NO4 μg/m3 5,563 1.39±1.00 1.10 (0.72, 1.73) 1.01
  NO3 μg/m3 5,569 0.81±0.77 0.55 (0.31, 1.06) 0.75
  SO4 μg/m3 5,572 3.88±2.96 2.94 (1.88, 4.87) 2.99
  Si ng/m3 4,932 94.51±112.37 68.16 (39.78, 110.79) 71.01
  K ng/m3 4,932 63.78±83.88 50.80 (35.28, 75.54) 40.26
  Ca ng/m3 4,932 36.42±29.71 29.32 (18.29, 44.71) 26.41
  Fe ng/m3 4,921 76.51±59.39 60.29 (39.64, 95.10) 55.47
  Zn ng/m3 4,880 11.35±11.16 8.84 (5.73, 13.31) 7.58
  WS Fe ng/m3 4,085 24.22±20.63 18.67 (10.81, 31.28) 20.46
7 April 2008–15 December 2013
  Na ng/m3 1,930 38.86±39.29 26.03 (14.66, 47.09) 32.43
  Al ng/m3 1,931 45.96±59.49 31.75 (17.37, 56.11) 38.74
  Ti ng/m3 1,931 4.47±3.91 3.62 (2.33, 5.36) 3.03
  Cu ng/m3 1,916 5.32±10.40 3.78 (2.44, 5.68) 3.23
  WS V ng/m3 805 0.20±0.19 0.14 (0.07, 0.26) 0.19
  WS Cr ng/m3 805 0.14±0.17 0.10 (0.06, 0.15) 0.09
  WS Mn ng/m3 796 1.20±0.98 0.94 (0.57, 1.54) 0.96
  WS Ni ng/m3 805 0.30±0.68 0.15 (0.09, 0.25) 0.16
  WS Cu ng/m3 790 2.83±4.56 1.84 (1.10, 3.06) 1.96
  WS Zn ng/m3 682 8.99±6.14 7.32 (4.69, 11.16) 6.47
  WS As ng/m3 805 0.68±0.53 0.56 (0.36, 0.80) 0.44
  WS Se ng/m3 805 0.72±0.59 0.55 (0.33, 0.92) 0.59
  WS Cd ng/m3 805 0.08±0.08 0.06 (0.04, 0.09) 0.05
  WS Ba ng/m3 805 3.24±3.23 2.45 (1.36, 4.10) 2.74
  WS Pb ng/m3 803 1.39±2.98 0.87 (0.56, 1.42) 0.86

Note: Criteria gases were measured daily, including 1-h maximum carbon monoxide (CO), 1-h maximum nitrogen dioxide (NO2), 1-h maximum sulfur dioxide (SO2), and 8-h maximum ozone (O3). PM2.5 and its major components, including organic carbon (OC), elemental carbon (EC), ammonium (NH4), nitrate (NO3), and sulfate (SO4), were measured daily using filter-based 24-h integrated Federal Reference Methods. Total elemental concentrations of PM2.5 metals and metalloids, including titanium (Ti), manganese (Mn), iron (Fe), copper (Cu), zinc (Zn), aluminum (Al), lead (Pb), silicon (Si), calcium (Ca), sodium (Na), and potassium (K), were analyzed from the daily PM2.5 filters using X-ray fluorescence. Water-soluble concentrations of PM2.5 metals, including water-soluble vanadium (WS V), water-soluble chromium (WS Cr), water-soluble manganese (WS Mn), water-soluble iron (WS Fe), water-soluble nickel (WS Ni), and water-soluble copper (WS Cu), were analyzed using inductively coupled plasma optical emission spectrometry (ICP-OES) during 14 August 1998–6 April 2008 and using inductively coupled plasma mass spectrometry (ICP-MS) starting on 7 April 2008. Water-soluble zinc (WS Zn), water-soluble cadmium (WS Cd), water-soluble lead (WS Pb), water-soluble selenium (WS Se), water-soluble arsenic (WS As), water-soluble barium (WS Ba), and water-soluble lanthanum (WS La) were reported starting on 7 April 2008 from ICP-MS analyses. All water-soluble measures were available daily before 2009 and one-in-three days after 2009.

Pearson correlations of the pollutants were also calculated over all seasons (see Table S3), in the warm season (see Table S4a), and in the cold season (see Table S4b). Over all seasons, PM2.5 was most correlated with SO4 (r=0.80), OC (r=0.74), EC (r=0.67), and WS Fe (r=0.65). Water-soluble Fe was most correlated with SO4 (r=0.61) and Fe (r=0.64). OC and EC were highly correlated with one another (r=0.79), and their correlations with other PM2.5 components were weak to moderate (r=0.170.58). PM2.5 was more strongly correlated with SO4 and O3 in the warm season, and with EC, OC, and metals in the cold season.

Summary statistics of cardiovascular ED visits are listed in Table 2. Briefly, average daily counts of cardiovascular ED visits were 76 during the full time period and 96 during the later time period. The average daily counts were similar in the warm and cold seasons in Atlanta.

Table 2. Summary statistics of emergency department visits for cardiovascular diseases.
Time period Total visits (n) Average visits [n (SD)] Min, max visits (n, n)
14 August 1998–15 December 2013 Year-round 426,252 76 (22) (21, 143)
Warm season 210,020 74 (21) (23, 136)
Cold Season 216,232 78 (23) (21, 143)
7 April 2008–15 December 2013 Year-round 199,343 96 (14) (55, 143)
Warm season 102,793 93 (13) (61, 136)
Cold season 96,550 99 (14) (55, 143)

Note: Daily counts of emergency department visits were aggregated from individual-level billing records from metropolitan Atlanta hospitals for patients living within the five-county Atlanta area (Clayton, Cobb, DeKalb, Fulton, and Gwinnett). We identified emergency department visits for cardiovascular diseases as those billing records with primary International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes for ischemic heart disease (ICD-9 410–414), cardiac dysrhythmias (ICD-9 427), congestive heart failure (ICD-9 428), or peripheral vascular and cerebrovascular disease (ICD-9 433–437, 440, 443–445, 451–453). Warm season includes May to October, and cold season includes November to April. Max, maximum; min, minimum.

Primary Analysis

We estimated the associations between cardiovascular ED visits and pollutants available during the full time period using single-pollutant models. The estimated RRs were positive for a number of pollutants, including criteria gases, PM2.5 mass, and PM2.5 components (OC, EC, NO3, Si, Ca, Fe, Zn, water-soluble Fe) (Figure 1). Among them, the estimated RR per IQR increase in water-soluble Fe was the highest [RR=1.012 (95% CI: 1.005, 1.019)].

Figures 1a and 1b are plots showing RR (95 percent confidence intervals) per IQR increase (y-axis) across the following single-pollutant models and two-pollutant models, respectively (x-axis): C O, N O 2, S O 2, O3, PM 2.5, OC, EC, N O 3, S O 4, Si, K, Ca, Fe, Zn, and WS Fe

Figure 1. Estimated associations between cardiovascular emergency department visits and pollutants available during 1998–2013, year-round analysis (3,303 d), Atlanta, Georgia. Results from single-pollutant models (a); results from two-pollutant models: water-soluble Fe controlling for each of the other pollutants (b). Note: Ca, calcium; CO, carbon monoxide; EC, elemental carbon; Fe, iron; IQR, interquartile range; K, potassium; NO2, nitrogen dioxide; NO3, nitrate; O3, ozone; OC, organic carbon; PM2.5, particulate matter with aerodynamic diameter 2.5 μm; RR, rate ratio; Si, silicon; SO2, sulfur dioxide; SO4, sulfate; WS, water soluble; Zn, zinc.

To assess whether the association for water-soluble Fe was confounded by other pollutants, we estimated the associations between cardiovascular ED visits and water-soluble Fe controlling for each of the other measured pollutants in two-pollutant models. The associations for water-soluble Fe changed little when controlling for any of the pollutants. In contrast, the associations for PM2.5 mass and PM2.5 components (OC, EC, NO3, Si, Ca, Fe, Zn) were weaker and consistent with the null when controlling for water-soluble Fe (Figure 1).

We performed analyses in the warm (May–October) and cold (November–April) seasons separately to see if the patterns of associations were similar. In the warm season, the estimated RR per IQR increase in water-soluble Fe was the highest. The associations for PM2.5 and a number of PM2.5 components (OC, EC, SO4, K) were consistent with the null (Figure 2). Although the estimated RRs for CO, Si, Ca, Fe, and Zn were positive in single-pollutant models, they were lower in two-pollutant models with water-soluble Fe (Figure 2).

Figures 2a and 2b are plots showing RR (95 percent confidence intervals) per IQR increase (y-axis) across the following single-pollutant models and two-pollutant models, respectively (x-axis): C O, N O 2, S O 2, O3, PM 2.5, OC, EC, N O 3, S O 4, Si, K, Ca, Fe, Zn, and WS Fe

Figure 2. Estimated associations between cardiovascular emergency department visits and pollutants available during 1998–2013, warm-season analysis (1,737 d). Results from single-pollutant models (a); results from two-pollutant models: water-soluble Fe controlling for each of the other pollutants (b). Note: Ca, calcium; CO, carbon monoxide; EC, elemental carbon; Fe, iron; IQR, interquartile range; K, potassium; NO2, nitrogen dioxide; NO3, nitrate; O3, ozone; OC, organic carbon; PM2.5, particulate matter with aerodynamic diameter 2.5 μm; RR, rate ratio; Si, silicon; SO2, sulfur dioxide; SO4, sulfate; WS, water soluble; Zn, zinc.

In the cold season, the estimated associations across pollutants were generally higher than those in the warm season (Figures 2 and 3). Among PM2.5 components, the estimated RR for water-soluble Fe was still the highest (Figure 3). The associations for CO, PM2.5, OC, EC, NO3, SO4, Si, K, and Ca were weaker and consistent with the null when controlling for water-soluble Fe. The association for water-soluble Fe was weaker in two-pollutant models with Fe (Figure 3).

Figures 3a and 3b are plots showing RR (95 percent confidence intervals) per IQR increase (y-axis) across the following single-pollutant models and two-pollutant models, respectively (x-axis): C O, N O 2, S O 2, O3, PM 2.5, OC, EC, N O 3, S O 4, Si, K, Ca, Fe, Zn, and WS Fe

Figure 3. Estimated associations between cardiovascular emergency department visits and pollutants available during 1998–2013, cold-season analysis (1,566 d). Results from single-pollutant models (a); results from two-pollutant models: water-soluble Fe controlling for each of the other pollutants (b). Note: Ca, calcium; CO, carbon monoxide; EC, elemental carbon; Fe, iron; IQR, interquartile range; K, potassium; NO2, nitrogen dioxide; NO3, nitrate; O3, ozone; OC, organic carbon; PM2.5, particulate matter with aerodynamic diameter 2.5 μm; RR, rate ratio; Si, silicon; SO2, sulfur dioxide; SO4, sulfate; WS, water soluble; Zn, zinc.

Measurements of an additional 15 PM2.5 metals were only available during the later time period. We estimated their associations with cardiovascular ED visits using single-pollutant models. The estimated RRs were the highest for water-soluble V [RR=1.012 (95% CI: 1.000, 1.025)] and Na [RR=1.008 (95% CI: 0.998, 1.017)]. We also estimated the association for water-soluble Fe in this later time period. The estimated RR (95% CI) for water-soluble Fe was 1.014 (0.988, 1.041) in the single-pollutant model, which was similar to that during the full time period, and the estimated RR had little change when controlling for any of the other metals in two-pollutant models (Figure 4).

Figures 4a and 4b are plots showing RR (95 percent confidence intervals) per IQR increase (y-axis) across the following single-pollutant models and two-pollutant models, respectively (x-axis): WS Fe, Na, Al, Ti, Cu, WS V, WS Cr, WS Mn, WS Ni, WS Cu, WS Zn, WS As, WS Cd, WS Ba, and WS Pb

Figure 4. Estimated associations between cardiovascular emergency department visits and pollutants only available during 2008–2013 and water-soluble Fe during 2008–2013, year-round analysis (628 d). Results from single-pollutant models (a); results from two-pollutant models: water-soluble Fe controlling for each of the other pollutants (b). Note: Al, aluminum; Cu, copper; IQR, interquartile range; Na, water-soluble sodium; RR, rate ratio; Ti, titanium; WS, water soluble; WS As, water-soluble arsenic; WS Ba, water-soluble barium; WS Cd, water-soluble cadmium; WS Cr, water-soluble chromium; WS Cu, water-soluble copper; WS Fe, water-soluble iron; WS Mn, water-soluble manganese; WS Pb, water-soluble lead; WS V, water-soluble vanadium; WS Zn, water-soluble zinc.

Sensitivity Analyses

For single-pollutant models in the year-round analysis, we evaluated model misspecification by estimating the associations between tomorrow’s pollutant levels and today’s ED visits, controlling for today’s pollutant and covariate levels. We found associations between cardiovascular ED visits and tomorrow’s levels of WS Ni and WS Mn, suggesting possible model misspecification when estimating these associations (see Figures S1 and S2). All other associations with tomorrow’s pollutant levels were consistent with the null, as expected under a well-specified model.

We restricted the primary analysis to days on which all pollutants were available. However, this led to reduced statistical power. We performed the same set of analyses without this restriction as a sensitivity analysis. We observed patterns of associations similar to those in the primary analysis, except that the association for SO4 in the cold season was more positive and the association for NO3 was more negative in this sensitivity analysis than in the primary analysis (see Figures S3–S6).

Discussion

In this study, we estimated acute cardiovascular effects of PM2.5 and its components, including a suite of water-soluble metals that are not routinely measured at the ambient level. We performed two-pollutant analysis to account for copollutant confounding, and compared the patterns of associations across pollutants in the warm and cold seasons.

Among the PM2.5 components we examined during the full time period (1998–2013), water-soluble Fe had the strongest estimated effect in both the warm and cold seasons. The associations for PM2.5 and other PM2.5 components were generally weak and consistent with the null when controlling for water-soluble Fe. Among PM2.5 components that were only measured during the later time period (2008–2013), water-soluble V was associated with cardiovascular ED visits.

Oxidative stress has been suggested as a central mechanism by which particulate matter affect health (Ghio et al. 2012). Transition metals can generate reactive oxygen species (ROS) in living systems, leading to oxidative stress (Ghio et al. 2012; Stohs and Bagchi 1995). Redox-active transition metals—such as Fe, Cu, Mn, and V—can act as catalysts of Fenton or Fenton-like reactions, facilitating the conversion of superoxide anion and hydrogen peroxide to hydroxyl radical (Chevion 1988; Stohs and Bagchi 1995). Because particle-bound metals need to dissolve and become metal ions to participate in these reactions, the water-soluble fractions of metals are thought to be more biologically relevant than total metals (Birmili et al. 2006; Urch et al. 2004). Recent studies have used cellular and cell-free assays to measure the oxidative potential of ambient particulate matter and have suggested that water-soluble metals—especially water-soluble Fe, water-soluble Cu, and water-soluble Mn—contribute to the ROS generation of particulate matter (Abrams et al. 2017; Cheung et al. 2012; Fang et al. 2015b; Landreman et al. 2008; Shen and Anastasio 2011; Verma et al. 2010). In our analysis, however, we observed positive associations with water-soluble Fe, but not with water-soluble Cu or water-soluble Mn. One reason could be that these species are less abundant than water-soluble Fe in the ambient air and thus could be more subject to measurement error, resulting in more underestimated health associations.

The observed associations with water-soluble Fe could indicate cardiovascular effects of certain pollution mixtures. Metals are released to the atmosphere from various sources, including natural processes acting on crustal minerals, resuspension of road dust and brake/tire wear abrasion during traffic, combustion of fossil fuels and wood, industrial processes, and waste incineration (Allen et al. 2001; Birmili et al. 2006; Duan et al. 2014; Fang et al. 2015a; Grigoratos and Martini 2015; Ito et al. 2004; Lin et al. 2015; Seinfeld 2006). Crustal species such as silicon, iron, calcium, sodium, aluminum, and potassium are largely found in the resuspension of road dust; meanwhile, copper, barium, manganese, iron, zinc, and chromium are commonly related to brake/tire wear debris; Nickel and vanadium are often attributed to residual oil combustion (Allen et al. 2001; Birmili et al. 2006; Duan et al. 2014; Fang et al. 2015a; Grigoratos and Martini 2015; Ito et al. 2004; Lin et al. 2015; Seinfeld 2006). The water-soluble fractions of these metals are partly from direct emission and partly from secondary processing of the primary insoluble metals by acid dissolution. A recent study in Atlanta investigated source contributions of a suite of water-soluble metals (Fang et al. 2015a). Roadway emissions, such as brake/tire wear debris and the resuspension of road dust followed by secondary processing by acid, were suggested as major contributors of a number of water-soluble metals, including water-soluble Fe, water-soluble Cu, water-soluble Mn, and water-soluble Zn. For water-soluble Fe, source apportionment attributed over 30% to mechanical abrasion of automobile brakes/tires and another 50% to secondary processing of Fe by acid (Fang et al. 2015a). Acid dissolution has been suggested as a major source of water-soluble Fe and other transition-metal ions in recent studies. Size distributions of soluble metals and particle pH have shown that sulfate plays a key role in producing highly acidic fine particles that are capable of dissolving primary transition metals (Fang et al. 2017). Single-particle analysis has shown that the majority of water-soluble Fe in Atlanta is in the form of iron sulfate (Longo et al. 2016; Oakes et al. 2012). We also found in this study that water-soluble Fe is mostly correlated with total Fe (r=0.64) and SO4 (r=0.61). These observations are consistent with the proposed mechanism of metal dissolution by acidic sulfate. The association we observed with water-soluble Fe points to certain aspects of roadway emission, when processed by acidic sulfate, as a mixture harmful for cardiovascular health. In our analysis, however, associations with other roadway-related metal species, such as water-soluble Cu, water-soluble Mn, water-soluble Zn, and water-soluble Ba, were consistent with the null. Again, these species are less abundant than water-soluble Fe in the ambient air and thus could be more subject to measurement error, resulting in more underestimated health associations.

The co-influence of the two sources on the levels of water-soluble Fe—primary roadway emission and secondary processing of the roadway emission by acidic sulfate—is reflected in the temporal trends of total Fe, water-soluble Fe, and sulfate in our study (Figure 5). Fe (i.e., total Fe) has no discernable seasonal or long-term trend. In contrast, the temporal trend of water-soluble Fe follows that of the sulfate: sulfate has a distinct seasonal trend with peaks in summer and a long-term decreasing trend potentially due to SO2 controls on coal-fired electrical generating units as well as their replacements (i.e., natural gas-fired units) (de Gouw et al. 2014). These observed trends illustrate how complex interactions between differing pollutant sources could affect the levels of potentially harmful components in PM2.5 and a co-benefit of SO2 reduction.

Three graphical representations respectively plotting concentrations of Fe in nanograms per cubic meter, water-soluble Fe in nanograms per cubic meter, and sulfate measured in micrograms per cubic meter (y-axis) across years 1998 to 2013 (x-axis).

Figure 5. Daily concentrations of Fe, water-soluble Fe, and sulfate measured at the Atlanta Jefferson Street ambient monitor, 14 August 1998–15 December 2013. Note: Fe (i.e., total Fe) was analyzed from daily PM2.5 filters using X-ray fluorescence. Water-soluble Fe was analyzed using ICP-OES during 14 August 1998–6 April 2008 and using ICP-MS starting on 7 April 2008. Measurements of water-soluble Fe was daily before 2009 and one-in-three days after 2009. Sulfate was measured daily using filter-based 24-h integrated Federal Reference Methods. Fe, iron; ICP-MS, inductively coupled plasma mass spectrometry; ICP-OES, inductively coupled plasma optical emission spectrometry; PM2.5, particulate matter with aerodynamic diameter 2.5 μm.

Fe (i.e., total Fe) and water-soluble Fe were both included in our analysis over the full time period, and their associations with cardiovascular ED visits were similar in single-pollutant models. In the warm season, the association with total Fe was consistent with the null when controlling for water-soluble Fe, suggesting that the water-soluble fraction was driving the association of Fe. This is expected if iron is a causal agent and its water-soluble fraction is more biologically accessible. However, we did not observe this pattern of associations in the cold season. One reason could be that the concentrations of water-soluble Fe in the cold seasons were much lower than in the warm seasons (see Table S2), and thus could be more subject to measurement error compared with total Fe, whose concentrations were similar in cold and warm seasons.

In fact, in the cold season, other PM2.5 components, such as EC and OC had stronger associations with cardiovascular ED visits than in the warm season. Although the associations for EC and OC were weaker when controlling for water-soluble Fe, the association of water-soluble Fe was also slightly weaker in two-pollutant models with these pollutants. EC and OC are partly from tailpipe emissions, and together with roadway-related species such as total Fe and water-soluble Fe, these pollutants may all contribute to cardiovascular effects of traffic pollution in the cold season.

Epidemiologic evidence on cardiovascular effects of water-soluble metals is sparse. In a time-series study in Edinburgh, Scotland, Heal et al. (2009) estimated the associations between cardiovascular hospital admissions and a number of PM2.5 total and water-soluble metals, including Cu, Fe, Ni, V, and Zn. However, direct measurements of these species were only available for 1 y, during which they did not find significant associations with total or water-soluble metals, nor with PM2.5 mass. Huang et al. (2003) exposed a panel of 38 healthy adults to concentrated ambient particles (CAP) from Chapel Hill, North Carolina, and reported that water-soluble metals in CAP (the V/Cu/Zn factor by principal component analysis) was associated with increased blood fibrinogen levels.

A number of studies have provided general evidence for acute cardiovascular effects of PM2.5 metals, although they only considered total elemental concentrations, not water-soluble fractions of metals (Basagaña et al. 2015; Bell et al. 2014; Bilenko et al. 2015; Huang et al. 2003; Ito et al. 2011; Lippmann et al. 2013; Morishita et al. 2015; Ostro et al. 2007; Suh et al. 2011; Urch et al. 2004; Zhang et al. 2016; Zhou et al. 2011). Suh et al. (2011) combined Cu, Mn, Zn, Ti, and Fe in a transition-metal category and reported positive associations with cardiovascular hospital admissions in a time-series study in Atlanta. In a time-series study in New York City, Ito et al. (2011) reported positive associations between cardiovascular hospital admissions and a number of PM2.5 components (OC, EC, SO4, Ni, V, Zn, Se, Br). Similarly in a time-series study of 64 U.S. counties, Lippmann et al. (2013) found positive associations between cardiovascular hospital admissions and OC, EC, SO4, Fe, V, and Zn. Zhang et al. (2016) reported that short-term exposures to transition metals (Cr, Fe, Cu, Mn, and Ni) in the ambient air were associated with decreased microvascular function in a panel of adults in Los Angeles, California. Morishita et al. (2015) found that a number of PM2.5 metals (As, Ca, Ce, Fe, Mg, Mn, S, Se, Ti) were associated with heart rate in a panel of adults in Dearborn, Michigan.

Some studies reported stronger associations with carbonaceous components than metals (Bell et al. 2014; Sarnat et al. 2015). In a time-series study in the St. Louis, Missouri–Illinois, area, Sarnat et al. (2015) found positive associations between cardiovascular ED visits and carbonaceous constituents (OC, EC, and certain hopanes), but not with metals (Si, K, Ca, Fe, Cu, Zn, and Pb). Bell et al. (2014), in a time-series study in four New England counties, observed positive associations between cardiovascular hospital admissions and black carbon, Ca, Zn, and V, where the association with black carbon was stronger than with the metals and was robust to copollutant adjustment of these metals. The inconsistencies between our study and these previous studies may be due to a number of factors, including the specific components being examined, copollutant confounding, pollutant interactions, nonlinear dose response, differences in population susceptibility, and measurement error. In particular, these studies considered only total elemental, not water-soluble, metals; besides, given that OC is itself a mixture of organic compounds, its health effects also depend on its composition, which likely varies by study location. In addition, previous studies have suggested synergism between organic compounds and metals in generating reactive oxygen species (Ghio et al. 2012; Li et al. 2009). Health associations of organic pollutants could depend on the levels of metals, and vice versa, which further complicates the comparison of health effects across PM components.

There are several limitations to our study. Our results are subject to spatial misalignment and instrument measurement error, and the degree of these sources of error likely differs by pollutant. Compared with pollutants dominated by secondary origins (e.g., O3, PM2.5, NO3, SO4, water-soluble metals), primary pollutants (e.g., EC, Fe, Cu, Zn) are likely more subject to spatial misalignment due to their greater spatiotemporal heterogeneity, and thus their estimated associations may be more biased towards the null. Additionally, pollutants with a lower ambient concentration may be more subject to instrument measurement error, leading to an underestimation of effects.

Conclusions

Our study suggests cardiovascular effects of certain water-soluble metals, particularly water-soluble Fe, which has not been well studied previously. Our findings further elucidate the link between traffic emissions, atmospheric secondary processing, and cardiovascular health, and contribute to the ongoing effort to identify causal mixtures in air pollution. The co-influence of two sources on the levels of water-soluble metals, roadway emission and secondary processing of the roadway emission by acidic sulfate, has implications for pollution control strategies.

Acknowledgments

The authors acknowledge the contributions of members of the Southeastern Center for Air Pollution and Epidemiology (SCAPE) research group. This publication is based in part upon information obtained from the Georgia Hospital Association and individual hospitals; we are grateful for the support of all participating hospitals.

Research reported in this publication was supported by funding from the Electric Power Research Institute (EPRI, 10002467). This publication was also made possible by a Clean Air Research Center grant to Emory University and the Georgia Institute of Technology from the U.S. Environmental Protection Agency (EPA, RD834799), as well as by grants to Emory University from the U.S. EPA (R82921301), the National Institute of Environmental Health Sciences (R01ES11294), and the EPRI (EP-P27723/C13172 and EP-P4353/C2124).

The content of this publication is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the U.S. EPA. Further, the U.S. EPA does not endorse the purchase of any commercial products or services mentioned in the publication.

References

Abrams JY, Weber RJ, Klein M, Sarnat SE, Chang HH, Strickland MJ, et al. 2017. Associations between ambient fine particulate oxidative potential and cardiorespiratory emergency department visits. Environ Health Perspect 125(10):107008, PMID: 29084634, 10.1289/EHP1545.

Allen AG, Nemitz E, Shi JP, Harrison RM, Greenwood JC. 2001. Size distributions of trace metals in atmospheric aerosols in the United Kingdom. Atmos Environ 35(27):4581–4591, 10.1016/S1352-2310(01)00190-X.

Atkinson RW, Mills IC, Walton HA, Anderson HR. 2015. Fine particle components and health—a systematic review and meta-analysis of epidemiological time series studies of daily mortality and hospital admissions. J Expo Sci Environ Epidemiol 25(2):208–214, PMID: 25227730, 10.1038/jes.2014.63.

Basagaña X, Jacquemin B, Karanasiou A, Ostro B, Querol X, Agis D, et al. 2015. Short-term effects of particulate matter constituents on daily hospitalizations and mortality in five South-European cities: results from the MED-PARTICLES project. Environ Int 75:151–158, PMID: 25461424, 10.1016/j.envint.2014.11.011.

Bell ML, Ebisu K, Leaderer BP, Gent JF, Lee HJ, Koutrakis P, et al. 2014. Associations of PM2.5 constituents and sources with hospital admissions: analysis of four counties in Connecticut and Massachusetts (USA) for persons ≥65 years of age. Environ Health Perspect 122(2):138–144, PMID: 24213019, 10.1289/ehp.1306656.

Bell ML, Ebisu K, Peng RD, Samet JM, Dominici F. 2009. Hospital admissions and chemical composition of fine particle air pollution. Am J Respir Crit Care Med 179(12):1115–1120, PMID: 19299499, 10.1164/rccm.200808-1240OC.

Bilenko N, Brunekreef B, Beelen R, Eeftens M, de Hoogh K, Hoek G, et al. 2015. Associations between particulate matter composition and childhood blood pressure—the PIAMA study. Environ Int 84:1–6, PMID: 26186643, 10.1016/j.envint.2015.07.010.

Birmili W, Allen AG, Bary F, Harrison RM. 2006. Trace metal concentrations and water solubility in size-fractionated atmospheric particles and influence of road traffic. Environ Sci Technol 40(4):1144–1153, PMID: 16572768, 10.1021/es0486925.

Brook RD. 2008. Cardiovascular effects of air pollution. Clin Sci 115(6):175–187, PMID: 18691154, 10.1042/CS20070444.

Cheung K, Shafer MM, Schauer JJ, Sioutas C. 2012. Diurnal trends in oxidative potential of coarse particulate matter in the Los Angeles Basin and their relation to sources and chemical composition. Environ Sci Technol 46(7):3779–3787, PMID: 22380575, 10.1021/es204211v.

Chevion M. 1988. A site-specific mechanism for free-radical induced biological damage—the essential role of redox-active transition-metals. Free Radical Bio Med 5(1):27–37, PMID: 2060841, 10.1016/0891-5849(88)90059-7.

de Gouw JA, Parrish DD, Frost GJ, Trainer M. 2014. Reduced emissions of CO2, NOx, and SO2 from U.S. power plants owing to switch from coal to natural gas with combined cycle technology. Earths Future 2(2):75–82, 10.1002/2013EF000196.

Dominici F, Peng RD, Bell ML, Pham L, McDermott A, Zeger SL, et al. 2006. Fine particulate air pollution and hospital admission for cardiovascular and respiratory diseases. JAMA 295(10):1127–1134, PMID: 16522832, 10.1001/jama.295.10.1127.

Duan JC, Tan JH, Hao JM, Chai FH. 2014. Size distribution, characteristics and sources of heavy metals in haze episode in Beijing. J Environ Sci (China) 26(1):189–196, PMID: 24649706.

Fang T, Guo H, Verma V, Peltier RE, Weber RJ. 2015a. PM2.5 water-soluble elements in the southeastern United States: automated analytical method development, spatiotemporal distributions, source apportionment, and implications for heath studies. Atmos Chem Phys 15(12):11667–11682, 10.5194/acp-15-11667-2015.

Fang T, Guo HY, Zeng LH, Verma V, Nenes A, Weber RJ. 2017. Highly acidic ambient particles, soluble metals, and oxidative potential: a link between sulfate and aerosol toxicity. Environ Sci Technol 51(5):2611–2620, PMID: 28141928, 10.1021/acs.est.6b06151.

Fang T, Verma V, Bates JT, Abrams J, Klein M, Strickland MJ, et al. 2015b. Oxidative potential of ambient water-soluble PM2.5 measured by dithiothreitol (DTT) and ascorbic acid (AA) assays in the southeastern United States: contrasts in sources and health associations. Atmos Chem Phys Discuss 15(21):30609–30644, 10.5194/acpd-15-30609-2015.

Flanders WD, Klein M, Darrow LA, Strickland MJ, Sarnat SE, Sarnat JA, et al. 2011. A method for detection of residual confounding in time-series and other observational studies. Epidemiology 22(1):59–67, PMID: 21068669, 10.1097/EDE.0b013e3181fdcabe.

Franklin M, Koutrakis P, Schwartz P. 2008. The role of particle composition on the association between PM2.5 and mortality. Epidemiology 19(5):680–689, PMID: 18714438, 10.1097/EDE.0b013e3181812bb7.

Ghio AJ, Carraway MS, Madden MC. 2012. Composition of air pollution particles and oxidative stress in cells, tissues, and living systems. J Toxicol Environ Health B Crit Rev 15(1):1–21, PMID: 22202227, 10.1080/10937404.2012.632359.

Grigoratos T, Martini G. 2015. Brake wear particle emissions: a review. Environ Sci Pollut Res Int 22(4):2491–2504, PMID: 25318420, 10.1007/s11356-014-3696-8.

Hansen DA, Edgerton E, Hartsell B, Jansen J, Burge H, Koutrakis P, et al. 2006. Air quality measurements for the Aerosol Research and Inhalation Epidemiology Study. J Air Waste Manag Assoc 56(10):1445–1458, PMID: 17063867, 10.1080/10473289.2006.10464549.

Heal MR, Elton RA, Hibbs LR, Agius RM, Beverland IJ. 2009. A time-series study of the health effects of water-soluble and total-extractable metal content of airborne particulate matter. Occup Environ Med 66(9):636–638, PMID: 19451142, 10.1136/oem.2008.045310.

Heal MR, Hibbs LR, Agius RM, Beverland IJ. 2005. Total and water-soluble trace metal content of urban background PM10, PM2.5 and black smoke in Edinburgh, UK. Atmos Environ 39(8):1417–1430, 10.1016/j.atmosenv.2004.11.026.

Huang YC, Ghio AJ, Stonehuerner J, McGee J, Carter JD, Grambow SC, et al. 2003. The role of soluble components in ambient fine particles-induced changes in human lungs and blood. Inhal Toxicol 15(4):327–342, PMID: 12635002, 10.1080/08958370304460.

Ito K, Mathes R, Ross Z, Nádas A, Thurston G, Matte T. 2011. Fine particulate matter constituents associated with cardiovascular hospitalizations and mortality in New York City. Environ Health Perspect 119(4):467–473, PMID: 21463978, 10.1289/ehp.1002667.

Ito K, Xue N, Thurston G. 2004. Spatial variation of PM2.5 chemical species and source-apportioned mass concentrations in New York City. Atmos Environ 38(31):5269–5282, 10.1016/j.atmosenv.2004.02.063.

Kelly FJ, Fussell JC. 2012. Size, source and chemical composition as determinants of toxicity attributable to ambient particulate matter. Atmos Environ 60:504–526, 10.1016/j.atmosenv.2012.06.039.

Landreman AP, Shafer MM, Hemming JC, Hannigan MP, Schauer JJ. 2008. A macrophage-based method for the assessment of the reactive oxygen species (ROS) activity of atmospheric particulate matter (PM) and application to routine (daily-24 h) aerosol monitoring studies. Aerosol Sci Tech 42(11):946–957, 10.1080/02786820802363819.

Levy JI, Diez D, Dou Y, Barr CD, Dominici F. 2012. A meta-analysis and multisite time-series analysis of the differential toxicity of major fine particulate matter constituents. Am J Epidemiol 175(11):1091–1099, PMID: 22510275, 10.1093/aje/kwr457.

Li QF, Wyatt A, Kamens RM. 2009. Oxidant generation and toxicity enhancement of aged-diesel exhaust. Atmos Environ 43(5):1037–1042, 10.1016/j.atmosenv.2008.11.018.

Lin YC, Tsai CJ, Wu YC, Zhang R, Chi KH, Huang YT, et al. 2015. Characteristics of trace metals in traffic-derived particles in Hsuehshan Tunnel, Taiwan: size distribution, potential source, and fingerprinting metal ratio. Atmos Chem Phys 15(8):4117–4130, 10.5194/acp-15-4117-2015.

Lippmann M. 2014. Toxicological and epidemiological studies of cardiovascular effects of ambient air fine particulate matter (PM2.5) and its chemical components: coherence and public health implications. Crit Rev Toxicol 44(4):299–347, PMID: 24494826, 10.3109/10408444.2013.861796.

Lippmann M, Chen LC, Gordon T, Ito K, Thurston GD. 2013. National Particle Component Toxicity (NPACT) Initiative: Integrated Epidemiologic and Toxicologic Studies of the Health Effects of Particulate Matter Components. Research report. Boston, MA:Health Effects Institute, 5–13.

Longo AF, Vine DJ, King LE, Oakes M, Weber RJ, Huey LG, et al. 2016. Composition and oxidation state of sulfur in atmospheric particulate matter. Atmos Chem Phys 16(21):13389–13398, 10.5194/acp-16-13389-2016.

Lough GC, Schauer JJ, Park JS, Shafer MM, Deminter JT, Weinstein JP. 2005. Emissions of metals associated with motor vehicle roadways. Environ Sci Technol 39(3):826–836, PMID: 15757346, 10.1021/es048715f.

Metzger KB, Tolbert PE, Klein M, Peel JL, Flanders WD, Todd K, et al. 2004. Ambient air pollution and cardiovascular emergency department visits. Epidemiology 15(1):46–56, PMID: 14712146, 10.1097/01.EDE.0000101748.28283.97.

Morishita M, Bard RL, Kaciroti N, Fitzner CA, Dvonch T, Harkema JR, et al. 2015. Exploration of the composition and sources of urban fine particulate matter associated with same-day cardiovascular health effects in Dearborn, Michigan. J Expo Sci Environ Epidemiol 25(2):145–152, PMID: 24866265, 10.1038/jes.2014.35.

Oakes M, Ingall ED, Lai B, Shafer MM, Hays MD, Liu ZG, et al. 2012. Iron solubility related to particle sulfur content in source emission and ambient fine particles. Environ Sci Technol 46(12):6637–6644, PMID: 22621615, 10.1021/es300701c.

Ostro B, Broadwin R, Green S, Feng WY, Lipsett M. 2006. Fine particulate air pollution and mortality in nine California counties: results from CALFINE. Environ Health Perspect 114(1):29–33, PMID: 16393654, 10.1289/ehp.8335.

Ostro B, Feng WY, Broadwin R, Green S, Lipsett M. 2007. The effects of components of fine particulate air pollution on mortality in California: results from CALFINE. Environ Health Perspect 115(1):13–19, PMID: 17366813, 10.1289/ehp.8335.

Peng RD, Bell ML, Geyh AS, McDermott A, Zeger SL, Samet JM, et al. 2009. Emergency admissions for cardiovascular and respiratory diseases and the chemical composition of fine particle air pollution. Environ Health Perspect 117(6):957–963, PMID: 19590690, 10.1289/ehp.0800185.

Pope CA III, Dockery DW. 2006. Health effects of fine particulate air pollution: lines that connect. J Air Waste Manage 56(6):709–742, 10.1080/10473289.2006.10464485.

Rohr AC, Wyzga RE. 2012. Attributing health effects to individual particulate matter constituents. Atmos Environ 62:130–152, 10.1016/j.atmosenv.2012.07.036.

Sarnat JA, Marmur A, Klein M, Kim E, Russell AG, Sarnat SE, et al. 2008. Fine particle sources and cardiorespiratory morbidity: an application of chemical mass balance and factor analytical source-apportionment methods. Environ Health Perspect 116(4):459–466, PMID: 18414627, 10.1289/ehp.10873.

Sarnat SE, Sarnat JA, Mulholland J, Isakov V, Özkaynak H, Chang HH, et al. 2013. Application of alternative spatiotemporal metrics of ambient air pollution exposure in a time-series epidemiological study in Atlanta. J Expos Sci Environ Epidemiol 23(6):593–605, 10.1038/jes.2013.41.

Sarnat SE, Winquist A, Schauer JJ, Turner JR, Sarnat JA. 2015. Fine particulate matter components and emergency department visits for cardiovascular and respiratory diseases in the St. Louis, Missouri–Illinois, metropolitan area. Environ Health Perspect 123(5):437–444, PMID: 25575028, 10.1289/ehp.1307776.

Seinfeld JPS. 2006. Atmospheric Chemistry and Physics: From Air Pollution to Climate Change. Hoboken, NJ:Wiley-Interscience.

Shen H, Anastasio C. 2011. Formation of hydroxyl radical from San Joaquin Valley particles extracted in a cell-free surrogate lung fluid. Atmos Chem Phys 11(18):9671–9682, PMID: 22121357, 10.5194/acp-11-9671-2011.

Stafoggia M, Samoli E, Alessandrini E, Cadum E, Ostro B, Berti G, et al. 2013. Short-term associations between fine and coarse particulate matter and hospitalizations in Southern Europe: results from the MED-PARTICLES project. Environ Health Perspect 121(9):1026–1033, PMID: 23777832, 10.1289/ehp.1206151.

Stohs SJ, Bagchi D. 1995. Oxidative mechanisms in the toxicity of metal ions. Free Radic Biol Med 18(2):321–336, PMID: 7744317, 10.1016/0891-5849(94)00159-H.

Suh HH, Zanobetti A, Schwartz J, Coull BA. 2011. Chemical properties of air pollutants and cause-specific hospital admissions among the elderly in Atlanta, Georgia. Environ Health Perspect 119(10):1421–1428, PMID: 21708510, 10.1289/ehp.1002646.

Tolbert PE, Klein M, Metzger KB, Peel J, Flanders WD, Todd K, et al. 2000. Interim results of the study of particulates and health in Atlanta (SOPHIA). J Expo Anal Environ Epidemiol 10(5):446–460, PMID: 11051535, 10.1038/sj.jea.7500106.

Urch B, Brook JR, Wasserstein D, Brook RD, Rajagopalan S, Corey P, et al. 2004. Relative contributions of PM2.5 chemical constituents to acute arterial vasoconstriction in humans. Inhal Toxicol 16(6–7):345–352, PMID: 15204750, 10.1080/08958370490439489.

Verma V, Shafer MM, Schauer JJ, Sioutas C. 2010. Contribution of transition metals in the reactive oxygen species activity of PM emissions from retrofitted heavy-duty vehicles. Atmos Environ 44(39):5165–5173, 10.1016/j.atmosenv.2010.08.052.

Winquist A, Grundstein A, Chang HH, Hess J, Sarnat SE. 2016. Warm season temperatures and emergency department visits in Atlanta, Georgia. Environ Res 147:314–323, PMID: 26922412, 10.1016/j.envres.2016.02.022.

Ye D, Klein M, Chang HH, Sarnat JA, Mulholland JA, Edgerton ES, et al. 2017. Estimating acute cardiorespiratory effects of ambient volatile organic compounds. Epidemiology 28(2):197–206, PMID: 27984424, 10.1097/EDE.0000000000000607.

Zanobetti A, Franklin M, Koutrakis P, Schwartz J. 2009. Fine particulate air pollution and its components in association with cause-specific emergency admissions. Environ Health 8:58, PMID: 20025755, 10.1186/1476-069X-8-58.

Zhang X, Staimer N, Tjoa T, Gillen DL, Schauer JJ, Shafer MM, et al. 2016. Associations between microvascular function and short-term exposure to traffic-related air pollution and particulate matter oxidative potential. Environ Health 15(1):81, PMID: 27460097, 10.1186/s12940-016-0157-5.

Zhou J, Ito K, Lall R, Lippmann M, Thurston G. 2011. Time-series analysis of mortality effects of fine particulate matter components in Detroit and Seattle. Environ Health Perspect 119(4):461–466, PMID: 21193387, 10.1289/ehp.1002613.

Cadmium Body Burden and Gestational Diabetes Mellitus: A Prospective Study

Author Affiliations open

1Key Laboratory of Environment and Health (HUST), Ministry of Education & Ministry of Environmental Protection, and State Key Laboratory of Environmental Health (Incubation), School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China

2Wuhan Medical and Health Center for Women and Children, Wuhan, Hubei, China

3Department of Environmental Health and Food Safety, Wuhan Centers for Disease Prevention and Control, Wuhan, Hubei, China

PDF icon PDF Version (264 KB)

  • Background:
    Several studies have reported that cadmium (Cd) is associated with type 2 diabetes. However, little is known about Cd exposure and the risk of gestational diabetes mellitus (GDM).
    Objective:
    We examined the association between Cd body burden in early pregnancy and the risk of GDM.
    Methods:
    We conducted a prospective study of 2,026 pregnant women from a single tertiary medical center between 2013 and 2016 in Wuhan, China. Cd body burden was reflected by Cd concentrations in urine samples collected between gestational weeks 8 and 14. GDM was diagnosed according to International Association of Diabetes and Pregnancy Study Groups Consensus Panel (IADPSG) recommendations.
    Results:
    The geometric mean of Cd concentrations in maternal urine of all pregnant women was 0.59 μg/L. A total of 198 (9.8%) women were diagnosed with GDM. After adjustment for potential confounders, the risk ratios (RRs) of GDM were 1.04 (95% CI: 0.74, 1.44) for the middle tertile of Cd levels and 1.36 (95%: CI: 0.98, 1.90) for the top tertile compared with the bottom tertile. In addition, we found a significant interaction between fetal sex and maternal Cd levels on the risk of GDM (pforinteraction=0.03). Among women carrying male fetuses, the RR of GDM was 1.86 (95% CI: 1.14, 2.93) for the top tertile of Cd levels compared with the bottom tertile.
    Conclusions:
    To our knowledge, this is the first report of an association between urinary Cd levels in early pregnancy and GDM. Our findings suggest that Cd body burden increases the risk of GDM and that the association may be modified by fetal sex. https://doi.org/10.1289/EHP2716
  • Received: 21 August 2017
    Revised: 26 December 2017
    Accepted: 27 December 2017
    Published: 8 February 2018

    Address correspondence to Y. Li, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430030, People’s Republic of China. Telephone: 86 (27) 83693417. Email: liyuanyuan@hust.edu.cn

    Supplemental Material is available online (https://doi.org/10.1289/EHP2716).

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (567 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (79 KB)



Introduction

Gestational diabetes mellitus (GDM) is defined as any degree of glucose intolerance that is first recognized during pregnancy (American Diabetes Association 2011). The prevalence of GDM has been steadily increasing in many countries, including China (Albrecht et al. 2010; Leng et al. 2015). GDM may lead to serious adverse maternal outcomes such as high cesarean section rate and preeclampsia and to detrimental infant outcomes such as macrosomia, infant respiratory distress syndrome, and neonatal hypoglycemia (Poel et al. 2012). GDM also increases the long-term risks of type 2 diabetes mellitus (Bellamy et al. 2009), obesity, and metabolic syndrome for both mothers and infants (Metzger 2007). It is noteworthy that pregnancy women are at high risk of glucose intolerance because of the insulin-desensitizing effects of hormonal products of the placenta (Buchanan and Xiang 2005). Some characteristics such as maternal age and high prepregnancy body mass index (BMI) (Leng et al. 2015) have been well-established as important risk factors. However, there is increasing evidence indicating that GDM might be caused by environmental chemical exposures, which has earned less attention than the traditional risk factors (Ehrlich et al. 2016; Shapiro et al. 2015).

Metals such as cadmium (Cd) and arsenic (As) have been reported to be involved in the etiology of type 2 diabetes or GDM in previous studies (Beck et al. 2017; Edwards and Ackerman 2016; Farzan et al. 2016). Cd is a toxic heavy metal that is widely used in batteries, pigments, coatings and plating, and stabilizers for plastics, among other applications (ATSDR 2012). Cd, which has a high soil-to-plant transfer rate, is easily emitted to the environment by nonferrous metal mining and refining, manufacturing and application of phosphate fertilizers, fossil fuel combustion, and waste incineration and disposal (ATSDR 2012). Cd enters the human body mainly through smoking and food ingestion, and the diet is the main source of environmental Cd exposure in nonsmokers in most parts of the world (Järup and Akesson 2009; Satarug et al. 2010). Absorbed Cd accumulates mainly in the liver and the kidney (Orłowski and Piotrowski 2003) and is primarily eliminated from the body in urine (ATSDR 2012). As is a naturally occurring element that is widely distributed in the earth’s crust, and people are exposed to As worldwide (ATSDR 2007). Simultaneous exposure to Cd and As is common in real-world exposure scenarios (Tchounwou et al. 2012). In animal studies, Cd and As have been demonstrated to have diabetogenic effects, such as damage to pancreatic βcells and impairment of insulin secretion (Edwards and Ackerman 2016; Hectors et al. 2011). A growing number of population-based studies have investigated the association between Cd body burden and type 2 diabetes in the general population. Some (Afridi et al. 2008, 2013; Haswell-Elkins et al. 2007; Kolachi et al. 2011; Li et al. 2017; Schwartz et al. 2003; Son et al. 2015; Wallia et al. 2014), but not all (Barregard et al. 2013; Liu et al. 2016; Moon 2013; Nie et al. 2016; Swaddiwudhipong et al. 2010, 2012, 2015), studies have suggested that an increased risk of diabetes is associated with higher Cd exposure. However, evidence regarding the association of Cd body burden with GDM is limited. Only two previous studies have investigated the association between Cd and GDM: A nested case–control study in China suggested that Cd residues in meconium, which were measured after delivery, were associated with GDM risk (Peng et al. 2015), but a study from Canada using Cd levels in maternal blood collected in the first trimester did not find such an association (Shapiro et al. 2015). However, the Cd concentration in blood is believed to reflect recent exposure, whereas urinary Cd more approximately reflects total body burden; the half-life of Cd is 6–38 y (ATSDR 2012). Evidence from epidemiologic studies has also suggested the association of GDM with As exposure (Ettinger et al. 2009; Farzan et al. 2016; Peng et al. 2015), even at relatively low exposure levels (Shapiro et al. 2015).

In the present study, we collected early-pregnancy urine samples from 2,026 pregnant women and examined whether an increased risk of GDM was associated with higher urinary Cd levels, which provide a better estimation of Cd body burden. We also investigated whether the association between Cd exposure and GDM was modified by fetal sex because women carrying male fetuses have been reported to have a higher risk of GDM than those carrying female fetuses (Jaskolka et al. 2015). We also examined the potential interaction between Cd and As exposure on GDM.

Methods

Study Population

The present study was conducted between October 2013 and April 2016 at Wuhan Women and Children’s Medical Care Center, a major tertiary medical center in Wuhan, China. A total of 2,592 pregnant women who met the following criteria were recruited: a) <16 wk of pregnancy with a singleton gestation at the time of enrollment; b) resident of Wuhan City; c) willing to have prenatal care and give birth at the study hospital. In this study, five women were excluded because of their family histories of type 2 diabetes, four were excluded because they had type 2 diabetes before pregnancy, 338 were excluded because they did not donate urine samples, and 219 were excluded because they did not take oral glucose tolerance tests (OGTTs). We only included the first delivery records for women who had two separate deliveries. All participants provided written informed consent at enrollment. The research protocol was approved by the ethics committees of Tongji Medical College, Huazhong University of Science and Technology [No. (2012)07], and Wuhan Women and Children’s Medical Care Center (No. 2012003).

Urine Sample Collection and Cd Measurements

Spot urine samples were collected at 13 wk of gestation, on average [range 8–14 wk, standard deviation (SD)=0.68], and stored in polypropylene tubes at 20°C for further analysis. Urinary Cd levels were determined by inductively coupled plasma mass spectrometry (ICP-MS) (Agilent 7700, Agilent Technologies). Urinary total As levels were also measured. Urine samples were brought to room temperature (20–25°C) before analysis. After being thoroughly vortex mixed, urine samples were acidified with 1.2% (v/v) nitric acid (HNO3). The resulting samples were digested at 40°C for 1 h, and then Cd and As were analyzed. The operating parameters for ICP-MS were as follows: radio frequency (RF) power, 1,550 W; auxiliary gas flow, 0.8 L/min; carrier gas flow, 0.8 L/min; plasma gas flow, 15.00 L/min; resolution (peak high 10%), 0.65–0.80 amu; sample uptake rate, 0.4 mL/min; unimodal residence time, 0.99 s; repetitions, 3 times.

We implemented stringent laboratory quality controls to ensure the accuracy of the analyses. An external quality-control sample (SRM2670a Toxic Elements in Urine, a standard reference material from the National Institute of Standards and Technology, Gaithersburg, MD, USA) was analyzed with every batch to improve the accuracy of the measurements, and the concentrations measured were within the certified range (5%). The limit of detection (LOD) values for Cd and As were 0.001 and 0.018 μg/L, respectively. Two samples were below the LOD for Cd, and one sample was below the LOD for As. These three samples were assigned a value of one-half LOD for the analyses. For Cd and As, the between-assay coefficients of variation (CVs) were 0.36% and 0.28%, respectively, and the within-assay CVs were 0.89% and 0.68%, respectively.

Concentrations of Cd and As were adjusted for variation in dilution by urinary specific gravity (SG) according to the following formula: Pc=Pi[(SGm1)/(SGi1)], where Pc=specific gravity(SG)adjusted metabolite concentration (μg/L), Pi=observed metabolite concentration, SGi=SG of the urine sample, and SGm=median SG of the cohort (SGm=1.014) (Just et al. 2010). SG was measured using a pocket refractometer while preparing urinary samples for Cd and As analysis (Atago PAL-10S; Atago).

Data Collection

Standard face-to-face interviews were conducted by trained nurses within three days before or after delivery to collect information on sociodemographic characteristics (e.g., maternal age, occupation, and education) and lifestyle habits during pregnancy (e.g., smoking, passive smoking, and alcohol consumption). Passive smoking was defined as exposure of nonsmoking women to tobacco smoke during pregnancy (from her family or other people in the household or workplace) (Vardavas et al. 2016). Information on history of pregnancy complications and fetal sex was obtained from medical records. Gestational age was calculated based on the last menstrual period. Prepregnancy BMI was calculated from prepregnancy body weight and height, where self-reported prepregnancy body weight was extracted from records of the first prenatal visit to the hospital, and maternal height was measured using a stadiometer in the hospital.

GDM Diagnosis

One-step GDM screening, using a 75-g OGTT, was routinely administered between gestational weeks 24 and 28 in the study hospital. Women were diagnosed with GDM according to International Association of Diabetes and Pregnancy Study Group (IADPSG) recommendations (American Diabetes Association 2011): fasting plasma glucose (FPG) 5.1mmol/L (92 mg/dL), 1-h plasma glucose (1-h PG) 10.0mmol/L (180 mg/dL), or 2-h plasma glucose (2-h PG) 8.5mmol/L (153 mg/dL).

Statistical Analysis

The distribution of Cd concentrations was examined, and the Wilcoxon rank test was used to compare Cd concentrations between women with and without GDM owing to the left-skewed distribution of Cd concentrations. We compared the frequency distributions of sociodemographic and lifestyle characteristics between women with and without GDM. The associations between risk of GDM and SG-corrected urinary Cd concentrations were evaluated by calculating risk ratios (RRs) and 95% confidence intervals (CIs) using Poisson regression with a robust error variance with generalized estimating equations (GEE) estimation (Zou 2004). Models were fit using SG-corrected urinary Cd concentrations as categorical variables based on tertile distribution of urinary Cd in all women, and the bottom tertile was assigned as the reference. We conducted trend tests using the median value within each tertile of urinary Cd as the score variable and evaluated the statistical significance of this predictor using the Wald test. We further examined the associations between urinary Cd levels [log10transformed SG-corrected urinary Cd concentrations (Log10Cd)] and continuous plasma glucose (PG) concentrations (mmol/L) using multiple linear regression. A composite OGTT measure, which was the sum of PG z-scores for FPG, 1-h PG, and 2-h PG, was also used as an outcome variable. The z-score was calculated by subtracting the mean from each woman’s glucose measurements in our study and dividing by the corresponding standard deviation (Lowe et al. 2012). Bivariate summary analyses were conducted for all variables. Inclusion of covariates in final multivariable models was based on a) covariates associated with GDM in bivariate analyses (p0.1) and b) a priori knowledge of the associations with Cd levels and GDM.

Maternal age (<25 y, 25–29 y, 30–34 y, 35 y), maternal education (more than high school, high school, less than high school), parity (primiparous, multiparous), prepregnancy BMI (<18.5kg/m2, 18.523.9kg/m2, 24.0kg/m2), hypertensive disorder in pregnancy, fetal sex (Peng et al. 2015), and passive smoking (Shaham et al. 1996) during pregnancy were included in models. Some studies have suggested that As exposure is associated with risk of GDM (Farzan et al. 2016; Shapiro et al. 2015); therefore, we included SG-corrected As levels (<17.12μg/LSG, 17.1228.62μg/LSG, 28.62μg/LSG) in the final models. Because pregnant women with GDM were more likely to be multiparous than non-GDM women in our study (p=0.01, shown in Table 1), we further conducted stratified analyses by parity.

Table 1. Selected characteristics of study population [n (%)].
Characteristic Total (n=2,026) Non-GDM (n=1,828) GDM (n=198) p-Valuea
Age (y) <0.01
<25 152 (7.50) 143 (7.82) 9 (4.55)
 25–29 1,226 (60.52) 1,125 (61.56) 101 (51.01)
 30–34 526 (25.96) 464 (25.37) 62 (31.31)
35 122 (6.02) 96 (5.25) 26 (13.13)
Prepregnancy BMI (kg/m2) <0.01
<18.5 366 (18.07) 350 (19.15) 16 (8.08)
 18.5–23.9 1,380 (68.31) 1,262 (69.04) 122 (61.62)
24 276 (13.62) 216 (11.81) 60 (30.30)
Parity 0.01
 Primiparous 1,763 (87.02) 1,602 (87.64) 161 (81.31)
 Multiparous 263 (12.98) 226 (12.36) 37 (18.69)
Education 0.49
 More than high school 1,645 (81.19) 1,490 (81.51) 155 (78.28)
 High school 282 (13.92) 249 (13.62) 33 (16.67)
 Less than high school 99 (4.89) 89 (4.87) 10 (5.05)
Occupation 0.53
 Employed 1,356 (66.93) 1,228 (67.18) 128 (64.65)
 Unemployed 664 (32.77) 594 (32.49) 70 (35.35)
 Missing 6 (0.30) 6 (0.33) 0 (0.00)
Passive smoking during pregnancy 0.76
 No 1,450 (71.57) 1,310 (71.68) 140 (70.71)
 Yes 574 (28.33) 516 (28.21) 58 (29.29)
 Missing 2 (0.1) 2 (0.11) 0 (0.0)
Hypertensive disorder in pregnancy 0.12
 No 1,953 (96.40) 1,766 (96.61) 187 (94.44)
 Yes 73 (3.60) 62 (3.39) 11 (5.56)
Fetal sex 0.90
 Male 1,066 (52.62) 961 (52.57) 105 (53.03)
 Female 960 (47.38) 867 (47.43) 93 (46.97)

Note: BMI, body mass index; GDM, gestational diabetes mellitus.

ap-Values for difference according to chi-squared test.

To evaluate the potential effect modification by fetal sex, we used two approaches: a) We stratified the models by fetal sex, and b) we included an interaction term between Cd and fetal sex in the models, where Cd was modeled in both as tertiles. In addition, because prepregnancy BMI is known to be related to the risk of GDM (Leng et al. 2015), we conducted similar interaction analyses to explore the effect modification of prepregnancy BMI on Cd-GDM association. We assessed the interaction between Cd and As on GDM by including a cross-product term of Cd and As in the model, where Cd and As were categorized into high and low levels based on the median values of their concentrations. Owing to the potential contribution of tobacco smoke to Cd exposure (ATSDR 2012), we also performed a sensitivity analysis excluding women who were exposed to passive smoke during pregnancy. All statistical analyses were performed using SAS (version 9.4; SAS Institute Inc.). Two-sided p<0.05 was considered statistically significant. All presented CIs were calculated at the 95% level.

Results

Among the 2,026 participants, 198 (9.8%) women were diagnosed with GDM. Pregnant women with GDM were older (29.9 vs. 28.5 y), had greater prepregnancy BMI (22.4 vs. 20.8kg/m2), and were more likely to be multiparous (18.7% vs.12.4%), than the non-GDM women. There were no significant differences between women with and without GDM in educational attainment, occupational status, passive smoking, hypertensive disorder in pregnancy, and fetal sex (Table 1). Six women were missing occupation data, and two women were missing passive smoking data. No women reported smoking or alcohol consumption during pregnancy in this study.

The geometric means (GMs) of Cd and SG-corrected Cd concentrations in maternal urine of all pregnant women were 0.59 μg/L and 0.67 μg/L, respectively (Figure 1). The median value (interquartile range) of urinary Cd concentrations among all women was 0.62 (0.33–1.08) μg/L. There was no significant difference in SG-corrected Cd concentrations between women with (GM=0.66 μg/L) and without GDM (GM=0.73 μg/L).

Figures 1A and 1B are plots showing Cd concentrations in micrograms per liter and SG-corrected Cd concentrations in micrograms per liter, respectively (y-axis), among total number of women, women with GDM, and women without GDM (x-axis).
Figure 1. Distributions of urinary cadmium (Cd) and specific gravity (SG)-corrected Cd concentrations among total women and women with or without gestational diabetes mellitus (GDM). (A) Urinary Cd concentrations; (B) SG-corrected Cd concentrations. Squares represent median values. Triangles and diamonds represent the 25th and 75th percentiles. Solid vertical lines span the fifth to 95th percentiles. The geometric means (confidence interval) of Cd and SG-corrected Cd are 0.59 (0.48–0.70) and 0.67(0.58–0.76) μg/L, respectively.

There was a significant increase in the risk of GDM across increasing tertiles of SG-corrected Cd in crude GEE models [crude RR: low=1; medium=1.11 (95% CI: 0.77, 1.59); high=1.43 (95% CI: 1.02, 2.01); pfortrend=0.02]. After adjustment for a range of potential confounders (maternal age, maternal education, parity, prepregnancy BMI, hypertensive disorder in pregnancy, passive smoking, and fetal sex), the association between maternal Cd and GDM was slightly attenuated, but it remained borderline significant for the top tertile of Cd [adjusted RR: low=1; medium=1.04 (95% CI: 0.74, 1.44); high=1.36 (95% CI: 0.98, 1.90); pfortrend=0.05] (Table 2).

Table 2. Associations between maternal urinary Cd levels and GDM.
Cd concentrations (μg/LSG) GDM/Total RR (95% CI)a RR (95% CI)b
All (n=2,026)
 Low (<0.51) 56/676 1.00 1.00
 Medium (0.51–0.86) 62/675 1.11 (0.77, 1.59) 1.04 (0.74, 1.44)
 High (0.86) 80/675 1.43 (1.02, 2.01) 1.36 (0.98, 1.90)
p for trendc 0.02 0.05
Male (n=1,066)
 Low (<0.51) 25/350 1.00 1.00
 Medium (0.51–0.86) 31/363 1.20 (0.72, 1.98) 1.14 (0.70, 1.87)
 High (0.86) 49/353 1.94 (1.23, 3.07) 1.86 (1.14, 2.93)
p for trendc <0.01 0.01
Female (n=960)
 Low (<0.51) 31/326 1.00 1.00
 Medium (0.51–0.86) 31/312 1.04 (0.65, 1.68) 0.97 (0.61, 1.55)
 High (0.86) 31/322 1.01 (0.63, 1.63) 0.98 (0.60, 1.60)
p for trendc 0.99 0.94
p for interactiond 0.05 0.03

Note: Cd, cadmium; CI, confidence interval; GDM, gestational diabetes mellitus; RR, risk ratio; SG, specific gravity.

aUnadjusted risk ratio.

bAdjusted for maternal age, education, maternal prepregnancy body mass index (BMI), parity, passive smoking, total arsenic level, and hypertensive disorder in pregnancy. Estimates for all women were also adjusted for fetal sex.

cp-Values for trend were derived using a continuous variable with the median value of each tertile.

dp-Values for the interaction term between maternal urinary Cd and fetal sex.

Given the evidence that fetal sex is associated with risk of GDM, we explored the effect modification by fetal sex on the association between urinary Cd and GDM. Among women carrying male fetuses, the risk of GDM increased with increasing tertiles of Cd [adjusted RR for the top vs. bottom tertile=1.86 (95% CI: 1.14, 2.93); ptrend=0.01], whereas there was no association between Cd and GDM among women carrying female fetuses [adjusted RR for the top vs. bottom tertile=0.98 (95% CI: 0.61, 1.60); ptrend=0.94] (Table 2). The interaction between urinary Cd and fetal sex on the risk of GDM was significant (pforinteraction=0.03) (Table 2).

Among women with normal prepregnancy BMI (18.523.9kg/m2, n=1,384), adjusted RRs increased with increasing Cd tertile, with a significant association for the top vs. bottom tertile [adjusted RR=1.62 (95% CI: 1.04, 2.53); ptrend=0.03] (Table 3). Estimates were imprecise for the high-BMI group (n=276) and without a consistent trend [top vs. bottom tertile adjusted RR=1.14 (95% CI: 0.64, 2.04); ptrend=0.40). The interaction between BMI and Cd was not significant (p=0.42).

Table 3. Associations between maternal urinary Cd levels and GDM stratified by prepregnancy BMI (kg/m2).
Cd concentrations (μg/LSG) GDM/Total RR (95% CI)a RR (95% CI)b
18.5<BMI23.9 (n=1,384)
 Low (<0.51) 31/455 1.00 1.00
 Medium (0.51–0.86) 39/465 1.25 (0.77, 2.04) 1.21 (0.77, 1.91)
 High (0.86) 52/464 1.73 (1.08, 2.75) 1.62 (1.04, 2.53)
p for trendc 0.02 0.03
BMI24.0 (n=276)
 Low (<0.51) 19/88 1.00 1.00
 Medium (0.51–0.86) 16/90 0.82 (0.45, 1.50) 0.70 (0.39, 1.25)
 High (0.86) 25/98 1.18 (0.70, 1.99) 1.14 (0.64, 2.04)
p for trendc 0.34 0.40
p for interactiond 0.36 0.42

Note: BMI, body mass index; Cd, cadmium; CI, confidence interval; GDM, gestational diabetes mellitus; RR, risk ratio; SG, specific gravity.

aUnadjusted risk ratio.

bAdjusted for maternal age, education, parity, passive smoking, fetal sex, total arsenic level, and hypertensive disorder in pregnancy.

cp-Values for trend were derived using a continuous variable with the median value of each tertile.

dp-Values for the interaction term between maternal urinary Cd and prepregnancy BMI.

We also investigated the association between continuous urinary Cd levels (Log10Cd) and PG concentrations. Among the overall study population, we found significant associations between Log10Cd and 2-h PG [β=0.18 (95% CI: 0.04, 0.33)], as well as the PG z-score sum [β=0.37 (95% CI: 0.09, 0.64)]. No significant associations were observed between Log10Cd and FPG or 1-h PG. Estimated associations with Log10Cd and 2-h PG or PG z-score sum were similar for women carrying male fetuses and women carrying female fetuses, although the estimates were statistically significant only for the larger subgroup of women carrying male fetuses (see Figure S1). When stratified by prepregnancy BMI, associations between Log10Cd and normal-weight women’s 2-h PG and PG z-score sum were similar for women with normal and high BMI but were statistically significant only for the larger group of normal-BMI women (see Figure S2).

We also explored whether the Cd-GDM associations were different between nulliparous and multiparous women by stratified analyses. No significant associations of Cd with GDM were found in either stratum. However, the RRs for GDM appeared larger in multiparous women than in nulliparous women (see Table S1). Estimates from a model including a cross-product term for Cd and As were consistent with a synergistic effect (i.e., a stronger association than expected was observed for the combined exposures based on the observed associations with Cd only and As only), although the interaction was not significant (p=0.09) (see Table S2). In the sensitivity analysis that excluded women with passive smoking during pregnancy, the risk of GDM still increased across increasing tertiles of SG-corrected Cd levels, although the associations did not reach statistical significance (see Table S3).

Discussion

In this study, we observed marginal associations between Cd body burden and the risk of GDM in the overall study population. Among women with male fetuses, the relative risk for those in the top versus bottom tertile of urinary Cd was 1.86 (95% CI: 1.14, 2.93). The association between higher urinary Cd and risk of GDM appeared to be limited to women with normal prepregnancy BMI, although differences between women with high versus normal BMI were not significant.

We are aware of only two published studies addressing the association of Cd exposure with GDM (Peng et al. 2015; Shapiro et al. 2015). Our study is most similar to the study from Canada in its prospective study design (Shapiro et al. 2015). Compared with their population, urinary Cd levels in our population were higher: GM in normal-glucose women=0.2 μg/L and GM in GDM women=0.3 μg/L (Shapiro et al. 2015) versus GM in normal-glucose women=0.66 μg/L and GM in GDM women=0.73 μg/L (the present study). Although Shapiro et al. (2015) did not suggest an association between blood Cd level in the first trimester and increased risk of GDM, the relationship remained borderline significant after adjustment for confounders [crude odds ratio (OR)=2.9 (95% CI: 1.2, 7.0); adjusted OR=2.5 (95% CI: 1.0, 6.4) for highest vs. lowest quartile]. In the present study, we also found a marginally significant association of Cd level with the risk of GDM after adjustment for confounders. A nested case–control study suggested that Cd residues in meconium were associated with a high risk of GDM [adjusted OR=16.87 (95% CI: 4.19, 67.86) for the third quartile; adjusted OR=11.95 (95% CI: 2.97, 48.04) for the highest quartile] (Peng et al. 2015). However, those authors assessed maternal Cd exposure by Cd in meconium, which was measured after the outcome measurement. Moreover, Cd is reported to largely accumulate in the placenta; thus, Cd levels in meconium may not be an accurate reflection of total body burden (Vilahur et al. 2015). In our study, we further examined associations of Cd levels with continuous PG concentrations. Among the three glucose measures (FPG, 1-h PG, and 2-h PG), only 2-h PG had a significant association with Cd level. However, the PG z-score sum was also significantly correlated with Cd level. We speculate that might be the case because PG z-score sum is a composite measure of PG, to some extent similar to the IADPSG criteria.

More studies have addressed associations of Cd exposure with type 2 diabetes than with GDM. Although study findings are inconsistent, most suggest that Cd exposure would increase the risk of type 2 diabetes. In an analysis of NHANES III (Third National Health and Nutrition Examination Survey, 1988–1994) data, urinary Cd levels were associated with impaired fasting glucose and diabetes in a dose-dependent manner [OR=1.24 (95% CI: 1.06, 1.45) for the middle versus bottom tertile; OR=1.45 (95% CI: 1.07, 1.97) for the top versus bottom tertile; p for trend <0.0001] (Schwartz et al. 2003). Another study (Wallia et al. 2014) of NHANES (2005–2010) reached the same conclusion [OR=1.67 (95% CI: 1.12, 2.47) for the highest quintile of Cd levels]. A recent meta-analysis also supported the positive association between Cd exposure and risk of type 2 diabetes (Li et al. 2017).

Potential mechanisms for associations between Cd and GDM are not clear. Cd is reported to induce pancreatic βcell death via increased reactive oxygen species (Chang et al. 2013). Notably, βcells normally increase their insulin secretion to compensate for insulin resistance during pregnancy (Buchanan and Xiang 2005). βcells undergo hyperplasia, resulting in increases in both insulin secretion and insulin sensitivity in early pregnancy, followed by progressive insulin resistance. In addition, we found evidence of effect modification on the association between Cd and GDM by fetal sex, and the association was more evident and stronger among women carrying male fetuses. To the best of our knowledge, no other study has explored effect modification on the Cd-GDM association by fetal sex. Women carrying male fetuses have been reported to have lower βcell function and higher risk of GDM than women carrying female fetuses (Jaskolka et al. 2015; Retnakaran et al. 2015), and it has been proposed that this phenomenon might be explained by differences in the influence of male versus female fetuses on placental secretion of hormones or proteins involved in βcell compensation (Retnakaran et al. 2015). Given the potential damage to βcells by Cd, we cautiously speculate that pancreatic βcells of women carrying male fetuses might be more vulnerable to Cd exposure. Future studies are needed to clarify the influence of fetal sex on the Cd-GDM relationship, as well as the underlying mechanisms.

Evidence of effect modification by prepregnancy BMI was inconclusive, in part because of the relatively small number of women with high prepregnancy BMI. However, associations between urinary Cd and GDM appeared to be limited to normal-weight women, perhaps as a consequence of the higher baseline risk of GDM in obese women compared with normal weight women, which might negate or obscure a smaller effect of Cd exposure (Li et al. 2014). However, potential differences in the association of Cd with GDM between normal- and high-BMI women need to be confirmed in a larger study population. We also performed sensitivity analyses that excluded women who were exposed to passive smoke during pregnancy. The risk of GDM still increased across increasing Cd tertiles, although the observed associations were weakened and became nonsignificant, which might be due to the reduced sample size; another reason for the weakening of the association might be that the women without passive smoking who remained in the analysis had relatively low Cd exposure. Finally, we found preliminary evidence of a synergistic effect of Cd and As on the risk of GDM. In vitro, Cd (Chang et al. 2013; El Muayed et al. 2012) and As (Lu et al. 2011; Yang et al. 2012) have both been shown to cause oxidative stress, impaired glucose-stimulated insulin release, and pancreatic βcell death, which could lead to diabetes (Edwards and Ackerman 2016). However, although a synergistic effect of As and Cd is plausible, future studies are needed to confirm our findings.

Our study has several strengths. First, this prospective cohort study enabled us to assess Cd exposure during early pregnancy, before the onset of GDM, which helped us to avoid the potential bias caused by misclassification of Cd exposure. Second, urinary Cd provides a better reflection of the Cd body burden than blood Cd. We excluded women with type 2 diabetes, which may lead to diabetes-related changes in renal function, thereby increasing urinary excretion of Cd. Third, the interviews, medical record abstraction, and urinary total As concentration provided extensive data on potential confounders. Nevertheless, our study has some limitations. First, we did not account for the impact of micronutrient (i.e., iron, zinc and calcium) intake because participants’ nutritional levels were not registered. Second, the interviews were conducted at delivery, which was after the diagnosis of GDM, which may have led to measurement error in the confounders. Third, although we excluded pregnant women with a family history of diabetes from the study population, the information on family history of diabetes was self-reported and may not have been completely accurate. Finally, some unmeasured or unknown coexposure may contribute to the risk of GDM, although we did control urinary total As.

Conclusion

Our findings suggest that Cd body burden may be a potential risk factor for GDM and that the association may be modified by fetal sex. However, additional studies are needed to confirm these findings in other study populations.

Acknowledgments

We thank all the participants in the study and all collaborators in the study hospital.

This work was supported by the National Natural Science Foundation of China (21437002, 81372959, 81402649, and 91643207), the National Key Research and Development Plan (2016YFC0206203 and 2016YFC0206700), and the Fundamental Research Funds for the Central Universities, Huazhong University of Science and Technology (HUST) (2016YXZD043).

References

Afridi HI, Kazi TG, Brabazon D, Naher S, Talpur FN. 2013. Comparative metal distribution in scalp hair of Pakistani and Irish referents and diabetes mellitus patients. Clin Chim Acta 415:207–214, PMID: 23123286, 10.1016/j.cca.2012.10.029.

Afridi HI, Kazi TG, Kazi N, Jamali MK, Arain MB, Jalbani N. 2008. Evaluation of status of toxic metals in biological samples of diabetes mellitus patients. Diabetes Res Clin Pract 80(2):280–288, PMID: 18276029, 10.1016/j.diabres.2007.12.021.

Albrecht SS, Kuklina EV, Bansil P, Jamieson DJ, Whiteman MK, Kourtis AP, et al. 2010. Diabetes trends among delivery hospitalizations in the U.S., 1994–2004. Diabetes Care 33(4):768–773, PMID: 20067968, 10.2337/dc09-1801.

American Diabetes Association, 2011. Diagnosis and classification of diabetes mellitus. Diabetes Care 34(suppl1):S62–S69, PMID: 21193628, 10.2337/dc11-S062.

ATSDR (Agency for Toxic Substances and Disease Registry). 2007. Toxic substances portal–arsenic. https://www.atsdr.cdc.gov/phs/phs.asp?id=18&tid=3 [accessed 13 December 2017].

ATSDR. 2012. Toxic substances portal–cadmium. http://www.atsdr.cdc.gov/ToxProfiles/tp.asp?id=48&tid=15 [accessed 4 August 2017].

Barregard L, Bergström G, Fagerberg B. 2013. Cadmium exposure in relation to insulin production, insulin sensitivity and type 2 diabetes: a cross-sectional and prospective study in women. Environ Res 121:104–109, PMID: 23261793, 10.1016/j.envres.2012.11.005.

Beck R, Styblo M, Sethupathy P. 2017. Arsenic exposure and type 2 diabetes: microRNAs as mechanistic links?. Curr Diab Rep 17(3):18, PMID: 28275977, 10.1007/s11892-017-0845-8.

Bellamy L, Casas JP, Hingorani AD, Williams D. 2009. Type 2 diabetes mellitus after gestational diabetes: a systematic review and meta-analysis. Lancet 373(9677):1773–1779, PMID: 19465232, 10.1016/S0140-6736(09)60731-5.

Buchanan TA, Xiang AH. 2005. Gestational diabetes mellitus. J Clin Invest 115(3):485–491, PMID: 15765129, 10.1172/JCI24531.

Chang KC, Hsu CC, Liu SH, Su CC, Yen CC, Lee MJ, et al. 2013. Cadmium induces apoptosis in pancreatic β-cells through a mitochondria-dependent pathway: the role of oxidative stress-mediated c-Jun N-terminal kinase activation. PloS One 8(2):e54374, PMID: 23405080, 10.1371/journal.pone.0054374.

Edwards J, Ackerman C. 2016. A review of diabetes mellitus and exposure to the environmental toxicant cadmium with an emphasis on likely mechanisms of action. Curr Diabetes Rev 12(3):252–258, PMID: 26264451.

Ehrlich S, Lambers D, Baccarelli A, Khoury J, Macaluso M, Ho SM. 2016. Endocrine disruptors: a potential risk factor for gestational diabetes mellitus. Am J Perinatol 33(13):1313–1318, PMID: 27490770, 10.1055/s-0036-1586500.

El Muayed M, Raja MR, Zhang X, MacRenaris KW, Bhatt S, Chen X, et al. 2012. Accumulation of cadmium in insulin-producing β cells. Islets 4(6):405–416, PMID: 23466887, 10.4161/isl.23101.

Ettinger AS, Zota AR, Amarasiriwardena CJ, Hopkins MR, Schwartz J, Hu H, et al. 2009. Maternal arsenic exposure and impaired glucose tolerance during pregnancy. Environ Health Perspect 117(7):1059–1064, PMID: 19654913, 10.1289/ehp0800533.

Farzan SF, Gossai A, Chen Y, Chasan-Taber L, Baker E, Karagas M. 2016. Maternal arsenic exposure and gestational diabetes and glucose intolerance in the New Hampshire Birth Cohort Study. Environ Health 15(1):106, PMID: 27825389, 10.1186/s12940-016-0194-0.

Haswell-Elkins M, Imray P, Satarug S, Moore MR, O’Dea K. 2007. Urinary excretion of cadmium among Torres Strait Islanders (Australia) at risk of elevated dietary exposure through traditional foods. J Expo Sci Environ Epidemiol 17(4):372–377, PMID: 16912696, 10.1038/sj.jes.7500520.

Hectors TL, Vanparys C, van der Ven K, Martens GA, Jorens PG, Van Gaal LF, et al. 2011. Environmental pollutants and type 2 diabetes: a review of mechanisms that can disrupt beta cell function. Diabetologia 54(6):1273–1290, PMID: 21442161, 10.1007/s00125-011-2109-5.

Järup L, Akesson A. 2009. Current status of cadmium as an environmental health problem. Toxicol Appl Pharmacol 238(3):201–208, PMID: 19409405, 10.1016/j.taap.2009.04.020.

Jaskolka D, Retnakaran R, Zinman B, Kramer CK. 2015. Sex of the baby and risk of gestational diabetes mellitus in the mother: a systematic review and meta-analysis. Diabetologia 58(11):2469–2475, PMID: 26253767, 10.1007/s00125-015-3726-1.

Just AC, Adibi JJ, Rundle AG, Calafat AM, Camann DE, Hauser R, et al. 2010. Urinary and air phthalate concentrations and self-reported use of personal care products among minority pregnant women in New York city. J Expo Sci Environ Epidemiol 20(7):625–633, PMID: 20354564, 10.1038/jes.2010.13.

Kolachi NF, Kazi TG, Afridi HI, Kazi N, Khan S, Kandhro GA, et al. 2011. Status of toxic metals in biological samples of diabetic mothers and their neonates. Biol Trace Elem Res 143(1):196–212, PMID: 20963639, 10.1007/s12011-010-8879-7.

Leng J, Shao P, Zhang C, Tian H, Zhang F, Zhang S, et al. 2015. Prevalence of gestational diabetes mellitus and its risk factors in Chinese pregnant women: a prospective population-based study in Tianjin, China. PloS One 10(3):e0121029, PMID: 25799433, 10.1371/journal.pone.0121029.

Li W, Zhang S, Liu H, Wang L, Zhang C, Leng J, et al. 2014. Different associations of diabetes with β-cell dysfunction and insulin resistance among obese and nonobese Chinese women with prior gestational diabetes mellitus. Diabetes Care 37(9):2533–2539, PMID: 24914241, 10.2337/dc14-0573.

Li Y, Zhang Y, Wang W, Wu Y. 2017. Association of urinary cadmium with risk of diabetes: a meta-analysis. Environ Sci Pollut Res Int 24(11):10083–10090, PMID: 28233200, 10.1007/s11356-017-8610-8.

Liu B, Feng W, Wang J, Li Y, Han X, Hu H, et al. 2016. Association of urinary metals levels with type 2 diabetes risk in coke oven workers. Environ Pollut 210:1–8, PMID: 26689646, 10.1016/j.envpol.2015.11.046.

Lowe LP, Metzger BE, Dyer AR, Lowe J, McCance DR, Lappin TR, et al. 2012. Hyperglycemia and Adverse Pregnancy Outcome (HAPO) Study: associations of maternal A1C and glucose with pregnancy outcomes. Diabetes Care 35(3):574–580, PMID: 22301123, 10.2337/dc11-1687.

Lu TH, Su CC, Chen YW, Yang CY, Wu CC, Hung DZ, et al. 2011. Arsenic induces pancreatic β-cell apoptosis via the oxidative stress-regulated mitochondria-dependent and endoplasmic reticulum stress-triggered signaling pathways. Toxicol Lett 201(1):15–26, PMID: 21145380, 10.1016/j.toxlet.2010.11.019.

Metzger BE. 2007. Long-term outcomes in mothers diagnosed with gestational diabetes mellitus and their offspring. Clin Obstet Gynecol 50(4):972–979, PMID: 17982340, 10.1097/GRF.0b013e31815a61d6.

Moon SS. 2013. Association of lead, mercury and cadmium with diabetes in the Korean population: the Korea National Health and Nutrition Examination Survey (KNHANES) 2009–2010. Diabet Med 30(4):e143–e148, PMID: 23278294, 10.1111/dme.12103.

Nie X, Wang N, Chen Y, Chen C, Han B, Zhu C, et al. 2016. Blood cadmium in Chinese adults and its relationships with diabetes and obesity. Environ Sci Pollut Res Int 23(18):18714–18723, PMID: 27312901, 10.1007/s11356-016-7078-2.

Orłowski C, Piotrowski JK. 2003. Biological levels of cadmium and zinc in the small intestine of non-occupationally exposed human subjects. Hum Exp Toxicol 22(2):57–63, PMID: 12693828, 10.1191/0960327103ht326oa.

Peng S, Liu L, Zhang X, Heinrich J, Zhang J, Schramm K-W, et al. 2015. A nested case–control study indicating heavy metal residues in meconium associate with maternal gestational diabetes mellitus risk. Environ Health 14:19, PMID: 25888735, 10.1186/s12940-015-0004-0.

Poel YH, Hummel P, Lips P, Stam F, van der Ploeg T, Simsek S. 2012. Vitamin D and gestational diabetes: a systematic review and meta-analysis. Eur J Intern Med 23(5):465–469, PMID: 22726378, 10.1016/j.ejim.2012.01.007.

Retnakaran R, Kramer CK, Ye C, Kew S, Hanley AJ, Connelly PW, et al. 2015. Fetal sex and maternal risk of gestational diabetes mellitus: the impact of having a boy. Diabetes Care 38(5):844–851, PMID: 25693837, 10.2337/dc14-2551.

Satarug S, Garrett SH, Sens MA, Sens DA. 2010. Cadmium, environmental exposure, and health outcomes. Environ Health Perspect 118(2):182–190, PMID: 20123617, 10.1289/ehp.0901234.

Schwartz GG, Il’yasova D, Ivanova A. 2003. Urinary cadmium, impaired fasting glucose, and diabetes in the NHANES III. Diabetes Care 26(2):468–470, PMID: 12547882, 10.2337/diacare.26.2.468.

Shaham J, Meltzer A, Ashkenazi R, Ribak J. 1996. Biological monitoring of exposure to cadmium, a human carcinogen, as a result of active and passive smoking. J Occup Environ Med 38(12):1220–1228, PMID: 8978513, 10.1097/00043764-199612000-00007.

Shapiro GD, Dodds L, Arbuckle TE, Ashley-Martin J, Fraser W, Fisher M, et al. 2015. Exposure to phthalates, bisphenol A and metals in pregnancy and the association with impaired glucose tolerance and gestational diabetes mellitus: the MIREC study. Environ Int 83:63–71, PMID: 26101084, 10.1016/j.envint.2015.05.016.

Son HS, Kim SG, Suh BS, Park DU, Kim DS, Yu SD, et al. 2015. Association of cadmium with diabetes in middle-aged residents of abandoned metal mines: the first health effect surveillance for residents in abandoned metal mines. Ann Occup Environ Med 27:20, PMID: 26306202, 10.1186/s40557-015-0071-2.

Swaddiwudhipong W, Limpatanachote P, Mahasakpan P, Krintratun S, Punta B, Funkhiew T. 2012. Progress in cadmium-related health effects in persons with high environmental exposure in northwestern thailand: a five-year follow-up. Environ Res 112:194–198, PMID: 22033168, 10.1016/j.envres.2011.10.004.

Swaddiwudhipong W, Mahasakpan P, Limpatanachote P, Krintratun S. 2010. Correlations of urinary cadmium with hypertension and diabetes in persons living in cadmium-contaminated villages in northwestern Thailand: a population study. Environ Res 110(6):612–616, PMID: 20561611, 10.1016/j.envres.2010.06.002.

Swaddiwudhipong W, Nguntra P, Kaewnate Y, Mahasakpan P, Limpatanachote P, Aunjai T, et al. 2015. Human health effects from cadmium exposure: comparison between persons living in cadmium-contaminated and non-contaminated areas in northwestern Thailand. Southeast Asian J Trop Med Public Health 46(1):133–142, PMID: 26513915.

Tchounwou PB, Yedjou CG, Patlolla AK, Sutton DJ. 2012. Heavy metal toxicity and the environment. EXS 101:133–164, PMID: 22945569, 10.1007/978-3-7643-8340-4_6.

Vardavas CI, Hohmann C, Patelarou E, Martinez D, Henderson AJ, Granell R, et al. 2016. The independent role of prenatal and postnatal exposure to active and passive smoking on the development of early wheeze in children. Eur Respir J 48(1):115–124, PMID: 26965294, 10.1183/13993003.01016-2015.

Vilahur N, Vahter M, Broberg K. 2015. The epigenetic effects of prenatal cadmium exposure. Curr Environ Health Rep 2(2):195–203, PMID: 25960943, 10.1007/s40572-015-0049-9.

Wallia A, Allen NB, Badon S, El Muayed M. 2014. Association between urinary cadmium levels and prediabetes in the NHANES 2005–2010 population. Int J Hyg Environ Health 217(8):854–860, PMID: 25043455, 10.1016/j.ijheh.2014.06.005.

Yang B, Fu J, Zheng H, Xue P, Yarborough K, Woods CG, et al. 2012. Deficiency in the nuclear factor E2-related factor 2 renders pancreatic β-cells vulnerable to arsenic-induced cell damage. Toxicol Appl Pharmacol 264(3):315–323, PMID: 23000044, 10.1016/j.taap.2012.09.012.

Zou G. 2004. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol 159(7):702–706, PMID: 15033648, 10.1093/aje/kwh090.

In Memoriam: James M. Melius, MD, DrPH

PDF icon PDF Version (215 KB)

  • Published: 8 February 2018

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

James M. Melius, MD, DrPH, was an occupational physician and a national and international leader in occupational medicine and epidemiology. He was born in Great Barrington, Massachusetts, in 1948 and died of cardiac arrest in Copake Falls, New York, on 1 January 2018 at the age of 69.

Photograph of James M. MeliusJames M. Melius, 1948–2018.
Image: Courtesy of Laborers’ International Union of North America.

Melius was the principal architect of the James Zadroga 9/11 Health and Compensation Act of 2011, the federal law that supports an extensive program of medical monitoring and health care for first responders, volunteers, and survivors of the attacks on the World Trade Center and the Pentagon, as well as the Shanksville, Pennsylvania, crash site of 11 September 2001. This act also reopened the September 11th Victim Compensation Fund, which has provided more than $3 billion in compensation to injured and ill 9/11 responders and survivors to date. With his knowledge of medicine and health policy, his multiple connections to legislators and labor leaders, and an uncanny sense of political timing, Melius drafted the version of the Zadroga Bill that was successfully passed by Congress in a dramatic lame-duck session in the last days of 2010 and was signed into law by President Barack Obama in January 2011. Melius worked with labor leaders across the United States, especially with the firefighters’, police, and construction workers’ unions; with Senators Kirsten Gillibrand (D), Chuck Schumer (D), and Hillary Clinton (D); and with Representatives Carolyn Maloney (D-Manhattan), Jerrold Nadler (D-Manhattan), and Peter King (R-Long Island) to support passage of this landmark bipartisan legislation.

Melius dedicated his professional life to protecting the health and safety of working men and women through both research and action. From 1980 to 1987, he directed the National Institute for Occupational Safety and Health’s (NIOSH’s) renowned Health Hazards Evaluations and Technical Assistance Branch based in Cincinnati. From 1987 to 1994, he served under Governor Mario Cuomo as director of the Center for Environmental Health in the New York State Department of Health, where he oversaw the establishment of a statewide network of Centers of Excellence in Occupational Health and Safety. This network, the only one of its kind in the United States, continues to this day; it formed the backbone of the medical response to 9/11. From 1994 until his death, he was the administrator of the New York State Laborers’ Health and Safety Trust Fund and the research director for the Laborers’ Health and Safety Fund of North America, organizations affiliated with the Laborers’ International Union of North America.

Melius developed a special relationship with the unions for the building and construction trades and spent much of his career improving safety and health on construction sites. Until the late 1980s, this industrial sector, which contains some of the most hazardous workplaces in America, had been neglected by researchers and policy makers alike. Melius helped end that neglect by assisting the Laborers’ International Union in the development of a unique national program that encompassed not only occupational safety and health but also health promotion activities such as smoking cessation. This was the first—and still the only—national worker protection program to use health insurance premiums to support occupational safety and health. He also forged an agreement between NIOSH and the construction industry to create the National Construction Safety and Health Research Program. As a result of these efforts, 500 fewer workers die each year on construction sites today than in 1990.

In 1983, Melius was appointed chairman of the Medical Advisory Board for the International Association of Fire Fighters. There, he conducted research on occupationally induced hearing loss, chronic obstructive pulmonary disease, asbestosis, and cardiovascular disease in firefighters and was instrumental in securing passage in many states and Canadian provinces of laws that presume cardiac deaths or cancer deaths in firefighters to be occupationally related and therefore deserving of compensation. Melius also championed the development and implementation of medical monitoring programs for first responders across North America.

At the time of his death, Melius was chair of the Presidential Advisory Board on Radiation and Worker Health, which addresses compensation for cancers caused by ionizing radiation in workers employed in nuclear weapons facilities in the United States. In New York City, he was chair of the steering committee for the World Trade Center Medical Monitoring and Treatment Program and was a founding member of the board of directors of 9/11 Health Watch. He served on multiple advisory committees to the New York state and federal governments and to the National Academy of Sciences.

Melius was for many years a fellow, and since 2012 the treasurer, of the Collegium Ramazzini, an independent international society in occupational and environmental health headquartered at the Castello di Bentivoglio near Bologna, Italy. The Collegium is dedicated to the protection of occupational and environmental health; it is named in honor of Bernardino Ramazzini, who was an Italian physician during the 17th century and is considered the father of occupational and environmental medicine. In 2012, Melius received the Collegium Ramazzini’s Irving J. Selikoff Memorial Award in recognition of his “lifetime’s work of protecting working men and women from occupational hazards and his heroic service on behalf of the 9/11 rescue workers.”

Melius graduated from Brown University in 1970 with an AB in biology, obtained an MD from the University of Illinois College of Medicine in Chicago in 1974, and obtained a DrPH degree from the University of Illinois School of Public Health in 1984. His clinical training took place at Cook County Hospital in Chicago; he was board certified in General Preventive Medicine and Occupational Medicine and published extensively in top-ranked occupational and public health journals. He served for 20 years as an adjunct faculty member of the Department of Environmental Medicine and Public Health at the Icahn School of Medicine at Mount Sinai in New York City. He is survived by his wife, Melanie, and two sons, Jeremy and Ehren.

The beneficiaries of Jim Melius’s lifetime of dedication to occupational health and safety are the tens of thousands of workers across America who have been spared injury and premature death because of his work, and the firefighters, police officers, paramedics, construction workers, and volunteers who participated in the rescue, recovery, and rebuilding operations at Ground Zero, the site of the World Trade Center, in the days, weeks, and months after 9/11.

Cleanup in the Gulf: Oil Spill Dispersants and Health Symptoms in Deepwater Horizon Responders

PDF icon PDF Version (1.4 MB)

  • Published: 7 February 2018

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.



Related EHP Article

Respiratory, Dermal, and Eye Irritation Symptoms Associated with Corexit™ EC9527A/EC9500A following the Deepwater Horizon Oil Spill: Findings from the GuLF STUDY

Craig J. McGowan, Richard K. Kwok, Lawrence S. Engel, Mark R. Stenzel, Patricia A. Stewart, and Dale P. Sandler

Cleanup crews applied approximately 1.8 million gallons of chemical oil dispersant to the Gulf of Mexico following the 2010 Deepwater Horizon oil spill.1 A study in Environmental Health Perspectives explores the relationship between potential exposure to the dispersants used and health complaints in a large cohort of workers who participated in the Deepwater Horizon response and cleanup.2 The results show an association between exposure to two chemical dispersants and self-reported health outcomes.

Previous studies have linked oil spill cleanup work to symptoms including eye and throat irritation, respiratory symptoms, and skin rashes.3,4 Yet until now, none attempted to distinguish between potential effects of dispersants versus crude oil exposure.5 The results reported in the new study held even after adjusting for estimated exposures to crude oil.2

Two Corexit™ brand dispersants were used after the Deepwater Horizon spill, EC9527A and EC9500A. Chemical dispersants lower the surface tension between oil and water. They break up oil slicks into smaller droplets that can be more readily dispersed by wind and waves. Although dispersants have been part of the oil spill cleanup arsenal since the 1960s, no previous spill involved the volume of dispersant use seen in the Deepwater Horizon response and cleanup.6

“It is important to understand the potential toxicity of any chemical released in such large volumes to which workers or the public may be exposed,” says Bernard Goldstein, an environmental toxicologist at the University of Pittsburgh Graduate School of Public Health. Goldstein was not involved in the new research. Most of the research on dispersants to date has focused on marine species.6,7

Crewmen pulling an oil-covered boom onto a ship
Crewmen haul an oil-covered boom onto a Coast Guard cutter 18 days after the Deepwater Horizon explosion in April 2010. The GuLF STUDY included more than 30,000 of the Coast Guard personnel, federal employees, contractors, Gulf Coast residents, and volunteers who worked on the Deepwater Horizon response and cleanup. Image courtesy of U.S. Coast Guard.

For the study, between 27,659 and 29,468 oil spill cleanup workers provided responses via a telephone survey about respiratory, skin, and eye irritation symptoms during and after the Deepwater Horizon response. (The number of respondents varied depending on the outcome of interest.) The survey was conducted between 2011 and 2013—one to three years after the spill. All respondents were part of the Gulf Long-term Follow-up Study (GuLF STUDY), an ongoing investigation of potential long-term health outcomes among workers involved in the Deepwater Horizon oil spill cleanup.8

The researchers, led by epidemiologist Dale Sandler of the National Institute of Environmental Health Sciences, used data from the survey to classify participants as exposed to dispersants if they reported having worked directly with chemicals, having worked on a ship from which dispersants were applied, or having spent more than half their time working with any dispersant-related equipment. The exposure measure also took into consideration when and where the participants worked during the cleanup.

After adjusting for estimated co-exposures, including crude oil and other chemical decontaminants, the researchers found that workers with potential exposures to either EC9527A or EC9500A were more likely to report that they experienced certain systems during the cleanup. Dispersant-exposed workers were 61% more likely than unexposed workers to report burning in the nose, throat, or lungs, 58% more likely to report chest tightness, 49% more likely to report eye irritation, and about 40% more likely to report cough or wheeze.

The researchers found a weaker association between potential exposure and symptoms that were present at the time of the telephone survey, many months after the cleanup took place. “Finding that workers with potential exposure to dispersants had acute symptoms during the cleanup was consistent with what we expected based on laboratory studies. It was reassuring that many of those who reported symptoms during the spill no longer had them one to three years later,” Sandler says.

However, she adds, “We were surprised to see that some of the workers who were exposed to dispersants and did not report having symptoms while they were working on the cleanup reported having them when they enrolled in the study.” Sandler explains that people may be better at reporting how they feel in the moment than recalling how they felt in the past, but it is also possible that the association between dispersant exposure and current symptoms is due in part to people overreporting their symptoms.

Sandler points out that it is not clear yet what effect—if any—these symptoms may have on long-term health. “We do not yet know whether exposures during the oil spill will be associated with clinical disease down the road,” she says.

Future research should address “how we can get the most benefit out of dispersants while at the same time minimizing the risk,” Goldstein says. He adds that some attention should be given to the toxicity of mixtures of dispersant and weathered oil. The chemicals and physical properties of oil change as it interacts with sunlight, wind, and microbes in the environment—a process called weathering. “We know very little about the toxicity of these mixtures,” he says.


Lindsey Konkel is a New Jersey–based journalist who reports on science, health, and the environment.

References

1. U.S. Coast Guard, National Response Team. 2011. On Scene Coordinator Report: Deepwater Horizon Oil Spill. Washington, DC:U.S. Dept. of Homeland Security, U.S. Coast Guard.

2. McGowan CJ, Kwok RK, Engel LS, Stenzel MR, Stewart PA, Sandler DP, et al. 2017. Respiratory, dermal, and eye irritation symptoms associated with Corexit™ EC9527A/EC9500A following the Deepwater Horizon oil spill: findings from the GuLF STUDY. Environ Health Perspect 125(9):097015, PMID: 28934097, 10.1289/EHP1677.

3. Aguilera F, Méndez J, Pásaro E, Laffon B. 2010. Review on the effects of exposure to spilled oils on human health. J Appl Toxicol 30(4):291–301, PMID: 20499335, 10.1002/jat.1521.

4. Laffon B, Pásaro E, Valdiglesias V. 2016. Effects of exposure to oil spills on human health: updated review. J Toxicol Environ Health B Crit Rev 19(3–4):105–128, PMID: 27221976, 10.1080/10937404.2016.1168730.

5. Goldstein BD, Osofsky HJ, Lichtveld MY. 2011. The Gulf oil spill. N Engl J Med 364(14):1334–1348, PMID: 21470011, 10.1056/NEJMra1007197.

6. Wise J, Wise JP Sr. 2011. A review of the toxicity of chemical dispersants. Rev Environ Health 26(4):281–300, PMID: 22435326, 10.1515/REVEH.2011.035.

7. Kleindienst S, Seidel M, Ziervogel K, Grim S, Loftis K, Harrison S, et al. 2015. Chemical dispersants can suppress the activity of natural oil-degrading microorganisms. Proc Natl Acad Sci U S A 112(48):14900–14905, PMID: 26553985, 10.1073/pnas.1507380112.

8. Kwok RK, Engel LS, Miller AK, Blair A, Curry MD, Jackson WB II, et al. 2017. The GuLF STUDY: a prospective study of persons involved in the Deepwater Horizon oil spill response and clean-up. Environ Health Perspect 125(4):570–578, PMID: 28362265, 10.1289/EHP715.

Ambient Air Pollution and Chronic Bronchitis in a Cohort of U.S. Women

Author Affiliations open

1Department of Medicine, University of Washington, Seattle, Washington, USA

2Department of Epidemiology, School of Public Health, University of Washington, Seattle, Washington, USA

3Department of Biostatistics, School of Public Health, University of Washington, Seattle, Washington, USA

4Biostatistics and Computational Biology Branch, Division of Intramural Research, National Institute of Environmental Health Sciences, National Institutes of Health, Department of Health and Human Services, Research Triangle Park, North Carolina, USA

5Epidemiology Branch, Division of Intramural Research, National Institute of Environmental Health Sciences, National Institutes of Health, Department of Health and Human Services, Research Triangle Park, North Carolina, USA

6Department of Environmental and Occupational Health Sciences, School of Public Health, University of Washington, Seattle, Washington, USA

PDF icon PDF Version (421 KB)

  • Background:
    Limited evidence links air pollution exposure to chronic cough and sputum production. Few reports have investigated the association between long-term exposure to air pollution and classically defined chronic bronchitis.
    Objectives:
    Our objective was to estimate the association between long-term exposure to particulate matter (diameter <10μm, PM10; <2.5 μm, PM2.5), nitrogen dioxide (NO2), and both incident and prevalent chronic bronchitis.
    Methods:
    We estimated annual average PM2.5, PM10, and NO2 concentrations using a national land-use regression model with spatial smoothing at home addresses of participants in a prospective nationwide U.S. cohort study of sisters of women with breast cancer. Incident chronic bronchitis and prevalent chronic bronchitis, cough and phlegm, were assessed by questionnaires.
    Results:
    Among 47,357 individuals with complete data, 1,383 had prevalent chronic bronchitis at baseline, and 647 incident cases occurred over 5.7-y average follow-up. No associations with incident chronic bronchitis were observed. Prevalent chronic bronchitis was associated with PM10 [adjusted odds ratio (aOR) per interquartile range (IQR) difference (5.8μg/m3)=1.07; 95% confidence interval (CI): 1.01, 1.13]. In never-smokers, PM2.5 was associated with prevalent chronic bronchitis (aOR=1.18 per IQR difference; 95% CI: 1.04, 1.34), and NO2 was associated with prevalent chronic bronchitis (aOR=1.10; 95%CI=1.01,1.20), cough (aOR=1.10; 95% CI: 1.05, 1.16), and phlegm (aOR=1.07; 95% CI: 1.01, 1.14); interaction p-values (nonsmokers vs. smokers) <0.05.
    Conclusions:
    PM10 exposure was related to chronic bronchitis prevalence. Among never-smokers, PM2.5 and NO2 exposure was associated with chronic bronchitis and component symptoms. Results may have policy ramifications for PM10 regulation by providing evidence for respiratory health effects related to long-term PM10 exposure. https://doi.org/10.1289/EHP2199
  • Received: 15 May 2017
    Revised: 4 December 2017
    Accepted: 5 December 2017
    Published: 6 February 2018

    Address correspondence to L.G. Hooper, University of Washington, Division of Pulmonary and Critical Care Medicine, Box 354695, 1959 NE Pacific St., Seattle, WA 98195-0001 USA. Telephone: (206) 543-3166. Email: lghooper@uw.edu

    Supplemental Material is available online (https://doi.org/10.1289/EHP2199).

    *These authors had equal contribution directing this work.

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (167 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (111 KB)

Introduction

Chronic bronchitis is a common clinical condition defined by chronic cough and sputum production for at least 3 mo in 2 or more consecutive years (American Thoracic Society 1995). Prevalence estimates in the general population of adults range from 3.5 to 27% (Kim et al. 2011; Martinez et al. 2014; Montes De Oca et al. 2012). This wide range may reflect, in part, variability in case definitions. Chronic bronchitis is a phenotype of chronic obstructive pulmonary disease (COPD) (Kim and Criner 2013). Among persons with COPD, chronic bronchitis portends increased frequency and severity of exacerbations (Burgel et al. 2009; Kim et al. 2011). Among persons without COPD, chronic bronchitis symptoms predict an increased risk of developing COPD, lower health-related quality-of-life scores, and increased risk for all-cause mortality (de Marco et al. 2007; Guerra et al. 2009; Lindberg et al. 2005; Martinez et al. 2014).

Smoking is the primary risk factor for chronic bronchitis, but exposure to ambient air pollution may also contribute (Kim and Criner 2013). The relationship between short-term air pollution exposure and acute respiratory symptoms and hospitalizations is well established (Peacock et al. 2011; Peel et al. 2005; Sunyer 2001), but limited data suggest a relationship between long-term ambient pollution exposure and COPD (Schikowski et al. 2014). There is a paucity of data on the possible relationship between classically defined chronic bronchitis and long-term exposure to the criteria pollutants PM2.5, PM10 (particulate matter <2.5μm and <10μm in diameter, respectively), and nitrogen dioxide (NO2). The sparse existing data provide inconsistent support for an association between PM10 and chronic cough and phlegm, and between NO2 and chronic cough (Bentayeb et al. 2010b; Cai et al. 2014; Schikowski et al. 2005; Zemp et al. 1999).

To address these relationships in a larger study, using specific outcome definitions and advanced exposure assessments, we investigated the association between residential exposure to PM2.5, PM10, and NO2 and both incident and prevalent chronic bronchitis in a prospective nationwide cohort of more than 50,000 U.S. women participating in the National Institute of Environmental Health Sciences (NIEHS) Sister Study. We estimated exposure at individuals’ residential addresses. Taking advantage of the comprehensive survey, we uniformly classified cases of chronic bronchitis using the classical clinical definition.

Methods

Study Population

The NIEHS Sister Study is a longitudinal cohort study of U.S. women with a sister diagnosed with breast cancer, but no personal breast cancer diagnosis at time of baseline interview (n=50,884). Women were enrolled between August 2003 and March 2009, and completed a baseline computer-assisted telephone survey. Follow-up telephone surveys were performed every 2 to 3 y. We analyzed data through the second follow-up survey (data release 4, data available through August 2014). Baseline and follow-up surveys queried participants on a wide range of health diagnoses and symptoms.

Of the 50,884 women participating in the NIEHS Sister Study, 1,234 (2.4%) were excluded for missing exposure data due to residential locations outside the modeling region or addresses that could not be geocoded (Figure 1). After excluding those missing baseline data on cough and phlegm, 47,357 individuals remained for analysis of prevalent outcomes. Of the 45,955 participants without chronic bronchitis symptoms at baseline, 6,111 (12.3%) were missing data on cough or phlegm for at least one of the two follow-up questionnaires, leaving 39,844 individuals for analysis of incident outcomes.

Flowchart showing NIEHS Sister Study cohort for women participants

Figure 1. Study population with excluded/missing participants.
*Total number of participants with nonmissing covariates for prevalence analyses is 44,158.
†Total number of participants with nonmissing covariates for incidence analysis is 38,006.

The Institutional Review Boards of the University of Washington and the NIEHS approved this study; all participants provided written informed consent.

Outcome Assessment

Chronic bronchitis was defined according to the classical symptom-based definition of chronic cough productive of phlegm for at least 3 mo out of a year for a minimum of 2 consecutive years (American Thoracic Society 1995). Participants were asked about the presence of cough and phlegm independently, and the duration of each symptom was specified using questions derived from the British Medical Research Council adult respiratory symptom standardized questionnaire. Women with cough and phlegm symptoms, both present for at least 3 mo per year out of the previous 2 y, were considered to have chronic bronchitis. Prevalent chronic bronchitis was determined by meeting symptom-based criteria at the baseline questionnaire. In a sensitivity analysis, we included history of physician diagnosis of chronic bronchitis in the case definition. Incident chronic bronchitis was defined by satisfying the case definition at either the second follow-up survey, or both the first and second follow-ups among participants who did not have chronic bronchitis at baseline. Participants whose symptoms did not persist from first through second follow-up were not considered cases.

Secondary outcomes were chronic cough (3 or more months of cough for at least 2 consecutive years, regardless of phlegm production), chronic phlegm (3 or more months of phlegm production for at least 2 consecutive years, regardless of cough), and chronic cough or phlegm. Both prevalent chronic cough and chronic phlegm were defined by being present at baseline.

Ambient Air Pollution Exposure Assessment

Air pollution exposure was estimated using annual average PM2.5, PM10, and NO2 levels at each participant’s current primary residence. Home addresses of participants were geocoded using ArcGIS (version 10; Esri). We estimated long-term exposure using year 2000 annual mean concentration levels for all pollutants. Measurements of PM2.5, PM10, and NO2 concentrations from monitors using federal reference methods were obtained from the U.S. Environmental Protection Agency (EPA) Air Quality System database. After excluding locations with only seasonal coverage or large amounts of missing data, the observations were aggregated into annual averages. The annual averages were used to fit a universal kriging regression model for predicting at points within the contiguous United States. The models for PM2.5 (Sampson et al. 2013) and NO2 (Young et al. 2016) have been previously described in detail, and the model for PM10 was fit in the same manner as the PM2.5 model. Partial least squares, a dimension reduction technique, was used to select linear combinations of land use, roadway proximity, and other geographic covariates. The NO2 prediction model additionally incorporated satellite data (Young et al. 2016). Spatial smoothing was included via an exponential covariance function. This model therefore incorporated land-use regression and spatial smoothing of values observed in the monitoring network. Model performance was evaluated using 10-fold cross-validation and for the year 2000. The cross-validated R2 was 0.85 for NO2, 0.53 for PM10, and 0.77 for PM2.5. Exposure modeling was limited to the continental United States; participants from Alaska, Hawaii, and Puerto Rico were excluded (n=1,234).

Statistical Analysis

To estimate the association between outcomes and pollutant exposures, we used multivariable logistic regression. Covariates were selected a priori based on plausible relationships and review of existing literature. Potential confounders, measured at baseline, were age (continuous), ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other), body mass index (continuous), education (high school or less, some college, associate or technical degree, bachelor’s degree, graduate degree), household income (continuous), occupational exposure to dust (ever/never) or vapors/fumes (ever/never), smoking status (never, former, current), tobacco pack-years (continuous), and years of secondhand smoke exposure since age 19 (continuous). After exclusion of individuals missing any of these covariates, 44,158 individuals were available for the analysis of prevalent outcomes. For incident outcomes, among those without chronic bronchitis at baseline, 38,006 individuals were available after excluding missing covariates.

A model adjusted for age alone was first performed followed by a fully adjusted model including all a priori identified covariates. PM2.5, PM10, and NO2 were modeled separately. Given potential bias by length of follow-up time for the incident chronic bronchitis outcome, adjustment was made for duration of follow-up time (from baseline to second follow-up survey) using restricted cubic splines with four knots (Dinse and Lagakos 1983). We performed analyses stratified by baseline smoking status (ever/never) and tested for interaction using product terms for smoking status and pollutant exposure. A two-pollutant model was performed by including remaining copollutants in the fully adjusted model. In all instances, a p-value of <0.05 was considered significant.

Sensitivity analyses were prespecified and performed on the following subgroups in independent analyses: a) prevalent chronic bronchitis that included either symptom-based criteria or report of physician diagnosis; b) excluding baseline asthmatics [defined by history of physician diagnosis, recent (within 12 mo) asthma medication use, and self-reported current asthma], given clinical overlap between asthma and chronic bronchitis; and c) participants who lived at least 10 y at their primary residence. In the asthma sensitivity analysis, asthmatics who reported either current smoking or history of 10 pack-years at baseline were not excluded given possibility of smoking-related symptoms leading to asthma misdiagnosis. Given concern that seasonal variation may affect results, an additional sensitivity analysis was performed by adjusting for season at the time of baseline and follow-up questionnaires.

Statistical analysis was performed using the statistical program Stata (version 11; StataCorp).

Results

Participants were, on average, 55.4 y old at baseline [standarddeviation(SD)=8.9], 84.8% were white, 52.6% had a bachelor’s degree or higher level of education, 56.4% had never smoked cigarettes, and only 8% were current smokers. The proportion of black and Hispanic participants increased across tertiles of PM2.5, PM10, and NO2 (Table 1). The distributions of occupational exposures, smoking history, and cumulative tobacco smoke exposure did not vary materially by ambient air pollution exposure levels (Table 1).

Table 1. Participant characteristics at baseline by exposure tertiles for particulate matter PM2.5, PM10, and NO2.
Characteristic PM2.5 (μg/m3) NO2 (ppb) PM10 (μg/m3)
2.1–10.9 10.9–13.9 13.9–25.3 2.16–9.54 9.54–14.1 14.1–39.0 5.68–19.8 19.8–23.4 23.4–56.4
n 14,720 14,719 14,719 14,720 14,719 14,719 14,720 14,719 14,719
Age (years) 55.9±8.9 55.4±9 54.9±8.9 55.7±8.8 55.3±8.9 55.3±9 55.7±8.7 55.3±8.9 55.3±9
BMI 27.5±6 27.9±6.3 28.3±6.5 27.8±6.1 27.8±6.3 27.9±6.5 27.5±6.1 27.9±6.3 28.2±6.5
Race/ethnicity (%)
 White (non-Hispanic) 90.7 85.9 77.7 89.2 85.4 79.7 90.1 85.5 78.6
 Black (non-Hispanic) 2.4 8.7 16.8 5.6 9.4 12.9 5.6 9.4 12.9
 Hispanic 4.1 2.9 3 2.3 2.9 4.8 1.8 2.7 5.6
 Other 2.8 2.5 2.5 2.9 2.3 2.6 2.5 2.3 3
Education (%)
 HS or less 14.7 15.1 14.8 17.6 14.8 12.3 15.1 15.1 14.5
 Some college 35.2 33.5 32.1 36.8 33.3 30.8 33.7 32.7 34.3
 Bachelor’s 27.4 27 26.9 25.7 27.4 28.2 26.6 27.3 27.4
 Graduate 22.8 24.4 26.2 20 24.6 28.8 24.6 25 23.8
Household income (USD) 41,218±25,849 43,570±26,914 43,743±28,293 39,995±24,817 43,064±26,339 45,472±29,533 42,473±25,999 43,104±26,892 42,953±28,242
Occupational Exposures (%)
 Vapors/fumes (ever) 24.9 24.3 24 25.7 24.1 23.5 24.4 24.4 24.4
 Dust (ever) 22.2 22.7 23.8 23.5 22.4 22.9 22.7 22.6 23.5
Smoking status (%)
 Never 56.2 55.7 57.2 57.2 57.3 54.6 54.4 57.4 57.3
 Former 36.6 36.4 33.9 34.8 35 37.1 37.6 34.8 34.5
 Current 7.2 8 8.9 8.1 7.7 8.3 7.9 7.8 8.3
Pack-years among ever-smokers 14.5±15.3 14.7±15.2 14.8±15.5 15.1±15.6 14.3±15.1 14.5±15.2 14.8±15.5 14.5±15.2 14.6±15.2
Packs per day among current smokers 0.7±0.4 0.7±0.4 0.7±0.5 0.7±0.5 0.7±0.4 0.6±0.4 0.7±0.5 0.7±0.4 0.7±0.4
Adult secondhand smoke (years) 10.9±12.9 11.3±13.3 11.4±13.2 11.5±13.2 11.1±13.1 11.1±13.1 11.3±13.1 11.1±13.1 11.3±13.3
Number of years at primary residence 12.8±10.9 13.5±11.3 13.6±11.5 12.7±11 12.7±10.7 14.6±11.8 13.4±11 13.2±11.2 13.3±11.5
Lived at primary residence
10 y (%)
50.7 53.5 53.9 50.1 50.8 57.1 53.6 52.2 52.4
Asthma at baseline (%) 5.7 5.6 5.9 5 6 6.3 5.4 5.9 6
Physician diagnosis of COPD (%) 1.5 1.5 1.5 1.6 1.6 1.4 1.5 1.5 1.5
Physician diagnosis of chronic bronchitis (%) 7.4 8 8.4 8 8.1 7.8 7.5 8 8.4

Note: BMI, body mass index; COPD, chronic obstructive pulmonary disease; HS, high school; PM, particulate matter.

The mean follow-up time was 5.7 y from enrollment to the second follow-up survey. During the follow-up period, there were 638 incident cases of chronic bronchitis, giving an estimated incidence rate of 2.8 cases per 1,000 person-years. At baseline, 1,351 (3.1%) women met symptom-based criteria for chronic bronchitis, whereas 4,698 (10.6%) participants reported ever having had a physician diagnosis of chronic bronchitis. Prevalent chronic cough was reported by 3,749 (8.5%) and chronic phlegm by 2,776 (6.3%) participants at baseline.

The median estimated exposure concentrations were 12.4μg/m3 [interquartilerange(IQR)=4.4μg/m3] for PM2.5, 2.16μg/m3 (5.8μg/m3) for PM10, and 11.7ppb (7.3ppb) for NO2. The results of the age-adjusted and fully adjusted regression analyses are presented in Table 2. No statistically significant associations were found between incident chronic bronchitis and any of the air pollution exposures. Limiting the incidence analysis to long-term residents (>10 y) did not appreciably alter the effect estimates (Table S1).

Table 2. Odds ratios per interquartile range (IQR) increase in particulate matter PM2.5 (4.4μg/m3), NO2 (7.3ppb), and PM10 (5.8μg/m3).
Exposure and outcome Cases Age adjusted Fully adjusted
OR (95% CI) p-Value OR (95% CI) p-Value
PM2.5
 Incident chronic bronchitis 638 0.94 (0.84, 1.05) 0.256 0.94 (0.83, 1.06) 0.289
 Prevalent (at baseline)
  Chronic bronchitis 1,351 1.04 (0.97, 1.13) 0.276 1.04 (0.96, 1.13) 0.318
  Chronic cough 3,749 1.03 (0.98, 1.08) 0.213 1.04 (0.99, 1.10) 0.103
  Chronic phlegm 2,776 1.07 (1.02, 1.13) 0.010 1.04 (0.98, 1.10) 0.213
  Chronic cough or phlegm 5,271 1.05 (1.01, 1.10) 0.015 1.04 (1.00, 1.09) 0.067
NO2
 Incident chronic bronchitis 638 0.95 (0.87, 1.03) 0.198 1.00 (0.92, 1.09) 0.974
 Prevalent (at baseline)
  Chronic bronchitis 1,351 1.00 (0.95, 1.06) 0.923 1.05 (0.99, 1.11) 0.136
  Chronic cough 3,749 1.02 (0.99, 1.06) 0.215 1.06 (1.02, 1.10) 0.002
  Chronic phlegm 2,776 1.01 (0.97, 1.05) 0.730 1.02 (0.98, 1.07) 0.266
  Chronic cough or phlegm 5,271 1.02 (0.99, 1.05) 0.199 1.04 (1.01, 1.08) 0.008
PM10
 Incident chronic bronchitis 638 0.92 (0.85, 1.01) 0.066 0.98 (0.90, 1.08) 0.745
 Prevalent (at baseline)
  Chronic bronchitis 1,351 1.06 (1.01, 1.12) 0.027 1.07 (1.01, 1.13) 0.019
  Chronic cough 3,749 1.04 (1.00, 1.07) 0.045 1.04 (1.00, 1.08) 0.030
  Chronic phlegm 2,776 1.07 (1.03, 1.12) <0.001 1.07 (1.02, 1.11) 0.002
  Chronic cough or phlegm 5,271 1.05 (1.01, 1.08) 0.001 1.05 (1.02, 1.08) 0.003

Note: Each outcome was compared to all participants without that outcome. The total number of participants with nonmissing data on all covariates was 38,006 for the analysis of incident outcomes, 44,158 for prevalent outcomes. Fully adjusted model includes age, race/ethnicity, body mass index, education, household income, occupational exposure to vapors/fumes or dust (ever), smoking status, total pack-years, and environmental tobacco smoke exposure. Primary outcome (incident chronic bronchitis) analysis additionally adjusted for length of follow-up time. CI, confidence interval; IQR, interquartile range; OR, odds ratio; PM, particulate matter.

For prevalent chronic bronchitis, a statistically significant positive association was seen with PM10 [odds ratio (OR) per IQR increase in PM10=1.07; 95% confidence interval (CI): 1.01, 1.13] (Table 2). Similar magnitudes of association with prevalent chronic bronchitis were seen for NO2 (OR=1.05; 95% CI: 0.99, 1.11) and PM2.5 (OR=1.04; 95% CI: 0.96, 1.13), but were not statistically significant. PM10 was also statistically significantly associated with chronic cough (OR=1.04; 95% CI: 1.00, 1.08), chronic phlegm (OR=1.07; 95% CI: 1.02, 1.11), and chronic cough or phlegm (OR=1.05; 95% CI: 1.02, 1.08); coadjustment for PM2.5 did not alter these effect estimates (Table S2). Adjustment of the PM10 model for NO2 resulted in a general attenuation of associations between prevalent symptoms and PM10. This attenuation is likely due in part to the strong correlation between NO2 and PM10 (Pearson’s r: 0.59).

NO2 showed a significant positive association with chronic cough (OR = 1.06; 95% CI: 1.02, 1.10) and chronic cough or phlegm (OR=1.04; 95% CI: 1.01, 1.08). In the NO2 model, ORs were robust to coadjustment for PM2.5 (Table S2). Coadjustment for PM10 in the NO2 model resulted in a loss of precision for the association between NO2 and chronic cough or phlegm, and an overall decrease in size of effect estimates across all prevalent outcomes. The significant association with chronic cough was preserved (OR=1.05; 95% CI: 1.01, 1.10). ORs for all pollutants and outcomes were generally very similar between age-adjusted and fully adjusted models.

For prevalent chronic bronchitis, sensitivity analyses incorporating additional case requirements into the classical symptom-based definition showed comparable effect estimates in association with PM10 (Table 3). For example, PM10 was significantly associated with prevalent chronic bronchitis defined either by symptoms or including participant-reported physician diagnosis (OR per IQR increase=1.06; 95% CI: 1.02, 1.09). In an analysis of prevalent chronic bronchitis excluding baseline asthmatics, the effect estimate was similar but less precise, commensurate with the smaller sample size (OR for IQR increase in PM10=1.06; 95% CI: 0.99, 1.13). Similarly, exclusion of the 47% of participants who lived at their residence less than 10 y largely preserved the estimated association, but with loss of precision reflecting the smaller sample size (OR per IQR increase in PM10=1.07; 95% CI: 0.99, 1.16). Comparable sensitivity analyses involving PM2.5 or NO2 and prevalent chronic bronchitis yielded ORs that were similar in magnitude and direction to the primary models (Table 3). Effect estimates for all three pollutants were essentially unchanged by seasonal adjustment (Table S3).

Table 3. Sensitivity analyses evaluating case definitions with additional inclusion or exclusion criteria for association of prevalent (baseline) chronic bronchitis with ambient air pollutants: odds ratios per interquartile range (IQR) increase in particulate matter PM2.5 (4.4μg/m3), NO2 (7.3ppb), and PM10 (5.8μg/m3).
Exposure and case definitions (primary and sensitivity analyses) na Cases Adjusted OR (95% CI) p-Value
PM2.5
 Prevalent chronic bronchitis (primary case definition) 44,158 1,351 1.04 (0.96, 1.13) 0.318
 Including physician diagnosis 44,099 4,698 1.04 (0.99, 1.09) 0.104
 Excluding asthma at baseline 41,488 1,104 0.99 (0.90, 1.08) 0.798
 Excluding those living at residence
<10 y
23,273 720 0.97 (0.87, 1.09) 0.643
NO2
 Prevalent chronic bronchitis (primary case definition) 44,158 1,351 1.05 (0.99, 1.11) 0.136
 Including physician diagnosis 44,099 4,698 1.02 (0.99, 1.06) 0.191
 Excluding asthma at baseline 41,488 1,104 1.02 (0.96, 1.09) 0.492
 Excluding those living at residence
<10 y
23,273 720 1.03 (0.95, 1.11) 0.444
PM10
 Prevalent chronic bronchitis (primary case definition) 44,158 1,351 1.07 (1.01, 1.13) 0.019
 Including physician diagnosis 44,099 4,698 1.06 (1.02, 1.09) 0.001
 Excluding asthma at baseline 41,488 1,104 1.06 (0.99, 1.13) 0.077
 Excluding those living at residence
<10 y
23,273 720 1.07 (0.99, 1.16) 0.093

Note: For each case definition, the comparison group was all individuals without that outcome. Each analysis was performed independently for each case definition. Adjusted for age, race/ethnicity, body mass index, education, household income, occupational exposure to vapors/fumes or dust (ever), smoking status, total pack-years, and secondhand smoke exposure. CI, confidence interval; OR, odds ratio; PM, particulate matter.

aTotal number of individuals with nonmissing data on all covariates for analysis.

In smoking-stratified analyses, we found evidence for stronger associations between all three air pollutants and prevalent outcomes in never-smokers (Table 4). PM2.5 was strongly associated with prevalent chronic bronchitis among never-smokers (OR per IQR difference=1.18; 95% CI: 1.04, 1.34), and the difference by smoking status was statistically significant (pinteraction=0.013). Similarly with NO2, in never-smokers, significant associations were seen for all four prevalent outcomes: chronic bronchitis (OR=1.10; 95% CI: 1.01, 1.20), chronic cough (OR=1.10; 95% CI: 1.05, 1.16), chronic phlegm (OR=1.07; 95% CI: 1.01, 1.14), and chronic cough or phlegm (OR=1.09; 95% CI: 1.04, 1.13), and the differences by smoking status were statistically significant for both cough (pinteraction=0.020), phlegm (pinteraction=0.017), and cough or phlegm (pinteraction=0.004). Corresponding ORs for PM2.5 and NO2 were close to null for ever-smokers. For PM10, results did not differ significantly by smoking status, although the same pattern of stronger associations in never-smokers was seen (Table 4).

Table 4. Chronic bronchitis in relation to air pollutants [particulate matter PM2.5, NO2, and PM10] by smoking status (never/ever): odds ratios per interquartile range (IQR) increase in PM2.5 (4.4μg/m3), NO2 (7.3ppb), and PM10 (5.8μg/m3).
Exposure and outcome Never-smoker Ever-smoker
Cases OR (95% CI) p-Value Cases OR (95% CI) p-Value pInteraction
PM2.5
 Incident chronic bronchitis 271 0.92 (0.77, 1.10) 0.382 367 0.95 (0.81, 1.11) 0.516 0.815
 Prevalent (at baseline)
  Chronic bronchitis 580 1.18 (1.04, 1.34) 0.011 771 0.96 (0.86, 1.07) 0.427 0.013
  Chronic cough 1,802 1.07 (1.00, 1.15) 0.053 1,947 1.02 (0.95, 1.10) 0.545 0.345
  Chronic phlegm 1,362 1.08 (1.00, 1.17) 0.059 1,414 1.00 (0.92, 1.09) 0.956 0.189
  Chronic cough or phlegm 2,632 1.06 (0.99, 1.12) 0.077 2,639 1.03 (0.97, 1.10) 0.292 0.638
NO2
 Incident chronic bronchitis 271 1.03 (0.91, 1.17) 0.609 367 0.97 (0.86, 1.10) 0.660 0.498
 Prevalent (at baseline)
  Chronic bronchitis 580 1.10 (1.01, 1.20) 0.029 771 1.00 (0.92, 1.08) 0.955 0.097
  Chronic cough 1,802 1.10 (1.05, 1.16) <0.001 1,947 1.01 (0.96, 1.06) 0.642 0.020
  Chronic phlegm 1,362 1.07 (1.01, 1.14) 0.014 1,414 0.97 (0.92, 1.03) 0.359 0.017
  Chronic cough or phlegm 2,632 1.09 (1.04, 1.13) <0.001 2,639 0.99 (0.95, 1.04) 0.806 0.004
PM10
 Incident chronic bronchitis 271 1.04 (0.91, 1.18) 0.587 367 0.95 (0.83, 1.08) 0.458 0.359
 Prevalent (at baseline)
  Chronic bronchitis 580 1.09 (1.00, 1.20) 0.055 771 1.06 (0.98, 1.14) 0.131 0.597
  Chronic cough 1,802 1.07 (1.01, 1.12) 0.015 1,947 1.02 (0.97, 1.08) 0.391 0.260
  Chronic phlegm 1,362 1.10 (1.04, 1.17) 0.002 1,414 1.04 (0.98, 1.10) 0.179 0.189
  Chronic cough or phlegm 2,632 1.08 (1.03, 1.13) <0.001 2,639 1.02 (0.98, 1.07) 0.352 0.081

Note: For each case definition, the comparison group was all individuals without that outcome. The total number of participants with nonmissing data on all covariates was 38,006 (21,527 never-smokers and 16,479 ever-smokers) for the analysis of incident outcomes and 44,158 (24,894 never-smokers and 19,264 ever-smokers) for prevalent outcomes. Adjusted for age, race/ethnicity, body mass index, education, household income, occupational exposure to vapors/fumes or dust (ever), total pack-years, and environmental tobacco smoke exposure. Incident analysis additionally adjusted for length of follow-up time. CI, confidence interval; IQR, interquartile range; OR, odds ratio; PM, particulate matter.

Discussion

To our knowledge, this is the largest study to investigate the association between classically defined chronic bronchitis and long-term ambient air pollution exposure using a validated national exposure model. We did not find an association between incident chronic bronchitis and any of the three air pollution measures. However, exposure to higher concentrations of PM10 was significantly associated with all prevalent outcomes: chronic bronchitis, chronic cough, chronic phlegm, and chronic cough or phlegm. These findings were robust to coadjustment for PM2.5 in a two-pollutant model (Table S2). We also found NO2 exposure was significantly associated with chronic cough and chronic cough or phlegm. To the best of our knowledge, no other study has shown an association between PM10 and classically defined chronic bronchitis. These findings provide evidence that long-term ambient air pollution exposure, particularly PM10, is a risk factor for chronic bronchitis and the chronic respiratory symptoms of cough and phlegm that define it.

Incident chronic bronchitis should be superior to prevalent chronic bronchitis for making causal inference regarding observed associations with air pollution. However, the relatively short follow-up duration (mean: 5.7 y) limited our power to detect an association between ambient air pollutants and incident chronic bronchitis. With the much larger number of cases of prevalent conditions, we had substantially higher power than for the incident analyses. One smaller study of nonsmoking Seventh Day Adventists in California has shown an association between incident chronic bronchitis and long-term exposure to PM2.5; however, levels were in excess of 20μg/m3, a concentration almost double that observed in our study (Abbey et al. 1995).

Comparison to previous studies is limited due to substantial variability in defining chronic bronchitis and exposure estimation methods. The observed incidence rate of 2.5 cases per 1,000 person-years and prevalence of 2.9% are at the low end of the range reported in the literature (Cai et al. 2014; Cerveri et al. 2001; Huchon et al. 2002; Kim et al. 2011; Sobradillo et al. 1999). However, our study population was more than half nonsmoking women, and our estimates are in agreement with study populations with similar demographics (Montes De Oca et al. 2012; Sunyer et al. 2006). National prevalence and incidence figures for chronic bronchitis are lacking because they rely on physician diagnosis rather than the classical symptom-based diagnostic criteria (American Lung Association 2013). Including participant-reported physician diagnosis greatly increases the prevalence of chronic bronchitis in this study and likely elsewhere (Schikowski et al. 2005).

Our study provides evidence that PM10 exposure is a risk factor for chronic bronchitis, while the existing literature suggests associations between PM10 and various respiratory symptoms. A large cross-sectional study in Switzerland found an association between increased prevalence of chronic cough and phlegm with PM10 exposure among never-smokers (Zemp et al. 1999). The European Study of Cohorts for Air Pollution Effects (ESCAPE) meta-analysis of five European cohorts similarly showed an association between PM10 and prevalent chronic phlegm, but not chronic bronchitis, in never-smokers (Cai et al. 2014). A French study of elderly adults demonstrated increased prevalence of chronic cough associated with PM10 exposure (Bentayeb et al. 2010a). Furthermore, in the Swiss Cohort Study on Air Pollution and Lung and Heart Diseases in Adults (SAPALDIA) cohort, decline in PM10 over time was associated with a reduction in chronic cough and phlegm (Schindler et al. 2009). Our study suggests long-term PM10 exposure is associated with prevalent chronic bronchitis, the distinct clinical entity, as well as the associated symptoms that define it.

In contrast to PM2.5, which deposits within the distal alveoli, the preferential deposition of coarse particles within the conducting airways of the tracheobronchial tree provides biologic plausibility for the association between PM10 and chronic bronchitis (Carvalho et al. 2011). Chronic airway epithelial inflammation and mucus metaplasia are the pathologic bases of chronic bronchitis (Kim and Criner 2013). Chronic bronchitis is associated with narrowing and mucus plugging of the nonalveolated conducting airways (Matsuba and Thurlbeck 1973). PM has been frequently implicated in triggering pro-inflammatory cascades within airway epithelial cells (Øvrevik et al. 2015). Certain PM10 components, including transition metals and endotoxins, have been shown to drive airway inflammation and mucus hypersecretion via upregulation of transcription factors and generation of reactive oxygen species and oxidative stress (Longphre et al. 2000; Øvrevik et al. 2015).

For NO2, we saw no associations with prevalent chronic bronchitis or chronic phlegm in the main analyses; however, there were positive associations with both chronic cough and chronic cough or phlegm. These associations were robust to coadjustment for PM2.5. Adjustment for PM10, in an attempt to isolate the effect of NO2, was limited by the strong correlation between these copollutants. Previous studies have shown inconsistent associations between respiratory symptoms and NO2. The ESCAPE meta-analysis showed no significant association between NO2 and chronic bronchitis, cough, or phlegm (Cai et al. 2014). The German Study on the Influence of Air Pollution on Lung, Inflammation and Aging (SALIA) cohort analysis showed association with cough, but not chronic bronchitis, while the Swiss SAPALDIA study showed no association overall (Schikowski et al. 2005; Zemp et al. 1999). In both the SALIA and SAPALDIA studies, conducted in the 1990s, the annual mean NO2 levels were more than double those in our study.

Among never-smokers, associations between prevalent symptoms and exposures were stronger for all pollutants. If we limit our interpretation of the stratified analyses to those with statistically significant interactions, we find stronger evidence for associations between both NO2 and PM2.5 and our prevalent outcomes among never-smokers. The finding of stronger associations with these two pollutants and outcomes in never-smokers is consistent with some previous literature findings. In the Swiss SAPALDIA study, NO2 was related to chronic bronchitis only among nonsmokers. For PM2.5, the ESCAPE meta-analysis found a positive association between chronic cough and PM2.5 only among never-smokers (Cai et al. 2014). NO2 was associated with chronic cough in a study of about 4,700 women in Germany, 74% of whom were never-smokers (Schikowski et al. 2005).

The reason for stronger associations with NO2 in never-smokers in our study and others is unclear. Perhaps the airways of smokers are more tolerant to the irritant effects of ambient NO2 than nonsmokers who, as a result, are dissuaded from smoking because of greater sensitivity to these effects. Alternatively, the effects of compounds in cigarette smoke might swamp the effects of ambient NO2 exposure. Stronger associations with PM2.5 in never-smokers could reflect, in part, the high dose of fine particles inhaled by smokers. The biologic effects of cigarette smoke may overwhelm the effects from long-term, low-level ambient air pollution and thus mask any association. For PM10, we saw clear associations with chronic bronchitis in all subjects, possibly related to PM10 favoring deposition in the conducting airways that are responsible for producing bronchitic symptoms. This association is apparent in the aggregate sample. In contrast, the associations between chronic bronchitis and PM2.5 and NO2 were only apparent in never-smokers. Theses pollutants have distribution patterns that tend to bypass the conducting airways for deposition and adsorption in the distal alveoli.

Outcome misclassification was reduced by using the symptom-defined definition of chronic bronchitis. However, the symptom-based questionnaire still has limitations; due to recall bias, it likely results in inclusion of cases with recent, but not necessarily chronic, symptoms. Overlap with asthma remains possible given the clinical similarities of these conditions. Sensitivity analyses excluding subjects with a physician diagnosis of asthma and active asthma symptoms at baseline attenuated the association with all pollutants. However, the exclusion of these participants with overlapping chronic bronchitis symptoms may have eliminated true cases of chronic bronchitis and thus reduced power.

The sensitivity analysis including individuals reporting doctor diagnosis of chronic bronchitis showed preserved associations with increased precision. Self-report of doctor diagnosis is likely to include more individuals with symptom duration shorter than the 2-y minimum required for the chronic bronchitis definition (i.e., may include participants who have received a diagnosis of acute bronchitis in the past). Given that chronic bronchitis is defined by duration of symptoms, directly asking subjects questions on cough and phlegm is preferable to asking about physician diagnosis, which requires accurate reporting by both parties. However, the association between PM10 and prevalent chronic bronchitis remains robust, even with the less strict definition, suggesting that presence of symptoms is driving the relationship, rather than the specific duration of symptoms.

The objection has been raised that chronic bronchitis prevalence as reported on questionnaires may reflect recent symptoms and that prevalence and/or severity might then vary by season. Therefore, we undertook a sensitivity analysis adjusting for season of both baseline and follow-up questionnaire (Table S3). No change was observed in the effect estimates for outcomes associated with PM2.5 or NO2. The associations between prevalent chronic bronchitis and chronic cough and PM10 were no longer statistically significant after adjusting for season, but the effect estimates remained largely unchanged and in the anticipated direction. Season of questionnaire administration does not seem to contribute significant bias in reporting of chronic bronchitis.

Our air pollution exposure estimates are based on a validated national model using land-use regression and spatial smoothing to capture within- and between-region air pollution variability and minimize exposure misclassification. This model is a considerable improvement over road proximity, regional fixed-monitor averages, and simple land-use regression models employed in prior chronic bronchitis investigations (Keller et al. 2014; Young et al. 2016). In addition, seasonal bias in the exposure should be mitigated by using annual averages and a chronic outcome whose case definition dictates that symptoms must span a minimum of 2 consecutive years. Air pollution estimates used year 2000 annual averages, predating baseline enrollment for all participants. While concentrations of criteria pollutants are declining nationally, spatial differences of annual average pollution concentrations account for the majority of variability in PM2.5 measurement and were relatively stable across the study period (Kim et al. 2017). However, it is acknowledged that variability in the decline in pollution levels may contribute to exposure misclassification, and the resulting biases are difficult to predict. It is plausible that our observed lack of association for incident outcomes and positive effects for prevalent symptoms could be related to variable change in pollution levels; e.g., if the most polluted regions experienced more dramatic declines in levels than the cleaner areas.

Additional limitations exist. This study is limited to women, and the findings may not be broadly applicable to men. Outdoor ambient pollutant concentrations may not reflect the indoor exposures. Exposure measurement error owing to our use of residential addresses to characterize exposure owing both to subjects’ residential mobility, time spent away from home or indoors, and spatiotemporal trends is a limitation both of this study and epidemiological studies of air pollution health effects in general. Exposure measurement of this nature error is generally expected to bias associations toward the null rather than producing false positive associations.

In 2006, the EPA revoked the National Ambient Air Quality Standard for annual PM10 due to insufficient data on health risks associated with long-term exposure to PM10 as opposed to the finer PM2.5 fraction (U.S. EPA 2006). The preceding long-term PM10 exposure standard was an annual average of 50μg/m3, roughly double the mean concentration experienced by participants in this study. This study provides evidence that chronic respiratory health effects occur with long-term exposure to PM10 at levels below the previous national standards. These results add to a limited body of evidence relating morbidity to long-term PM10 exposure and consequently may have policy implications both nationally and globally.

Acknowledgments

The authors thank A. Gassett and C. Sack for data management and support with this analysis, as well as the staff at Social and Scientific Systems, Inc. and Westat, Inc. for overall study support and management of Sister Study data acquisition. The authors also thank the University of Washington Center for Clean Air Research (UW CCAR) and the Multi-Ethnic Study of Atherosclerosis and Air Pollution (MESA Air) for providing the infrastructure and development of the exposure models. MESA Air is conducted and supported by the U.S. EPA through grant RD831697 to the UW. This publication has not been formally reviewed by the U.S. EPA. The views expressed in this document are solely those of the authors and do not necessarily reflect those of the agency. U.S. EPA does not endorse any products or commercial services mentioned in this publication. The authors also thank the thousands of women who participate in the Sister Study.

This work was supported by the University of Washington Pulmonary and Critical Care Medicine Training Grant (T32 HL007287), the Intramural Research Program of the National Institutes of Health (NIH), and the National Institute of Environmental Health Sciences (NIEHS) (Z01 ES044005 and ES043012). Exposure modeling was supported by the U.S. EPA (RD831697 and R834796), and Biostatistics, Epidemiologic, and Bioinformatic Training in Environmental Health Training Grant was from NIEHS (T32ES015459).

References

Abbey DE, Ostro BE, Petersen F, Burchette RJ. 1995. Chronic respiratory symptoms associated with estimated long-term ambient concentrations of fine particulates less than 2.5 microns in aerodynamic diameter (PM2.5) and other air pollutants. J Expo Anal Environ Epidemiol 5(2):137–159, PMID: 7492903.

American Lung Association. 2013. Trends in COPD (Chronic Bronchitis and Emphysema): Morbidity and Mortality. http://www.lung.org/assets/documents/research/copd-trend-report.pdf [accessed 9 May 2017].

American Thoracic Society. 1995. Standards for the diagnosis and care of patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 152(5 Part 2):S77–121, PMID: 7582322.

Bentayeb M, Helmer C, Raherison C, Dartigues JF, Tessier JF, Annesi-Maesano I. 2010. Bronchitis-like symptoms and proximity air pollution in French elderly. Respir Med 104(6):880–888, PMID: 20129767, 10.1016/j.rmed.2010.01.004.

Burgel PR, Nesme-Meyer P, Chanez P, Caillaud D, Carré P, Perez T, et al. 2009. Cough and sputum production are associated with frequent exacerbations and hospitalizations in COPD subjects. Chest 135(4):975–982, PMID: 19017866, 10.1378/chest.08-2062.

Cai Y, Schikowski T, Adam M, Buschka A, Carsin AE, Jacquemin B, et al. 2014. Cross-sectional associations between air pollution and chronic bronchitis: an ESCAPE meta-analysis across five cohorts. Thorax 69(11):1005–1014, PMID: 25112730, 10.1136/thoraxjnl-2013-204352.

Carvalho TC, Peters JI, Williams RO III. 2011. Influence of particle size on regional lung deposition–what evidence is there? Int J Pharm 406(1–2):1–10, PMID: 21232585, 10.1016/j.ijpharm.2010.12.040.

Cerveri I, Accordini S, Verlato G, Corsico A, Zoia MC, Casali L, et al. 2001. Variations in the prevalence across countries of chronic bronchitis and smoking habits in young adults. Eur Respir J 18(1):85–92, PMID: 11510810.

de Marco R, Accordini S, Cerveri I, Corsico A, Antó JM, Künzli N, et al. 2007. Incidence of chronic obstructive pulmonary disease in a cohort of young adults according to the presence of chronic cough and phlegm. Am J Respir Crit Care Med 175(1):32–39, PMID: 17008642, 10.1164/rccm.200603-381OC.

Dinse GE, Lagakos SW. 1983. Regression analysis of tumour prevalence data. J R Stat Soc Ser C Appl Stat 32(3):236, 10.2307/2347946.

Guerra S, Sherrill DL, Venker C, Ceccato CM, Halonen M, Martinez FD. 2009. Chronic bronchitis before age 50 years predicts incident airflow limitation and mortality risk. Thorax 64(10):894–900, PMID: 19581277, 10.1136/thx.2008.110619.

Huchon GJ, Vergnenègre A, Neukirch F, Brami G, Roche N, Preux PM. 2002. Chronic bronchitis among French adults: high prevalence and underdiagnosis. Eur Respir J 20(4):806–812, PMID: 12412668, 10.1183/09031936.02.00042002.

Keller JP, Olives C, Kim SY, Sheppard L, Sampson PD, Szpiro AA, et al. 2014. A unified spatiotemporal modeling approach for predicting concentrations of multiple air pollutants in the Multi-Ethnic Study of Atherosclerosis and Air Pollution. Environ. Health Perspect 123(4):301–309, PMID: 25398188, 10.1289/ehp.1408145.

Kim SY, Olives C, Sheppard L, Sampson PD, Larson TV, Keller JP, et al. 2017. Historical prediction modeling approach for estimating long-term concentrations of PM2.5 in cohort studies before the 1999 implementation of widespread monitoring. Environ Health Perspect 125(1):38–46, PMID: 27340825, 10.1289/EHP131.

Kim V, Criner GJ. 2013. Chronic bronchitis and chronic obstructive pulmonary disease. Am J Respir Crit Care Med 187(3):228–237, PMID: 23204254, 10.1164/rccm.201210-1843CI.

Kim V, Han MK, Vance GB, Make BJ, Newell JD, Hokanson JE, et al. 2011. The chronic bronchitic phenotype of COPD: an analysis of the COPDGene Study. Chest 140(3):626–633, PMID: 21474571, 10.1378/chest.10-2948.

Lindberg A, Jonsson AC, Rönmark E, Lundgren R, Larsson LG, Lundbäck B. 2005. Ten-year cumulative incidence of COPD and risk factors for incident disease in a symptomatic cohort. Chest 127(5):1544–1552, PMID: 15888826, 10.1378/chest.127.5.1544.

Longphre M, Li D, Li J, Matovinovic E, Gallup M, Samet JM, et al. 2000. Lung mucin production is stimulated by the air pollutant residual oil fly ash. Toxicol Appl Pharmacol 162(2):86–92, PMID: 10637131, 10.1006/taap.1999.8838.

Martinez CH, Kim V, Chen Y, Kazerooni EA, Murray S, Criner GJ, et al. 2014. The clinical impact of non-obstructive chronic bronchitis in current and former smokers. Respir Med 108(3):491–499, PMID: 24280543, 10.1016/j.rmed.2013.11.003.

Matsuba K, Thurlbeck WM. 1973. Disease of the small airways in chronic bronchitis. Am Rev Respir Dis 107(4):552–558, PMID: 4697664, 10.1164/arrd.1973.107.4.552.

Montes de Oca M, Halbert RJ, Lopez MV, Perez-Padilla R, Tálamo C, Moreno D, et al. 2012. The chronic bronchitis phenotype in subjects with and without COPD: the PLATINO study. Eur Respir J 40(1):28–36, PMID: 22282547, 10.1183/09031936.00141611.

Øvrevik J, Refsnes M, Låg M, Holme J, Schwarze PE. 2015. Activation of proinflammatory responses in cells of the airway mucosa by particulate matter: oxidant- and non-oxidant-mediated triggering mechanisms. Biomolecules 5(3):1399–1440, PMID: 26147224, 10.3390/biom5031399.

Peacock JL, Anderson HR, Bremner SA, Marston L, Seemungal TA, Strachan DP, et al. 2011. Outdoor air pollution and respiratory health in patients with COPD. Thorax 66(7):591–596, PMID: 21459856, 10.1136/thx.2010.155358.

Peel JL, Tolbert PE, Klein M, Metzger KB, Flanders WD, Todd K, et al. 2005. Ambient air pollution and respiratory emergency department visits. Epidemiology 16(2):164–174, PMID: 15703530, 10.1097/01.ede.0000152905.42113.db.

Sampson PD, Richards M, Szpiro AA, Bergen S, Sheppard L, Larson TV, et al. 2013. A regionalized national universal kriging model using Partial Least Squares regression for estimating annual PM2.5 concentrations in epidemiology. Atmos Environ (1994) 75:383–392, PMID: 24015108, 10.1016/j.atmosenv.2013.04.015.

Schikowski T, Mills IC, Anderson HR, Cohen A, Hansell A, Kauffmann F, et al. 2014. Ambient air pollution: a cause of COPD? Eur Respir J 43(1):250–263, PMID: 23471349, 10.1183/09031936.00100112.

Schikowski T, Sugiri D, Ranft U, Gehring U, Heinrich J, Wichmann HE, et al. 2005. Long-term air pollution exposure and living close to busy roads are associated with COPD in women. Respir Res 6:152, PMID: 16372913, 10.1186/1465-9921-6-152.

Schindler C, Keidel D, Gerbase MW, Zemp E, Bettschart R, Brändli O, et al. 2009. Improvements in PM10 exposure and reduced rates of respiratory symptoms in a cohort of Swiss adults (SAPALDIA). Am J Respir Crit Care Med 179(7):579–587, PMID: 19151198, 10.1164/rccm.200803-388OC.

Sobradillo V, Miravitlles M, Jiménez CA, Gabriel R, Viejo JL, Masa JF, et al. 1999. [Epidemiological study of chronic obstructive pulmonary disease in Spain (IBERPOC): prevalence of chronic respiratory symptoms and airflow limitation]. Arch. Bronconeumol 35(4):159–166, PMID: 10330536.

Sunyer J. 2001. Urban air pollution and chronic obstructive pulmonary disease: a review. Eur Respir J 17(5):1024–1033, PMID: 11488305, 10.1183/09031936.01.17510240.

Sunyer J, Jarvis D, Gotschi T, Garcia-Esteban R, Jacquemin B, Aguilera I, et al. 2006. Chronic bronchitis and urban air pollution in an international study. Occup Environ Med 63(12):836–843, PMID: 16847030, 10.1136/oem.2006.027995.

U.S. EPA (Environmental Protection Agency). 2006. National Ambient Air Quality Standards for Particulate Matter. Final Rule. 40 CFR 50. Docket No. EPA-HQ-OAR-2001-0017, FRL-8225-3. Fed Reg 71(200). https://www.gpo.gov/fdsys/pkg/FR-2006-10-17/pdf/06-8477.pdf [accessed 29 December 2017].

Young MT, Bechle MJ, Sampson PD, Szpiro AA, Marshall JD, Sheppard L, et al. 2016. Satellite-based NO2 and model validation in a national prediction model based on universal kriging and land-use regression. Environ Sci Technol 50(7):3686–3694, PMID: 26927327, 10.1021/acs.est.5b05099.

Zemp E, Elsasser S, Schindler C, Künzli N, Perruchoud AP, Domenighetti G, et al. 1999. Long-term ambient air pollution and respiratory symptoms in adults (SAPALDIA study). The SAPALDIA Team. Am J Respir Crit Care Med 159(4 Pt 1):1257–1266, PMID: 10194174, 10.1164/ajrccm.159.4.9807052.

Long-term Air Pollution Exposure, Genome-wide DNA Methylation and Lung Function in the LifeLines Cohort Study

Author Affiliations open

1Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands

2Groningen Research Institute for Asthma and COPD (GRIAC), University of Groningen, University Medical Center Groningen, Groningen, Netherlands

3Department of Genetics, University of Groningen, University Medical Center Groningen, Groningen, Netherlands

4Department of Pulmonary Diseases, University of Groningen, University Medical Center Groningen, Groningen, Netherlands

5Department of Epidemiology, Erasmus Medical Center, Rotterdam, Netherlands

6Institute of Epidemiology II, Helmholtz Zentrum MünchenInstitute of Epidemiology II, Neuherberg, Germany

7Environmental Public Health Division, U.S. Environmental Protection AgencyEnvironmental Public Health Division, Chapel Hill, North Carolina, USA

8Molecular Epidemiology Unit, Helmholtz Zentrum München, Neuherberg, Germany

9Institute for Risk Assessment Sciences, Utrecht UniversityInstitute for Risk Assessment Sciences, Utrecht, Netherlands

10Julius Center for Health Sciences and Primary Care, University Medical Center UtrechtJulius Center for Health Sciences and Primary Care, Netherlands

PDF icon PDF Version (264 KB)

  • Background:
    Long-term air pollution exposure is negatively associated with lung function, yet the mechanisms underlying this association are not fully clear. Differential DNA methylation may explain this association.
    Objectives:
    Our main aim was to study the association between long-term air pollution exposure and DNA methylation.
    Methods:
    We performed a genome-wide methylation study using robust linear regression models in 1,017 subjects from the LifeLines cohort study to analyze the association between exposure to nitrogen dioxide (NO2) and particulate matter (PM2.5, fine particulate matter with aerodynamic diameter &le2.5 μm; PM10, particulate matter with aerodynamic diameter 10 μm) and PM2.5absorbance, indicator of elemental carbon content (estimated with land-use-regression models) with DNA methylation in whole blood (Illumina® HumanMethylation450K BeadChip). Replication of the top hits was attempted in two independent samples from the population-based Cooperative Health Research in the Region of Augsburg studies (KORA).
    Results:
    Depending on the p-value threshold used, we found significant associations between NO2 exposure and DNA methylation for seven CpG sites (Bonferroni corrected threshold p<1.19×107) or for 4,980 CpG sites (False Discovery Rate<0.05). The top associated CpG site was annotated to the PSMB9 gene (i.e., cg04908668). None of the seven Bonferroni significant CpG-sites were significantly replicated in the two KORA-cohorts. No associations were found for PM exposure.
    Conclusions:
    Long-term NO2 exposure was genome-wide significantly associated with DNA methylation in the identification cohort but not in the replication cohort. Future studies are needed to further elucidate the potential mechanisms underlying NO2exposure–related respiratory disease. https://doi.org/10.1289/EHP2045
  • Received: 12 April 2017
    Revised: 13 December 2017
    Accepted: 13 December 2017
    Published: 6 February 2018

    Address correspondence to J.M. Vonk, Department of Epidemiology, University Medical Center Groningen, University of Groningen, Hanzeplein 1, 9700 RB Groningen, Netherlands. Telephone: +31 50 3610934. Email: j.m.vonk@umcg.nl

    Supplemental Material is available online (https://doi.org/10.1289/EHP2045).

    The views in this manuscript do not necessarily represent U.S. Environmental Protection Agency policy.

    The authors declare they have no actual or potential competing financial interest.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (698 KB)
    Zip icon Supplemental Code and Data Zip File (539 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (81 KB)

Introduction

Air pollution is a major concern in public health. Long-term air pollution exposure has been consistently associated with lower function in adults (Ackermann-Liebrich et al. 1997; Adam et al. 2015; Forbes et al. 2009), children (Barone-Adesi et al. 2015; Gehring et al. 2013; Urman et al. 2014) and in those with preexisting lung disease (Nitschke et al. 2016). In the LifeLines cohort study, a large representative sample of the north of Netherlands, a higher level of nitrogen dioxide (NO2) and particulate matter (PM) exposure was associated with lower forced expiratory volume in 1 s (FEV1) and even lower forced vital capacity (FVC), and as a consequence, a higher FEV1/FVC ratio (De Jong et al. 2016).

Activation of oxidant and pro-inflammatory pathways is suggested to be a potential mechanism underlying the acute toxicity of NO2 and PM (Lodovici and Bigagli 2011), yet the knowledge on exact mechanisms underlying the effect of long-term air pollution exposure on health is inconclusive. Although clear evidence exists for the association between genetics and lung function (Hobbs et al. 2017), the genetic background cannot entirely explain the phenotypic variability. Emerging evidence suggests that apart from specific genetic variants, epigenetic alterations in response to environmental exposure is an important determinant of respiratory health (Boezen 2009). DNA methylation, currently the most frequently studied epigenetic mechanism, occurs by the binding of a methyl group to a cytosine-guanine dinucleotide (CpG site) (Griffiths et al. 2011). Changes in DNA methylation may be induced by exposure to air pollution and may alter the gene expression profile. This change in DNA methylation is one plausible mechanism potentially mediating the adverse health effects of air pollution (Holloway et al. 2012). Differential blood DNA methylation in response to air pollution exposure has been reported in environmental (Bind et al. 2014; De Prins et al. 2013), occupational (Sanchez-Guerra et al. 2015; Tarantini et al. 2009), and experimental settings (Ding et al. 2016). The strongest epigenetic signals identified in response to air pollution exposure were found in candidate genes for inflammatory pathways (F3, ICAM-1, TLR-2, IFNγ, IL-6) (Bind et al. 2014), detoxification metabolism system (GST) (Madrigano et al. 2011), lung function (TLR2, GCR) (Lepeule et al. 2014), lung cancer (SATα, NBL2) (Hou et al. 2014), and biological aging (Ward-Caviness et al. 2016).

Despite these findings, no study to date has examined whether the association between long-term air pollution exposure and lung function is mediated by DNA methylation (Tarantini et al. 2009). In this study, we performed a genome-wide methylation study of the long-term air pollution exposure, i.e., NO2 and PM in adults from the LifeLines cohort study. We further investigated to what extent the significant air pollution-associated DNA methylation sites mediated the association between air pollution and lung function.

Methods

Study Population and Design

This study included subjects enrolled in the LifeLines cohort study. LifeLines is a large Dutch population-based cohort study designed to investigate chronic diseases and healthy aging (Scholtens et al. 2015). Detailed information about LifeLines can be obtained at the official website (http://www.lifelines.net). The LifeLines cohort study was approved by the Medical Ethical Committee of the University Medical Center Groningen, Groningen, Netherlands. All participants provided written informed consent.

A subgroup of 1,656 subjects of the LifeLines cohort was selected based on having complete data on lung function and specific environmental exposures as well as on covariates used in the analysis: sex, age, height, and smoking history.

Air Pollution Exposure Assessment

At the baseline visit (2007–2013), the home addresses of all LifeLines individuals were geocoded and GIS (geographic information system)-derived information on distance to the nearest road, traffic intensity, built-up land, population density, and altitude was acquired. The annual average exposure to NO2, PM10 (PM with aerodynamic diameter less than 10 μm), PM2.5 (particles with aerodynamic diameter less than 2.5 μm) and PM2.5absorbance (indicator of elemental carbon content) (Cyrys et al. 2003) was then estimated using land-use regression models, developed in the ESCAPE study. In these models, the GIS-derived information was combined with annual air pollution concentrations from an intensive monitoring campaign in the ESCAPE study (Beelen et al. 2013; Eeftens et al. 2012; Zijlema et al. 2016).

Lung Function Assessment

Lung function was measured with spirometry according to standard operating procedures using the Wellch Allyn SpiroPerfect device (Wellch Allyn version 1.6.0.489, personal computer-based SpiroPerfect with CardioPerfect® Workstation software). FEV1, FVC, FEV1/FVC, and forced expiratory flow between 25–75% of the FVC (FEF25–75) were used as lung-function outcomes.

Covariate Assessment

Age was calculated based on the date of birth as registered in the municipal registries and the date of the baseline visit. Body mass index (BMI) is calculated as weight divided by height squared, as measured during the baseline visit using standardized procedures. Smoking status and cumulative smoking exposure (pack-years of cigarettes smoked) were assessed using the standardized European Community Respiratory Health Survey (ECRHS) questionnaire (Burney et al. 1994). Current smoking was defined as smoking in the last month, and only current smokers with a smoking history greater than five pack-years were included. Never-smokers were defined as having a smoking history of zero pack-years. To optimize the smoking exposure contrast, ex-smokers were not included in this study.

Genome-wide DNA Methylation Assessment

Genome-wide DNA methylation levels were assessed from whole blood of 1,656 subjects using standard methods. Briefly, blood samples were bisulfite-treated (EZ-96 DNA Methylation Kit; Zymo Research Corp.) and subsequently subjected to whole-genome amplification. DNA methylation level for each CpG site was measured using the Illumina Infinium® Human Methylation 450K array (Illumina, Inc.) and expressed quantitatively as βvalue. βValues represent the ratio of the fluorescent signal intensity measured by methylated and unmethylated probes and range from 0 (all copies of the CpG site in the sample are unmethylated) to 1 (all copies of the CpG site in the sample are methylated). Quality control (QC) included removal of samples with probes with a detection pvalue>0.01 in <99% of probes, samples with incorrect sex or SNP prediction, as well as probes with a detection pvalue>0.01, sex chromosome probes, probes measuring SNPs, probes where the CpG itself or the single base extension (SBE) site is an SNP, and cross-reactive probes. A total of 420,938 CpG sites passed the QC filtering criteria. The data were normalized using DASEN implemented in the wateRmelon package in R (R Core Team) (Pidsley et al. 2013).

Replication Analysis

Replication of the top CpG sites associated with air pollution exposure was attempted in two independent samples from the population-based Cooperative Health Research in the Region of Augsburg studies [Kooperative Gesundheitsforschung in der Region Augsburg, Germany (KORA F3 and KORA F4)]. The KORA F3 examinations took place from 2004 to 2005 (Aulchenko et al. 2009; Wichmann et al. 2005), and KORA F4 examinations took place from 2006 to 2008 (Rückert et al. 2011). For both examinations, health surveys were administered, and biospecimens were collected by trained personnel per published methodologies. Informed consent was provided by all participating individuals. All KORA studies were approved by the ethics committee of the Bavarian Medical Association in Munich, Germany. The KORA F3 and KORA F4 samples used for this replication study were nonoverlapping. Air pollution at the residential address was estimated using land-use regression models as developed in the ESCAPE study (Pitchika et al. 2017). DNA methylation was assessed identically for the KORA F3 and KORA F4 samples. Genome-wide DNA methylation measurement at 485,577 genomic sites was performed using the Infinium® HumanMethylation450K BeadChip (Illumina, Inc.) and expressed quantitatively as βvalue. The laboratory process has been described previously (Zeilinger et al. 2013). To preprocess the DNA methylation data, first, 65 probes that represent SNPs were excluded. Next, background correction using minfi, version 1.6.0 (Aryee et al. 2014) was performed, and signals represented by fewer than three functional beads were removed. Data were normalized using quantile normalization (QN) on the raw signal intensities (Lehne et al. 2015). QN was performed on six stratified probe categories based on probe type and color channel (Bibikova et al. 2011) using the R package limma (version 3.16.5; R Core Team). Differences in the signal intensities from Infinium I vs. Infinium II probes designs were corrected using beta-mixture quantile normalization (BMIQ) (Teschendorff et al. 2012) via the R package wateRmelon, version 1.0.3 (R Core Team) (Pidsley et al. 2013). White blood cell (i.e., granulocytes, monocytes, B cells, CD4+ T cells, CD8+ T cells, and natural killer cells) proportions were estimated using the Houseman method (Houseman et al. 2012). To keep in concert with the discovery analyses, individuals with a detection pvalue>0.05 for >1% of the probes were removed. After quality control, 451 samples were retained in KORA F3 and 1,424 in KORA F4.

Statistical Analyses

Statistical analyses were performed using the SPSS statistics software version 23.0 (IBM) and R software version 3.2.4 revised (R Foundation). Robust linear regression models were used to test the cross-sectional association between air pollution (NO2, PM10, PM2.5, and PM2.5absorbance) exposure as a predictor and genome-wide DNA methylation levels as a response. The models were adjusted for sex, age, BMI, current smoking, pack-years, and covariates expected to influence the DNA methylation levels (technical covariates and blood cell composition). The potential technical bias was minimized using principal component analysis applied to the control probes included on the 450K chip (Lehne et al. 2015). We included all PCs that explained >1% of the variance. This resulted in the inclusion of the first 7 PCs that together explained 95.5% of total variance. Additionally, the model was adjusted for the measured white blood cell counts (eosinophils, neutrophils, basophils, lymphocytes, and monocytes) to correct for the cellular heterogeneity of blood samples (Jaffe and Irizarry 2014). We used the Bonferroni corrected threshold pvalue<1.19×107 (0.05/420,938) to correct for the number of CpG sites tested. Sites that passed this threshold were considered genome-wide significant and were investigated further in subsequent analyses. To investigate the sensitivity of the results of the analyses between air pollution exposure and methylation to the model specifications we conducted several sensitivity analyses to the top hits of our analyses (see Supplemental Material for details): a) exclusion of outliers in the DNA-methylation levels, b) additional adjustment for possible confounders (i.e., highest educational level, chronic obstructive pulmonary disease (COPD), asthma, use of respiratory medication), and c) stratification of the models by sex, BMI, and smoking. Replication of the top hits was attempted in two independent samples from the KORA study. In this replication analysis, associations were estimated using robust linear regression models and included the following covariates: sex, age, body mass index, current smoking, pack-years, estimated cell counts, and the first 20 PCs from the control probes to adjust for technical variation (Lehne et al. 2015). As in the discovery analysis, only, never, and current smokers were included in the analysis (ex-smokers were excluded). Significant replication is defined as a pvalue<0.05 in at least one of the replication cohorts, and the direction of the effect should be the same in the discovery and both replication cohorts. In addition, using the software tool provided at https://129.125.135.180:8080/GeneNetwork/pathway.html, we conducted a pathway analysis in which we included all genes annotated to CpG sites with an FDR pvalue<0.01, and we investigated the association between our top methylation-sites and gene expression by searching the tables provided at https://www.genenetwork.nl/biosqtlbrowser/.

Robust linear regression models adjusted for sex, age, height, BMI, sex*age interaction, sex*height interaction, current smoking, and pack-years were used to analyze the cross-sectional association between air pollution exposure and lung function levels, as measured by: FEV1, FVC, FEV1/FVC, and FEF25–75. Two-sided pvalues<0.05 were considered statistically significant.

The potential mediation by significant air pollution-associated methylation sites was assessed using mediation analysis. By applying the bootstrapping method in the “mediation” package in R (version 4.4.6; R Core Team) (Hayes 2009), we verified whether the total effect of a specific air pollutant on a lung function outcome was mediated by DNA methylation at the significant CpG sites. To test this mediation effect, two models were applied, and their estimates were used as input for the mediate function (Figure 1). The first model, the mediator model, assessed the effect of air pollution exposure on DNA methylation (association A) and, the second model, the outcome model, assessed the combined effect of air pollution exposure and the mediator (DNA methylation) on a lung function outcome (associations B and C). A total of 1,000 bootstraps were run to estimate the confidence intervals (CIs) (Mayer et al. 2014). Significant mediation by DNA methylation was considered present when the p-value of the average mediation effect (AME) was <0.05. A p-value between 0.05 and 0.10 was considered as borderline significant.

Flowchart showing association between pollutant exposure, methylation CpG, and lung function.

Figure 1. Model showing the associations tested in the mediation analysis. Association A: association between air pollution and DNA methylation; association B: association between DNA methylation and lung function; association C: association between air pollution and lung function with adjustment for DNA methylation.

Results

Descriptive Statistics

In total, 1,622 subjects passed the quality control of the DNA methylation assessment, and 1,017 subjects had complete data on air pollution exposure, DNA methylation levels, and all included covariates. Table 1 summarizes the baseline characteristics of the 1,017 subjects with complete data and 605 subjects with incomplete data. The mean age of the complete subjects was 47.3 y, 58% were male, and 56% were never smokers. In current smokers, the mean cumulative smoking exposure was 20.5 pack-years. Mean levels of lung function measured at the baseline visit were: FEV1 3.5L (98% of predicted normal values), FVC 3.5L (111% of predicted), FEV1/FVC 73.4%, and FEF25–75 2.8 L/s (70% of predicted) (Quanjer et al. 1993). No significant differences were found between the incomplete and complete subjects. The annual average estimated concentration (range) of exposure was 16.3(9.432.8)μg/m3 for NO2; 24.1(23.727.6)μg/m3 for PM10; 15.5(15.118.2)μg/m3 for PM2.5, and 0.9(0.81.6)×105m1 for PM2.5absorbance. The correlation (Pearson correlation coefficient) among the annual average concentration of pollutants is presented in Supplemental Material (Table S1).

Table 1. Characteristics of subjects with incomplete or complete data from the LifeLines Cohort Study.
Characteristic Incomplete data Complete data Difference
p-Value
N (%) 605 (37.3) 1,017 (62.7)
Age (y) 46.8±10.6 47.3±11.0 0.38
Height (cm) 176.4±9.1 177.2±9.1 0.08
BMI (kg/cm2) 26.4±4.1 26.0±3.8 0.07
Male 341 (56.4) 587 (57.7) 0.59
Smoking status
Never smoking 343 (56.7) 574 (56.4) 0.92
Current smokinga 262 (43.3) 443 (43.6)
Cumulative smokingb 22.0±11.4 20.5±11.8 0.09
Lung function
FEV1 (L) 3.4±0.9 3.5±0.9 0.07
FVC (L) 4.7±1.1 4.7±1.1 0.17
FEV1/FVC (%) 72.8±8.8 73.4±8.6 0.18
FEF25–75 (L/s) 2.7±1.3 2.8±1.3 0.08
Air pollution IQR
NO2 (μg/m3) NA 16.3±3.2 4.66
PM10 (μg/m3) NA 24.1±0.5 0.58
PM2.5 (μg/m3) NA 15.5±0.2 0.13
PM2.5absorbance (105/m) NA 0.9±0.1 0.13

Note: Data are presented as mean±standard deviation (SD) for continuous variables or N (%) for categorical variables; BMI, body mass index; FEV1, forced expiratory volume in 1 second; FCV, forced vital capacity; FEV1/FVC, forced expiratory volume in 1 sec to forced vital capacity ratio; FEF25–75, forced expiratory flow between 25–75% of the FVC; NO2, nitrogen dioxide; PM10, particles with aerodynamic diameter <10 μm; PM2.5, particles with aerodynamic diameter 2.5 μm; PM2.5absorbance, indicator of elemental carbon content.

aCurrent smoking is defined as smoking in the last month and only current smokers with a smoking history greater than five pack-years were included.

bCumulative smoking defined by pack-years of cigarettes smoked in current smokers.

Association between Air Pollution Exposure and Genome-Wide DNA Methylation

NO2 exposure was genome-wide significantly associated with differential DNA methylation at seven CpG sites (mapped to seven different genes) at the Bonferroni corrected threshold pvalue<1.19×107 and with 4,980 CpGs at the False Discovery Rate (FDR) pvalue<0.05 (Table 2, see also Excel Table S1 and Figure S1). Among these top signals, three CpG sites showed a negative association with NO2 exposure: cg04908668 (PSMB9, chr6), cg00344801 (TTC38, chr22), and cg02234653 (AP1S3, chr2); four showed a positive association: cg14938677 (ARF5, chr7), cg18379295 (GNG2, chr14), cg25769469 (PTCD2, chr5), and cg08500171 (BAT2, chr6). The results of the sensitivity analyses on these 7 CpG sites are presented in Tables S2, S3, and S4. Importantly, after removal of outliers in the methylation values 1 CpG (i.e., cg02234653) was no longer significant at the Bonferroni corrected threshold although the effect estimate remained similar (Table S2). None of the genome-wide significant CpG sites associated with NO2 exposure was successfully replicated either in the KORA F3 or in the KORA F4 cohort (Tables 3 and 4). The results of the pathway analyses are presented in Excel Tables S2 and S3. A look-up of these 7 CpG sites in the eQTM table provided at https://www.genenetwork.nl/biosqtlbrowser/ showed that 2 CpGs were associated with gene expression of 3 genes [i.e., cg04908668 was associated with lower expression of Proteasome Subunit Beta 9 (PSMB9) and of Transporter 1, ATP Binding Cassette Subfamily B Member (TAP1) genes, and cg00344801 was associated with higher expression of Tetratricopeptide Repeat Domain 38 (TTC38)] (Table S5). PM10, PM2.5, and PM2.5absorbance exposures were not genome-wide significantly associated with DNA methylation, either at the Bonferroni corrected threshold or at the FDR-threshold (see Table S6 and Excel Table S4 and Table S7 for CpG sites associated with a pvalue<1×105). Given that only NO2 exposure was genome-wide significantly associated with DNA methylation, the subsequent mediation analyses were restricted to NO2.

Table 2. Genome-wide differential DNA methylation associated with NO2 exposure (per 10μg/m3 ) in the LifeLines Cohort Study (n=1,017).
CpG site B±SEa p-Valueb Chr Bp position Gene Location in gene Relation to island
cg04908668 0.012±0.002 7.94×109 6 32823941 PSMB9 Body S_Shore
cg14938677 0.023±0.004 1.05×108 7 127231698 ARF5 3’UTR S_Shelf
cg00344801 0.028±0.005 2.38×108 22 46685728 TTC38 Body Island
cg18379295 0.020±0.004 3.50×108 14 52326155 GNG2 TSS1500 OpenSea
cg25769469 0.035±0.006 3.69×108 5 71643841 PTCD2 Body OpenSea
cg02234653 0.017±0.003 4.07×108 2 224625080 AP1S3 Body OpenSea
cg08500171 0.023±0.004 9.81×108 6 31590674 BAT2 Body S_Shore

aEffect estimates±standard error (SE) of DNA methylation per 10μg/m3 increase in NO2 concentration adjusted for sex, age, BMI, current smoking, pack-years, technical variance, and blood cell composition. Chr, chromosome.

bBonferroni corrected threshold pvalue<1.19×107.

Table 3. Characteristics of the subjects included in the discovery and replication analyses.
Characteristic Discovery cohort Replication cohorts
KORA F3 KORA F4
N 1,017 422 971
Age (y) 47.3±11.0 53.2±9.6 60.9±8.8
BMI (kg/cm2) 26.0±3.8 27.1±4.4 27.8±4.6
Male 587 (57.7) 261 (51.2) 383 (39.4)
Smoking status
Never smoking 574 (56.4) 205 (48.6) 721 (74.3)
Current smokinga 443 (43.6) 217 (51.4) 250 (25.7)
Cumulative smokingb 20.5±11.8 26.2±20.6 30.6±24.4
Air pollution
NO2 (μg/m3) 16.3±3.2 18.3±3.7 18.8±3.9

Note: Data are presented as mean±standard deviation (SD) for continuous variables or N (%) for categorical variables. BMI, body mass index; NO2, nitrogen dioxide.

aCurrent smoking is defined as smoking in the last month and only current smokers with a smoking history greater than five pack-years were included.

bCumulative smoking defined by pack-years of cigarettes smoked in current smokers.

Table 4. Replication analysis: robust linear regression analysis for the association between NO2 and DNA methylation at the 7 genome-wide significant CpG sites in the KORA F3 and F4 cohort.
CpG Discovery cohort (n=1,017) KORA F3 (n=422) KORA F4 (n=971)
B SE p-Value B SE p-Value B SE p-Value
cg04908668 0.012 0.002 7.94×109 0.002 0.002 0.260 0.000 0.002 0.446
cg14938677 0.023 0.004 1.05×108 0.008 0.003 0.013 0.001 0.003 0.394
cg00344801 0.028 0.005 2.38×108 0.002 0.004 0.258 0.001 0.003 0.421
cg18379295 0.020 0.004 3.50×108 0.004 0.004 0.147 0.002 0.003 0.233
cg25769469 0.035 0.006 3.69×108 0.003 0.006 0.292 0.008 0.005 0.056
cg02234653 0.017 0.003 4.07×108 0.003 0.003 0.198 0.002 0.002 0.149
cg08500171 0.023 0.004 9.81×108 0.003 0.004 0.202 0.002 0.003 0.269

Note: Effect estimate per 10μg/m3 increase in NO2 concentration. Robust linear regression analysis adjusted for sex, age, BMI, current smoking, pack-years, technical variance. and blood cell composition.

Association between Air Pollution Exposure and Lung Function

Table 5 shows the association between NO2 exposure and lung function levels (see Table S8 for the associations between all pollutants and lung function). NO2 exposure was borderline significantly associated with FVC (B per 10μg/m3 NO2=106.3, 95% CI=219.16.6, p=0.065) and with FEV1/FVC (B=1.5, 95% CI=0.13.0, p=0.060). FEV1 was not significantly associated with NO2 exposure, indicating that the positive association between NO2 and FEV1/FVC is driven by a stronger negative association with FVC than FEV1. Given that NO2 exposure was associated with FVC and with FEV1/FVC, we examined the mediation by DNA methylation for FEV1, FVC and FEV1/FVC.

Table 5. Effect estimates of NO2 exposure on lung function (FEV1 , FVC, FEV1/FVC and FEF25–75) for 1,017 subjects enrolled in the LifeLines Cohort Study.
Variable NO2
B (95% CI)a p-Value
FEV1 (mL)b 2.4 (94.8, 99.6) 0.962
FVC (mL)b 106.3 (219.1, 6.6) 0.065
FEV1/FVC (%)c 1.5 (0.1, 3.0) 0.06
FEF25–75 (mL/s)d 128.5 (69.8, 326.8) 0.204

aEffect estimates and 95% confidence interval (95% CI) for NO2 are given per 10μg/m3 increase.

bFor FEV1 and FVC, robust regression analysis adjusted for sex, age, height, BMI, sex*age interaction, sex*height interaction, current-smoking, and pack-years.

cFor FEV1/FVC, robust regression analysis adjusted for sex, age, BMI, sex*age interaction, current-smoking, and pack-years.

dFor FEF25–75, robust regression analysis adjusted for sex, age, height, BMI, sex*age interaction, sex*height interaction, current-smoking, pack-years, and FVC.

Mediation Analysis

Mediation analysis showed that one of the seven top CpG sites significantly mediated the association between NO2 exposure and FVC (cg14938677), and 2 CpG sites significantly mediated the association between NO2 and FEV1/FVC (cg14938677 and cg18379295) (Table S9).

Discussion

We performed a cross-sectional genome-wide methylation study in blood to investigate whether long-term air pollution exposure is associated with DNA methylation in the LifeLines cohort study. We further investigated whether the association between air pollution exposure and lung function was mediated by DNA methylation. In our genome-wide methylation study, we identified differential DNA methylation at seven CpG sites to be genome-wide significantly associated with NO2 exposure. After removal of outliers in the methylation values, six CpG sites remained significantly associated with NO2 levels. Unfortunately, none of these associations could be significantly replicated in two independent cohorts. Further, higher levels of NO2 exposure were borderline significantly associated with lower FVC and higher FEV1/FVC levels. Finally, we found one out of seven CpG sites (cg14938677 in ARF5) was a significant mediator between NO2 exposure and FVC, and two CpG sites (cg14938677 in ARF5 and cg18379295 in GNG2) were significant mediators of the association between NO2 exposure and FEV1/FVC.

The top-significant CpG site (cg04908668) identified in our genome-wide methylation study on NO2 maps to PSMB9 (chromosome 6) and is associated with lower gene expression of PSMB9 and TAP1 (https://www.genenetwork.nl/biosqtlbrowser). PSMB9 and TAP1 are suggested to be involved in the pathophysiological mechanisms underlying COPD. Fujino et al. (2012) report both PSMB9 and TAP1 to be differentially expressed in alveolar epithelial type II cells isolated from COPD patients, in comparison with healthy subjects. The putative function of all genes identified in our study is presented in the Supplemental Material (Table S10). Interestingly, 2 of the 7 genome-wide significant CpG sites (i.e., cg14938677 and cg00344801) were also described by Joehanes et al. (2016) in relation to smoking habits, which might indicate they are general markers of inhaled particle exposure.

To date, genome-wide DNA methylation analyses of NO2 allowing a hypothesis-free assessment of epigenetic modifications are scarce. However, relevant evidence comes from a study in children by Gruzieva et al. (2017). In this epigenome-wide meta-analysis of methylation, prenatal NO2 exposure was associated with differential DNA methylation of genes involved in mitochondria and antioxidant defense pathways. Interestingly, in our genome-wide methylation study on NO2, we also identified a CpG (cg25769469) in PTCD2 that is reported to be involved in the mitochondrial RNA metabolism.

The positive association between NO2 exposure and FEV1/FVC found in our study is in line with findings reported in a larger sample of the LifeLines cohort (n=51,855 subjects) (De Jong et al. 2016). In this larger sample, NO2 had a stronger negative association with FVC than with FEV1 resulting in a higher FEV1/FVC. In our current smaller sample, the lack of significant association between NO2 and FEV1 may be the result of low study power (due to smaller sample size and smaller air pollution ranges). FEV1 and FVC are considered early indicators of chronic respiratory disease and predictors for cardiorespiratory mortality (Lee et al. 2011). Clinically, a reduced FVC along with FEV1 within a normal range is indicative (but not specific) of restrictive ventilatory abnormalities (Pellegrino et al. 2005). A comparison with existing studies shows that this restrictive effect is not universally seen. For example, in the ESCAPE study, the negative association between NO2 exposure and FEV1 and FVC are of equal magnitude (Adam et al. 2015). Because the main parameter for the diagnosis of restriction is a low total lung capacity (TLC), further studies including TLC are warranted to better elucidate whether the observed ventilatory pattern associated with NO2 exposure corresponds to a restrictive, obstructive or both types of disorders.

We tested mediation by DNA methylation to confirm our hypothesis that NO2 exposure may affect lung function through effects on DNA methylation. Among the seven differentially methylated CpG sites, two showed suggestive evidence for mediation (cg14938677 in ARF5 and cg18379295 in GNG2). ARF5 is a member of the human ADP-ribosylation factor (ARF) gene family that encodes small guanine nucleotide binding proteins. These proteins activate the phospholipase D (PLD), a critical enzyme involved in various endothelial and epithelial cell functions, such as actin cytoskeleton, vesicle trafficking for secretion, and endocytosis and receptor signaling (Jenkins and Frohman 2005). A family member, PLA1, was found to be significantly increased in plasma membrane of NO2exposed pulmonary artery endothelial cells (Bhat et al. 1990; Sekharam et al. 1991). Furthermore, the redox regulation of bleomycin-induced PLD activation was reported to play a crucial role in the cytotoxicity underlying the idiopathic pulmonary fibrosis (Patel et al. 2011). The GNG2 (G protein subunit gamma 2) gene belongs to the heterotrimeric G protein family that underlies important pathways involved in cell migration, proliferation, differentiation, apoptosis, and responses to external signals (Olate and Allende 1991). Its distinct isoform subunits α, β, and γ are selectively expressed and enriched in different tissues including white blood cells and lung (Modarressi et al. 2000). To date, we do not know whether differential DNA methylation at this particular CpG site (cg18379295), located in the transcription start site of the gene, results in any functional variation in lung function. However, GNG2 was reported to be involved in airway hyper-responsiveness and inflammation elicited by an antigen challenge in a rabbit model of asthma (Nino et al. 2012). Furthermore, upon activation by G protein-coupled receptors (GPCRs), both free Gα and Gβγ subunits regulate important signaling pathways like the MAPK kinase cascade. This MAPK kinase cascade is involved in various immune and inflammatory cell functions and is a plausible mechanism linking air pollution exposure and respiratory and cardiovascular outcomes (Carmona et al. 2014).

A large number of studies have linked short- and mid-term PM exposure to global and gene-specific DNA methylation (Baccarelli et al. 2009; Bellavia et al. 2013; Chen et al. 2016; Peng et al. 2016; Wang et al. 2016). A genome-wide meta-analysis of DNA methylation and PM2.5 identified twelve genes regulating pathways involved in tumor development, inflammatory stimuli, pulmonary disorders and glucose metabolism (Panni et al. 2016). However, few studies have examined this association in the context of a long-term exposure window (Chi et al. 2016; Ward-Caviness et al. 2016). Although we found no genome-wide significant effect of PM exposure (considering all different size fractions) on DNA methylation, many CpG sites had suggestive effects, especially in response to PM2.5 (Table S8). Possibly, the relatively small range of PM levels and consequently a modest exposure contrast in LifeLines cohort may explain this lack of association. Future genome-wide methylation studies conducted in cohorts with a broader range of PM exposure is needed to clarify this association.

Our study is the largest genome-wide methylation study of air pollution exposure in adults, and the first study to assess the mediation effect of DNA methylation in the association between air pollution and the FEV1/FVC ratio. However, this study has some limitations. We used the individual’s home address as basis for the air pollution exposure estimates, ignoring the fact that a person could spend time in another environment (e.g., while traveling or working), which might lead to some degree of exposure misclassification at the individual level (Sunyer 2009). Interestingly, this exposure misclassification may lead to overestimation of the mediation effect (Valeri et al. 2017) when the methylation levels at our identified CpG sites are better biomarkers of personal NO2 exposure than the estimated NO2 exposure using land-use-regression models. The results of the mediation analyses should thus be interpreted with caution. In addition, because our study was cross-sectional in design, the inference of causality from these measures could be questionable.

We also recognize that ambient air pollution is a complex mixture and the effects attributed to some specific component might be influenced by the underlying toxicity of the full mixture of all pollutants. In this study, we estimated the association between various pollutants and DNA methylation, but only found genome-wide significant associations with NO2. The moderate to high correlation between NO2 and other pollutants prohibits the use of multipollutant models, and thus we cannot completely disentangle the independent pollutant effect. However, the top CpG sites differ for the different pollutants, indicating that each pollutant may have its own specific methylation target sites.

Another potential limitation of this study is the use of DNA methylation in blood samples when the outcome of interest is lung function. To what extent these epigenetic changes that we observe in peripheral blood cells reflect changes in DNA methylation in target tissues like the lung merits further investigation. Finally, the identified associations between NO2 exposure and DNA methylation did not replicate in two independent cohorts from the German KORA study. This lack of replication could be explained by the differences in age, gender, BMI, and smoking habits between the cohorts (see Table 3), and therefore more replication studies should be performed to validate these findings.

Conclusions

In the largest genome-wide methylation study to date, long-term NO2 exposure was associated with differential DNA methylation in blood in 1,017 subjects from the LifeLines cohort study. Among the significant NO2associated DNA methylation sites, 2 CpGs can be considered potential mediators of the association between NO2 exposure and lung function. In this perspective, replication of these findings in other cohorts is necessary to elucidate the suggested role of epigenetic variability in the pathogenesis of NO2exposure-related respiratory disease.

Acknowledgments

This study was funded by the Dutch Lung Foundation (grant number 4.1.13.007) and the Science Without Borders Brazilian program (grant number 234331/2014-3), as well as the São Paulo Research Foundation (FAPESP), (grant number 2016/13384-0) to AJFC Lichtenfels. The LifeLines Biobank initiative has been possible by funds from FES (Fonds Economische Structuurversterking), SNN (Samenwerkingsverband Noord Nederland), and REP (Ruimtelijk Economisch Programma).

References

Ackermann-Liebrich U, Leuenberger P, Schwartz J, Schindler C, Monn C, Bolognini G, et al. 1997. Lung function and long term exposure to air pollutants in Switzerland. Study on air pollution and lung diseases in adults (SAPALDIA) team. Am J Respir Crit Care Med 155(1):122–129, PMID: 9001300, 10.1164/ajrccm.155.1.9001300.

Adam M, Schikowski T, Carsin AE, Cai Y, Jacquemin B, Sanchez M, et al. 2015. Adult lung function and long-term air pollution exposure. ESCAPE: a multicentre cohort study and meta-analysis. Eur Respir J 45(1):38–50, PMID: 25193994, 10.1183/09031936.00130014.

Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, et al. 2014. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics 30(10):1363–1369, PMID: 24478339, 10.1093/bioinformatics/btu049.

Aulchenko YS, Ripatti S, Lindqvist I, Boomsma D, Heid IM, Pramstaller PP, et al. 2009. Loci influencing lipid levels and coronary heart disease risk in 16 European population cohorts. Nat Genet 41(1):47–55, PMID: 19060911, 10.1038/ng.269.

Baccarelli A, Wright RO, Bollati V, Tarantini L, Litonjua AA, Suh HH, et al. 2009. Rapid DNA methylation changes after exposure to traffic particles. Am J Respir Crit Care Med 179(7):572–578, PMID: 19136372, 10.1164/rccm.200807-1097OC.

Barone-Adesi F, Dent JE, Dajnak D, Beevers S, Anderson HR, Kelly FJ, et al. 2015. Long-term exposure to primary traffic pollutants and lung function in children: cross-sectional study and meta-analysis. PLoS One 10(11):e0142565, PMID: 26619227, 10.1371/journal.pone.0142565.

Beelen R, Hoek G, Vienneau D, Eeftens M, Dimakopoulou K, Pedeli X, et al. 2013. Development of NO2 and NOx land use regression models for estimating air pollution exposure in 36 study areas in Europe — The ESCAPE project. Atmospheric Environment 72:10–23, 10.1016/j.atmosenv.2013.02.037.

Bellavia A, Urch B, Speck M, Brook RD, Scott JA, Albetti B, et al. 2013. DNA hypomethylation, ambient particulate matter, and increased blood pressure: findings from controlled human exposure experiments. J Am Heart Assoc 2(3):e000212, PMID: 23782920, 10.1161/JAHA.113.000212.

Bhat GB, Patel M, Block ER. 1990. Exposure of pulmonary artery endothelial cells to nitrogen dioxide activates phospholipase A1. J Biochem Toxicol 5(1):67–69, PMID: 2402004, 10.1002/jbt.2570050111.

Bibikova M, Barnes B, Tsan C, Ho V, Klotzle B, Le JM, et al. 2011. High density DNA methylation array with single CpG site resolution. Genomics 98(4):288–295, PMID: 21839163, 10.1016/j.ygeno.2011.07.007.

Bind MA, Lepeule J, Zanobetti A, Gasparrini A, Baccarelli A, Coull BA, et al. 2014. Air pollution and gene-specific methylation in the Normative Aging Study: association, effect modification, and mediation analysis. Epigenetics 9(3):448–458, PMID: 24385016, 10.4161/epi.27584.

Boezen HM. 2009. Genome-wide association studies: what do they teach us about asthma and chronic obstructive pulmonary disease? Proc Am Thorac Soc 6(8):701–703, PMID: 20008879, 10.1513/pats.200907-058DP.

Burney PG, Luczynska C, Chinn S, Jarvis D. 1994. The European Community Respiratory Health Survey. Eur Respir J 7(5):954–960, PMID: 8050554, 10.1183/09031936.94.07050954.

Carmona JJ, Sofer T, Hutchinson J, Cantone L, Coull B, Maity A, et al. 2014. Short-term airborne particulate matter exposure alters the epigenetic landscape of human genes associated with the mitogen-activated protein kinase network: a cross-sectional study. Environ Health 13:94, PMID: 25395096, 10.1186/1476-069X-13-94.

Chen R, Meng X, Zhao A, Wang C, Yang C, Li H, et al. 2016. DNA hypomethylation and its mediation in the effects of fine particulate air pollution on cardiovascular biomarkers: a randomized crossover trial. Environ Int 94:614–619, PMID: 27397927, 10.1016/j.envint.2016.06.026.

Chi GC, Liu Y, MacDonald JW, Barr RG, Donohue KM, Hensley MD, et al. 2016. Long-term outdoor air pollution and DNA methylation in circulating monocytes: results from the Multi-Ethnic Study of Atherosclerosis (MESA). Environ Health 15(1):119, PMID: 27903268, 10.1186/s12940-016-0202-4.

Cyrys J, Heinrich J, Hoek G, Meliefste K, Lewné M, Gehring U, et al. 2003. Comparison between different traffic-related particle indicators: elemental carbon (EC), PM2.5 mass, and absorbance. J Expo Anal Environ Epidemiol 13(2):134–143, PMID: 12679793, 10.1038/sj.jea.7500262.

De Jong K, Vonk JM, Zijlema WL, Stolk RP, van der Plaat DA, Hoek G, et al. 2016. Air pollution exposure is associated with restrictive ventilatory patterns. Eur Respir J 48(4):1221–1224, PMID: 27492831, 10.1183/13993003.00556-2016.

De Prins S, Koppen G, Jacobs G, Dons E, Van de Mieroop E, Nelen V, et al. 2013. Influence of ambient air pollution on global DNA methylation in healthy adults: a seasonal follow-up. Environ Int 59:418–424, PMID: 23917442, 10.1016/j.envint.2013.07.007.

Ding R, Jin Y, Liu X, Zhu Z, Zhang Y, Wang T, et al. 2016. Characteristics of DNA methylation changes induced by traffic-related air pollution. Mutat Res Genet Toxicol Environ Mutagen 796:46–53, PMID: 26778509, 10.1016/j.mrgentox.2015.12.002.

Eeftens M, Beelen R, de Hoogh K, Bellander T, Cesaroni G, Cirach M, et al. 2012. Development of Land Use Regression models for PM(2.5), PM(2.5) absorbance, PM(10) and PM(coarse) in 20 European study areas; results of the ESCAPE project. Environ Sci Technol 46(20):11195–11205, PMID: 22963366, 10.1021/es301948k.

Forbes LJ, Kapetanakis V, Rudnicka AR, Cook DG, Bush T, Stedman JR, et al. 2009. Chronic exposure to outdoor air pollution and lung function in adults. Thorax 64(8):657–663, PMID: 19359266, 10.1136/thx.2008.109389.

Fujino N, Ota C, Takahashi T, Suzuki T, Suzuki S, Yamada M, et al. 2012. Gene expression profiles of alveolar type II cells of chronic obstructive pulmonary disease: a case-control study. BMJ Open 2(6):pii: e001553, PMID: 23117565, 10.1136/bmjopen-2012-001553.

Gehring U, Gruzieva O, Agius RM, Beelen R, Custovic A, Cyrys J, et al. 2013. Air pollution exposure and lung function in children: the ESCAPE project. Environ Health Perspect 121(11–12):1357–1364, PMID: 24076757, 10.1289/ehp.1306770.

Griffiths AJF, Wessler SR, Carroll SB, Doebley J. 2011. Regulation of gene expression in eukaryotes. In: Introduction to Genetic Analysis, 10th revised edition. New York, NY:W.H. Freeman & Co., 455–456.

Gruzieva O, Xu CJ, Breton CV, Annesi-Maesano I, Antó JM, Auffray C, et al. 2017. Epigenome-wide meta-analysis of methylation in children related to prenatal NO2 air pollution exposure. Environ Health Perspect 125(1):104–110, PMID: 27448387, 10.1289/EHP36.

Hayes AF. 2009. Beyond Baron and Kenny: statistical mediation analysis in the new millennium. Communication Monographs Volume 76 (4):408–420, 10.1080/03637750903310360, [accessed 8 November 2016].

Hobbs BD, de Jong K, Lamontagne M, Bossé Y, Shrine N, Artigas MS, et al. 2017. Genetic loci associated with chronic obstructive pulmonary disease overlap with loci for lung function and pulmonary fibrosis. Nat Genet 49(3):426–432, PMID: 28166215, 10.1038/ng.3752.

Holloway JW, Savarimuthu Francis S, Fong KM, Yang IA. 2012. Genomics and the respiratory effects of air pollution exposure. Respirology 17(4):590–600, PMID: 22404320, 10.1111/j.1440-1843.2012.02164.x.

Hou L, Zhang X, Zheng Y, Wang S, Dou C, Guo L, et al. 2014. Altered methylation in tandem repeat element and elemental component levels in inhalable air particles. Environ Mol Mutagen 55(3):256–265, PMID: 24273195, 10.1002/em.21829.

Houseman EA, Accomando WP, Koestler DC, Christensen BC, Marsit CJ, Nelson HH, et al. 2012. DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinformatics 13:86, PMID: 22568884, 10.1186/1471-2105-13-86.

Jaffe AE, Irizarry RA. 2014. Accounting for cellular heterogeneity is critical in epigenome-wide association studies. Genome Biol 15(2):R31, PMID: 24495553, 10.1186/gb-2014-15-2-r31.

Jenkins GM, Frohman MA. 2005. Phospholipase D: a lipid centric review. Cell Mol Life Sci 62(19-20):2305–2316. Review, PMID: 16143829, 10.1007/s00018-005-5195-z.

Joehanes R, Just AC, Marioni RE, Pilling LC, Reynolds LM, Mandaviya PR, et al. 2016. Epigenetic Signatures of Cigarette Smoking. Circ Cardiovasc Genet 9(5):436–447, PMID: 27651444, 10.1161/CIRCGENETICS.116.001506.

Lee HM, Truong ST, Wong ND. 2011. Evidence of lung function for stratification of cardiovascular disease risk. Korean Circ J 41(4):171–174, PMID: 21607165, 10.4070/kcj.2011.41.4.171.

Lehne B, Drong AW, Loh M, Zhang W, Scott WR, Tan ST, et al. 2015. A coherent approach for analysis of the Illumina HumanMethylation450 BeadChip improves data quality and performance in epigenome-wide association studies. Genome Biol 16:37, PMID: 25853392, 10.1186/s13059-015-0600-x.

Lepeule J, Bind MA, Baccarelli AA, Koutrakis P, Tarantini L, Litonjua A, et al. 2014. Epigenetic influences on associations between air pollutants and lung function in elderly men: the normative aging study. Environ Health Perspect 122(6):566–572, PMID: 24602767, 10.1289/ehp.1206458.

LifeLines Cohort Study. 2016 Home page. http://www.lifelines.net [accessed 8 November 2016].

Lodovici M, Bigagli E. 2011. Oxidative stress and air pollution exposure. J Toxicol 2011:487074, PMID: 21860622, 10.1155/2011/487074.

Madrigano J, Baccarelli A, Mittleman MA, Wright RO, Sparrow D, Vokonas PS, et al. 2011. Prolonged exposure to particulate pollution, genes associated with glutathione pathways, and DNA methylation in a cohort of older men. Environ Health Perspect 119(7):977–982, PMID: 21385671, 10.1289/ehp.1002773.

Mayer A, Thoemmes F, Rose N, Steyer R, West SG. 2014. Theory and analysis of total, direct, and indirect causal effects. Multivariate Behav Res 49(5):425–442. PMID: 26732357, http://www.tandfonline.com/doi/abs/10.1080/00273171.2014.931797 [accessed 8 November 2016].

Modarressi MH, Taylor KE, Wolfe J. 2000. Cloning, characterization, and mapping of the gene encoding the human G protein gamma 2 subunit. Biochem Biophys Res Commun 272(2):610–615, PMID: 10833460, 10.1006/bbrc.2000.2832.

Nino G, Hu A, Grunstein JS, McDonough J, Kreiger PA, Josephson MB, et al. 2012. G protein βγ-subunit signaling mediates airway hyperresponsiveness and inflammation in allergic asthma. PLoS One 7(2):e32078, PMID: 22384144, 10.1371/journal.pone.0032078.

Nitschke M, Appleton SL, Li Q, Tucker GR, Shah P, Bi P, et al. 2016. Lung function reductions associated with motor vehicle density in chronic obstructive pulmonary disease: a cross-sectional study. Respir Res 17(1):138, PMID: 27776510, 10.1186/s12931-016-0451-3.

Olate J, Allende JE. 1991. Structure and function of G proteins. Pharmacol Ther 51(3):403–419, PMID: 1792242, 10.1016/0163-7258(91)90068-W.

Panni T, Mehta AJ, Schwartz JD, Baccarelli AA, Just AC, Wolf K, et al. 2016. Genome-wide analysis of DNA methylation and fine particulate matter air pollution in three study populations: KORA F3, KORA F4, and the Normative Aging Study. Environ Health Perspect 124(7):983–990, PMID: 26731791, 10.1289/ehp.1509966.

Patel RB, Kotha SR, Sherwani SI, Sliman SM, Gurney TO, Loar B, et al. 2011. Pulmonary fibrosis inducer, bleomycin, causes redox-sensitive activation of phospholipase D and cytotoxicity through formation of bioactive lipid signal mediator, phosphatidic acid, in lung microvascular endothelial cells. Int J Toxicol 30(1):69–90, PMID: 21131602, 10.1177/1091581810388850.

Pellegrino R, Viegi G, Brusasco V, Crapo RO, Burgos F, Casaburi R, et al. 2005. Interpretative strategies for lung function tests. Eur Respir J 26(5):948–968, PMID: 16264058, 10.1183/09031936.05.00035205.

Peng C, Bind MC, Colicino E, Kloog I, Byun HM, Cantone L, et al. 2016. Particulate air pollution and fasting blood glucose in nondiabetic individuals: associations and epigenetic mediation in the Normative Aging Study, 2000-2011. Environ Health Perspect 124(11):1715–1721, PMID: 27219535, 10.1289/EHP183.

Pidsley R, Y Wong CC, Volta M, Lunnon K, Mill J, Schalkwyk LC. 2013. A data-driven approach to preprocessing Illumina 450K methylation array data. BMC Genomics 14:293, PMID: 23631413, 10.1186/1471-2164-14-293.

Pitchika A, Hampel R, Wolf K, Kraus U, Cyrys J, Babisch W, et al. 2017. Long-term associations of modeled and self-reported measures of exposure to air pollution and noise at residence on prevalent hypertension and blood pressure. Sci Total Environ 593-594:337–346, PMID: 28346907, 10.1016/j.scitotenv.2017.03.156.

Rückert IM, Heier M, Rathmann W, Baumeister SE, Döring A, Meisinger C. 2011. Association between markers of fatty liver disease and impaired glucose regulation in men and women from the general population: the KORA-F4-study. PLoS One 6(8):e22932, PMID: 21850244, 10.1371/journal.pone.0022932.

Quanjer PH, Tammeling GJ, Cotes JE, Pedersen OF, Peslin R, Yernault JC. 1993. Lung volumes and forced ventilatory flows. Eur Respir J Suppl 16:5–40, PMID: 8499054, 10.1183/09041950.005s1693.

Sanchez-Guerra M, Zheng Y, Osorio-Yanez C, Zhong J, Chervona Y, Wang S, et al. 2015. Effects of particulate matter exposure on blood 5-hydroxymethylation: results from the Beijing truck driver air pollution study. Epigenetics 10(7):633–642, PMID: 25970091, 10.1080/15592294.2015.1050174.

Scholtens S, Smidt N, Swertz MA, Bakker SJ, Dotinga A, Vonk JM, et al. 2015. Cohort profile: LifeLines, a three-generation cohort study and biobank. Int J Epidemiol 44(4):1172–1180, PMID: 25502107, 10.1093/ije/dyu229.

Sekharam KM, Patel JM, Block ER. 1991. Plasma membrane-specific phospholipase A1 activation by nitrogen dioxide in pulmonary artery endothelial cells. Toxicol Appl Pharmacol 107(3):545–554, PMID: 2000640, 10.1016/0041-008X(91)90317-8.

Sunyer J. 2009. Lung function effects of chronic exposure to air pollution. Thorax 64(8):645–646, PMID: 19638558, 10.1136/thx.2009.115071.

Tarantini L, Bonzini M, Apostoli P, Pegoraro V, Bollati V, Marinelli B, et al. 2009. Effects of particulate matter on genomic DNA methylation content and iNOS promoter methylation. Environ Health Perspect 117(2):217–222. Erratum in: Environ Health Perspect 2009. 117(4):A143, PMID: 19270791, 10.1289/ehp.11898.

Teschendorff AE, Marabita F, Lechner M, Bartlett T, Tegner J, Gomez-Cabrero D, et al. 2012. A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data. Bioinformatics 29(2):189–196, PMID: 23175756, 10.1093/bioinformatics/bts680.

Urman R, McConnell R, Islam T, Avol EL, Lurmann FW, Vora H, et al. 2014. Associations of children’s lung function with ambient air pollution: joint effects of regional and near-roadway pollutants. Thorax 69(6):540–547, PMID: 24253832, 10.1136/thoraxjnl-2012-203159.

Valeri L, Reese SL, Zhao S, Page CM, Nystad W, Coull BA, et al. 2017. Misclassified exposure in epigenetic mediation analyses. Does DNA methylation mediate effects of smoking on birthweight? Epigenomics 9(3):253–265, PMID: 28234025, 10.2217/epi-2016-0145.

Wang C, Chen R, Cai J, Shi J, Yang C, Tse LA, et al. 2016. Personal exposure to fine particulate matter and blood pressure: a role of angiotensin converting enzyme and its DNA methylation. Environ Int 94:661–666, PMID: 27397929, 10.1016/j.envint.2016.07.001.

Ward-Caviness CK, Nwanaji-Enwerem JC, Wolf K, Wahl S, Colicino E, Trevisi L, et al. 2016. Long-term exposure to air pollution is associated with biological aging. Oncotarget 7(46):74510–74525, PMID: 27793020, 10.18632/oncotarget.12903.

Wichmann HE, Gieger C, Illig T. 2005. KORA-gen–resource for population genetics, controls and a broad spectrum of disease phenotypes. Gesundheitswesen 67 (Suppl 1):S26–S30, PMID: 16032514, 10.1055/s-2005-858226.

Zeilinger S, Kühnel B, Klopp N, Baurecht H, Kleinschmidt A, Gieger C, et al. 2013. Tobacco smoking leads to extensive genome-wide changes in DNA methylation. PloS one 8(5):e63812, PMID: 23691101, 10.1371/journal.pone.0063812.

Zijlema WL, Smidt N, Klijs B, Morley DW, Gulliver J, de Hoogh K, et al. 2016. The LifeLines cohort study: a resource providing new opportunities for environmental epidemiology. Arch Public Health 74:32, PMID: 27482379, 10.1186/s13690-016-0144-x.

Concentrated Ambient PM2.5-Induced Inflammation and Endothelial Dysfunction in a Murine Model of Neural IKK2 Deficiency

Author Affiliations open

1Department of Medicine Cardiology Division, University of Maryland School of Medicine, Baltimore, Maryland, USA

2Department of Environmental Health, School of Public Health, Fudan University, Shanghai, China

3Department of Electrocardiography, The People’s Hospital of Guangxi Zhuang Autonomous Region, Nanning, China

4Department of Occupational and Environmental Health, School of Public Health, Nantong University, Nantong, China

5Department of Pathology, Hubei University of Science and Technology, Xianning, Hubei, China

6Department of Endocrinology, The People’s Hospital of Zhengzhou University (Henan Provincial People’s Hospital), Zhengzhou, Henan, China

7Department of Neurology, Zhongshan Hospital, Fudan University, Shanghai, China

8Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, Maryland, USA

PDF icon PDF Version (2.4 MB)

  • Background:
    Exposure to ambient fine particulate matter (PM2.5) is associated with cardiovascular mortality, but underlying pathophysiologic mechanisms are not fully understood. Hypothalamic inflammation, characterized by the activation of Inhibitor kappaB kinase 2/Nuclear factor kappaB (IKK2/NFκB) signaling pathway, may play an important role in the pathogenesis of cardiovascular diseases. We recently demonstrated that hypothalamic inflammation is increased in mice exposed to concentrated ambient PM2.5 (CAP).
    Objectives:
    In the present study, we used a neuron-specific IKK2 knockout mouse model to examine the role of neural IKK2 expression and hypothalamic inflammation in the pathophysiologic effects of PM2.5.
    Methods:
    We assessed inflammatory and vascular responses in NestincreIKK2flox/flox (IKK2NeuKO) and littermate NestincreIKK2flox/+ (control) mice after 4 mo of exposure to filtered air (FA) or CAP.
    Results:
    CAP exposure was associated with significantly higher tumor necrosis factor-α (TNFα) and interleukin (IL)-6 mRNA in the hypothalamus of control mice, but not IKK2NeuKO mice. In addition, CAP exposure–induced increases in bronchoalveolar lavage fluid (BALF) leukocytes, pulmonary macrophage infiltration and IL-6 expression, plasma TNFα and IL1β levels, adipose macrophage infiltration and IL1β expression, and endothelial dysfunction were reduced or absent in IKK2NeuKO mice compared with controls.
    Conclusions:
    Our findings support a role of neural IKK2 in CAP exposure–induced local and systemic pro-inflammatory cytokine expression, pulmonary and adipose inflammation, and endothelial dysfunction, thus providing insight into pathophysiologic mechanisms that may mediate effects of PM2.5 exposure. https://doi.org/10.1289/EHP2311
  • Received: 8 June 2017
    Revised: 11 December 2017
    Accepted: 12 December 2017
    Published: 5 February 2018

    Address correspondence to Z. Ying, HSFII S022, 20 Penn St., Baltimore, MD 21201, USA. Telephone: 410-328-2063. Fax: 410-328-1048. Email: yingzhekang@hotmail.com

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

Introduction

Exposure to ambient fine particulate matter (PM2.5) is associated with increased risk of cardiovascular morbidity and mortality (Brook et al. 2010). However, while the majority of inhaled PM2.5 are known to deposit in the airway and not enter the systemic circulation (U.S. EPA 2009), mechanisms that link PM2.5 with cardiovascular effects remain elusive. The putative mechanisms for this include a) egress from the lung of PM2.5 components, b) autonomic nervous system (ANS) dysfunction, and c) inflammation (Brook et al. 2010). Notably, despite that inhaled nanoparticles have been identified in remote organs, the pathophysiological role of these nanoparticles remain controversial, as their concentrations are low and inconsistent (Kreyling et al. 2009; Oberdörster et al. 2002; Oberdörster and Utell 2002; Semmler-Behnke et al. 2008). PM2.5 exposure is associated with decreased heart rate variability (Brook 2005; Chen and Hwang 2005; Godleski et al. 2000; Gong et al. 2008; Magari et al. 2001; Sivagangabalan et al. 2011; Tankersley et al. 2004), indicating sympathetic nervous system (SNS) activation and/or parasympathetic nervous system withdrawal (Brook 2008; Karemaker 1999). However, because the nervous system can adapt to chronic stimulation, ANS dysfunction has been thought to play a trivial role in the mediation of the long-term health effects of PM2.5 exposure (Brook et al. 2010). Evidence that cardiovascular effects of PM2.5 are mediated by inflammation is stronger, given the established role of PM2.5 in the of various cardiovascular diseases and evidence of inflammatory responses to PM2.5 exposure in humans (Peters et al. 2001; Pope 2000; Utell et al. 2002) and various animal models (Kennedy et al. 1998; Quay et al. 1998; van Eeden et al. 2001; Vogel et al. 2005).

Putative effect mechanisms that involve ANS dysfunction or inflammation are not mutually exclusive, but may affect the same or different stages of pathogenesis, and may have synergistic effects (Brook et al. 2010). For example, ANS dysfunction may impact various inflammatory responses (Lambert et al. 2015; Pavlov and Tracey 2017), and increases in circulating pro-inflammatory cytokines/mediators resulting from PM2.5related pulmonary or systemic inflammation may cause ANS dysfunction (Dunn 2000; Turnbull and Rivier 1999; Zhang et al. 2003). More importantly, there is emerging evidence that ANS dysfunction may also play a critical role in the long-term regulation of cardiovascular function (DiBona 2013; Froeschl et al. 2013; Simonds and Cowley 2013), which appears to be dependent on an inflammatory response in the hypothalamus. For example, rats and mice fed a high-fat diet (HFD) showed rapid increases in hypothalamic inflammation, which may contribute to HFD feeding-induced long-term effects, including obesity and insulin resistance (Thaler et al. 2012). Inhibition or deletion of IKK2 in some hypothalamic neurons was sufficient to abolish HFD feeding-induced ANS dysfunction, and subsequently, normalized blood pressure in a mouse model (Purkayastha et al. 2011). Collectively, these data have highlighted a crucial role of an IKK2-dependent hypothalamic inflammatory response in the pathophysiology of various cardiovascular diseases.

The IKK2/NFκB signaling pathway is pivotal in both acute and chronic inflammatory responses (Pahl 1999). Previous mechanistic studies by our laboratory (Kampfrath et al. 2011; Wang et al. 2017) and others (Feng et al. 2017; Jeong et al. 2017; Song et al. 2017; Zhang et al. 2017) have demonstrated that exposure to ambient PM2.5 activates the IKK2/NFκB pathway, which is believed to play a key role the pathophysiologic effects of PM2.5 (Brook et al. 2010) in various tissues. Notably, we recently reported that chronic exposure of C57Bl/6J mice to ambient PM2.5 resulted in activation of IKK2/NFκB pathway in the hypothalamus, abnormal SNS activation, and hypertension (Ying et al. 2014), while in the KKay mouse model of type 2 diabetes, intracerebroventricular injection of an IKK2 inhibitor reduced PM2.5 exposure–induced abnormalities in glucose homeostasis (Liu et al. 2014). These findings suggest that an IKK2-dependent hypothalamic inflammatory response may be crucial in the development of adverse health effects in response to ambient PM2.5 exposure. In the present study, to verify the role of hypothalamic inflammation in the pathophysiologic effects of ambient PM2.5, we exposed NestincreIKK2flox/flox (IKK2NeuKO) mice and littermate controls (NestincreIKK2flox/+) to filtered air (FA) or concentrated ambient PM2.5 (CAP), and assessed their inflammatory and vascular responses.

Methods

Animals

University of Maryland, Baltimore (UMB) is an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-accredited institution. All procedures of this study were approved by the Institutional Animal Care and Use Committee at UMB, and all the animals were treated humanely and with regard for alleviation of suffering. Nestin-cre transgenic mice in the C57Bl/6J background were obtained from Jackson Laboratories (Stock No.: 003,771). IKK2flox/flox mice were generated as previously described (Li et al. 2003) and back-crossed with C57Bl/6J for three generations. NestincreIKK2flox/flox (IKK2NeuKO) and littermate NestincreIKK2flox/+ (control) mice were generated through in-house crossing of NestincreKK2flox/+ and IKK2flox/flox. The pups were genotyped at the age of 1 wk by polymerase chain reaction (PCR), and the genotyping results were verified before exposure to FA/CAP.

Concentrated Ambient PM2.5 Exposure

IKK2NeuKO and control mice (male, 8–12 wk old) were grouped into four age-matched groups and randomly designated to be subjected to exposure to FA (six IKK2NeuKO and seven controls) or CAP (seven IKK2NeuKO and seven controls) from May 2016 to September 2016 for a total duration of 4 mo in a mobile trailer with a 12-h light/12-h dark cycle, temperatures of 18–25°C, and relative humidity of 40–60%. Because we failed to observe CAP exposure–induced impairment of glucose tolerance in female mice, male mice only were used in the present study. The mobile trailer was located on the campus of the UMB. Animal exposure and the monitoring of exposure atmosphere and ambient aerosol were performed as previously described using a versatile aerosol concentration enrichment system that was modified for long-term exposures (Geller et al. 2005; Maciejczyk et al. 2005). The exposure protocol comprised exposures for 6 h/d, 5 d/wk (no exposure took place during weekends). Ambient aerosol and exposure atmosphere were monitored as previously described (Maciejczyk et al. 2005). The elemental composition of CAP was determined by inductively coupled plasma mass spectroscopy (for trace element analysis as previously described (Mirowsky et al. 2013; Wang et al. 2017).

Bronchoalveolar Lavage and Lung Histopathology

After euthanasia by overdose of isoflurane, the mouse trachea was cannulated, and the right primary bronchus was closed off with a ligature. Through the tracheal cannula, 0.5 mL sterile phosphate-buffered saline with 0.1mM ethylenediaminetetraacetic acid (EDTA) was instilled and withdrawn to recover bronchoalveolar lavage fluid (BALF). This was repeated twice. The total number of cells in the collected BALF (around 1.5 mL) was estimated using a hemocytometer. Cytospin slides were prepared using Shandon Cytospin 3™ (Shandon) and stained with Diff-Quik solution (EMS). Differential cell counts for neutrophils, eosinophils, macrophages/monocytes, and lymphocytes were assessed by a pathologist who was blinded to the grouping.

Following BALF collection, the right lung was harvested. About one-third were cut and fixed with 4% paraformaldehyde, and the rest were snap frozen in liquid nitrogen and then kept at 80°C. To assess the inflammation in the lung, tissue blocks were embedded in paraffin, 5μm-thick sections were cut, and the sections were subjected to hematoxylin and eosin staining. Three consecutive sections per sample were used for histopathology. Images covering all the tissue area were taken by a laboratory technician who was blinded to the grouping, and all images were then sent to and quantitated by the pathologist (blinded to the grouping, too).

Plasma Cytokine Analysis

To rapidly harvest blood samples for cytokine analysis, mouse retroorbital blood samples were collected in EDTA-treated tubes and subjected to centrifugation for 15 min at 2,000×g using a refrigerated centrifuge. The supernatants (plasma) were then transferred to new tubes and stored at 80°C until assessments of cytokine levels. Plasma cytokine levels were assessed using BD™ Cytometric Bead Array Kit (BD Biosciences) per manufacturer’s instructions. Briefly, 25 μL/mouse plasma were incubated with the beads, and signaling was quantified by flow cytometry.

Real-Time Polymerase Chain Reactions

Mouse heads were removed before bronchoalveolar lavage. Hypothalamus, hippocampus, cortex, and olfactory bulb were immediately isolated. Epididymal adipose tissues were harvested after bronchoalveolar lavage. All these samples were snap frozen in liquid nitrogen and kept at 80°C. Total RNA was extracted and purified using the Trizol reagent (Invitrogen) from epididymal adipose tissue, lung, hypothalamus, hippocampus, cortex, and olfactory bulb. The quality of RNA was assessed by determination of the ratio of absorbance at 260nm to absorbance at 280nm by nanodrop. Two micrograms of total DNase-treated RNA were reverse transcribed into cDNA using High Capacity cDNA Reverse Transcription Kits (Applied Biosystems™) per manufacturer’s instruction. Real-time PCR was performed using LightCycler® 480 SYBR Green I Master in the LightCycler® (Roche). Reactions were performed in a total volume of 10 μL containing 1 μL cDNA, 0.2 μM of each primer, and 5 μL of the SYBR® Green reaction mix. The amplification protocol was as follows: 95°C/5min (95°C/10 s, 60°C/20 s, and 72°C/30 s)×45. Following amplification, a dissociation curve analysis was performed to ensure purity of PCR product. The specific sense and antisense primers for tumor necrosis factor-α (TNFα), interleukin (IL)-6, IL1β, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (the reference housekeeping gene) were previously described (Chen et al. 2017). In each sample, 2ΔΔCt was calculated and used to represent its gene expression level.

Western Blotting

Standard techniques were performed with primary antibodies: mouse Anti-IKK2 (Millipore), mouse AntiβActin (Sigma-Aldrich), and mouse Anti-GAPDH (Cell Signaling Technology). Signals were detected by chemiluminescence and analyzed by densitometry.

Vascular Function Assay

After euthanasia, mouse thoracic aorta was quickly removed and cleaned in physiological salt solution (PSS) containing (mM): NaCl, 130; NaHCO3, 14.9; KCl, 4.7; KH2PO4, 1.18; MgSO47H2O, 1.18; CaCl22H2O, 1.56; EDTA, 0.026; and glucose, 5.5. The aorta was cut into 2-mm rings. The aortic rings were then mounted in a muscle bath containing PSS at 37°C and bubbled with 95% O2–5% CO2. Isometric force generation was recorded with a Multi Myograph System (Danish Myo Technology). A resting tension of 4mN was imposed on each ring, and the rings were allowed to equilibrate for 1 h. To test vascular function, the contractile responses of aortic rings to phenylephrine (PE) in the absence or presence of an NOS inhibitor ωnitroLarginine methyl ester (L-NAME; 100μM) were assessed in an accumulative manner. To analyze the endothelial function, aortic rings were precontracted with PE (1 μM), and acetylcholine (ACh) or sodium nitroprusside (SNP) was added in an accumulative manner.

Statistics

All data are expressed as means±standard errors of the mean(SEMs) unless noted otherwise. Statistical tests were performed using one-way or two-way analysis of variance with Bonferroni post tests or unpaired t-tests using GraphPad Prism (version 4.1.2; GraphPad Software). The significance level was set at p<0.05.

Results

Exposure Characterization

The average PM2.5 concentrations in the FA and CAP chambers were 3.6±1.9 and 71.6±33.2μg/m3, respectively. Table 1 shows the elemental composition of CAP and ambient PM2.5. Although identifying PM2.5 emission sources is generally beyond the scope of this paper, the relatively high ratio of Na/Al reflects more contribution by the marine source than the crustal source, which is consistent with the geographic proximity of the study site to the ocean (Huang et al. 1999), whereas the relatively high sulfur suggests that the study site was most strongly affected by secondary aerosols, which are likely to include emissions from coal-fired utility boilers located regionally (Huang et al. 1999).

Table 1. Ambient particulate matter (PM)2.5 and concentrated ambient PM2.5 (CAP) in the exposure chambers were collected to Teflon filters weekly, and their elemental compositions were determined by inductively coupled plasma mass spectroscopy (ICP-MS).
Element CAP Ambient
Mean (ng/m3) SD Mean (%) SD
Na 35.14 19.66 11.95 3.24
Mg 8.62 4.81 2.87 0.48
Al 25.93 12.67 2.56 2.41
Si 78.53 39.39 8.53 5.76
P 1.00 0.50 0.18 0.10
S 249.91 183.42 48.31 6.90
K 20.38 11.57 3.19 0.86
Ca 39.30 19.86 5.40 1.86
Ti 3.55 2.09 0.59 0.14
V 0.26 0.15 0.03 0.04
Cr 0.28 0.15 0.07 0.02
Mn 1.19 0.75 0.21 0.03
Fe 47.19 30.10 7.92 0.84
Ni 0.65 0.40 0.11 0.05
Cu 1.45 0.93 0.36 0.10
Zn 13.77 6.82 2.63 1.41
As 0.19 0.10 0.03 0.02
Se 0.13 0.06 0.01 0.02
Br 8.42 5.22 1.42 0.35
Sr 0.24 0.13 0.18 0.05
Ag 0.55 0.19 0.06 0.13
Sn 1.25 0.53 0.19 0.16
Ba 1.66 1.16 0.32 0.23
Ce 0.24 0.25 0.04 0.08
Pr 0.30 0.13 0.04 0.06
Er 0.98 0.75 0.20 0.06
Lu 1.33 1.19 0.32 0.12
W 0.41 0.26 0.12 0.08
Ir 0.11 0.04 0.01 0.02
Pt 0.20 0.09 0.02 0.03
Au 0.22 0.14 0.03 0.03
Hg 0.34 0.05 0.07 0.04
Tl 0.11 0.04 0.01 0.02
Pb 0.93 0.32 0.14 0.04

Note: CAP, concentrated ambient particulate matter (PM)2.5; SD, standard deviation.

Characterization of IKK2Neu-KO and Evaluation of Pro-Inflammatory Cytokines after Exposure to Filtered Air or Concentrated Ambient PM2.5

In order to examine the role of neural inflammation in the PM2.5 exposure–related pathophysiology, we generated NestincreIKK2flox/flox (IKK2NeuKO) and littermate NestincreIKK2flox/+ (control). Except for lower body weight at 12 wk of age (mean 22.3±1.1 vs. 20.3±0.6 g for controls and IKK2NeuKO, respectively, n=8/group, student’s t-test p<0.05), adult IKK2NeuKO mice were similar in appearance to adult control mice. Figure 1A shows the deletion of IKK2 in the brain of IKK2NeuKO. To confirm the tissue-specific deletion of IKK2 in IKK2NeuKO, we assessed IKK2 protein levels in different tissues. Figures 1B and C reveal that compared to control, IKK2NeuKO had significantly decreased expression of IKK2 in the brain but not lung, heart, and epididymal adipose tissue, corroborating the neuron-specific deletion of IKK2 in these mice. Consistent with our previous study (Ying et al. 2014), Figures 1D and E show that CAP exposure significantly increased the hypothalamic expression of TNFα and IL-6 mRNAs. Neural IKK2 knockout did not significantly alter the hypothalamic expression levels of TNFα and IL-6 mRNAs in FA-exposed mice, but significantly reduced them in CAP-exposed mice. Given that recent studies have suggested that some components of ambient PM2.5 may enter the brain through the olfactory system (Maher et al. 2016), we isolated olfactory bulb, cortex, and hippocampus, and assessed their expression of TNFα and IL-6 mRNAs. Except for a marked trend of increase in the expression of IL-6 mRNA in the hippocampus of CAP-exposed control mice vs. that of FA-exposed control mice, no any other remarkable difference was observed (Figures 1F–K).

Figure 1A is a PCR assay. Figure 1B is a Western blot showing the protein expression in brain, fat, lung, and heart tissues from male control mice and IKK2 super Neu KO mice. Figure 1C is a stacked bar graph with standard deviation plotting protein expression (percentage of WT) (y-axis) in brain, fat, lung, and heart tissues (x-axis) from male control mice and IKK2 super Neu KO mice. Figures 1D to 1K are stacked bar graphs with standard deviation plotting hypothalamic TNF alpha mRNA (AU), hypothalamic IL 6 mRNA (AU), hippocampus TNF alpha mRNA (AU), hippocampus IL 6 mRNA (AU), cortex TNF alpha mRNA (AU), cortex IL 6 mRNA (AU), olfactory TNF alpha mRNA (AU), and olfactory IL 6 mRNA (AU), respectively (y-axis), from male control mice and IKK2 super Neu KO mice as exposed to CAP and FA (x-axis).

Figure 1. Tissue-specific IKK2 expression and brain region–specific higher tumor necrosis factor-α (TNFα) and interleukin (IL)-6 mRNA expression. (A) Polymerase chain reaction (PCR) assay of flox allele (F), wild-type allele (W), and deletion allele (D) according to genotype in 13-wk-old male mice. Lane 1: IKK2flox/+; Lane 2, NestincreIKK2+/+; Lanes 3–5, NestincreIKK2flox/flox (IKK2NeuKO); Lanes 6–8, NestincreIKK2flox/+ (control); Lane 9, marker. (BC) Representative Western blot (B) and mean±standarddeviation(SD) IKK2 protein expression in brain, fat, lung, and heart tissues from male control (n=8) and IKK2NeuKO (n=7) mice at 10–16 wk of age. *p<0.05 vs. control, one-way analysis of variance (ANOVA). (DK) TNFα and IL-6 mRNA expression in hypothalamus, hippocampus, cortex, and olfactory bulb tissues harvested from control and IKK2NeuKO mice (n=67/group) at 8–12 wk of age following 4 mo of exposure to concentrated ambient particulate matter (PM)2.5 (CAP) or filtered air (FA). *p<0.05 vs. FA, #p<0.05 vs. control, ANOVA.

Evaluation of Pulmonary Inflammation after Exposure to Filtered Air or Concentrated Ambient PM2.5

Central inflammation has been implicated in the regulation of peripheral inflammatory responses (Liu et al. 2014), and pulmonary inflammation is widely believed to be essential in the pathophysiology due to exposure to ambient PM2.5 (Brook et al. 2010). To determine whether central inflammation plays a role in the inflammatory response of the lung to PM2.5 inhalation, we performed BALF cell differentiation. Figures 2A–D reveal that exposure to CAP significantly increased BALF total cells and macrophages in control mice. Consistent with many previous studies (Filep et al. 2016; Yoshizaki et al. 2017), although neutrophil infiltration is the hallmark of inflammation, we did not observe significant increase in BALF neutrophils in CAP-exposed mice (Figure 2D). This may be a reflection of the crucial role of neutrophils in acute but not chronic inflammatory response. BALF cell numbers were not significantly different between CAP- vs. FA-exposed neural IKK2–deficient mice, which suggests that CAP exposure–induced pulmonary inflammation may be regulated by a central mechanism. To confirm this, we performed lung histological analysis. The percentage of alveolar macrophages was significantly higher in CAP-exposed vs. FA-exposed IKK2NeuKO mice and control mice, but was significantly lower in CAP-exposed IKK2NeuKO mice vs. CAP-exposed controls (Figure 2E–F). Pulmonary TNFα and IL-6 expression was higher in CAP-exposed vs. FA mice, though the difference was significant only for controls (Figure 2G–I). In addition, pulmonary IL-6 mRNA expression was significantly lower in CAP-exposed IKK2NeuKO mice than in CAP-exposed controls (Figure 2H).

Figures 2A to 2D are stacked bar graphs with standard deviation plotting total cells in BALF (times 10 super 5), macrophages in BALF (times 10 super 5), lymphocytes in BALF (times 10 super 5), and neutrophils in BALF (times 10 super 5), respectively (y-axis), from male control mice and IKK2 super Neu KO mice as exposed to CAP and FA (x-axis). Figure 2E is a photomicrograph showing macrophages. Figures 2F to 2I are stacked bar graphs with standard deviation plotting percentage of macrophages, TNF alpha mRNA (fold), IL 6 mRNA (fold), and IL 1 beta mRNA (fold), respectively (y-axis), from male control mice and IKK2 super Neu KO mice as exposed to CAP and FA (x-axis).

Figure 2. Markers of pulmonary inflammation after 4 mo of exposure to concentrated ambient particulate matter (PM)2.5 (CAP) or filtered air (FA) in male NestincreIKK2flox/+ control and NestincreIKK2flox/flox (IKK2NeuKO) mice (n=67/group). (AD) Total cells, macrophages, lymphocyte, and neutrophil cell counts [means±standarddeviation(SD)] in bronchoalveolar lavage fluid (BALF). (EF) Representative histologic sections (red arrows indicate macrophages), and means±SD % alveolar macrophages, respectively, according to exposure and genotype (scalebar=50 μm). (GI) Tumor necrosis factor-α (TNFα), interleukin (IL)-6, and IL1β mRNA expression in pulmonary tissue samples according to exposure and genotype (means±SD). *p<0.05 vs. FA, #p<0.05 vs. control, analysis of variance (ANOVA).

Evaluation of Systemic and Adipose Inflammation after Exposure to Filtered Air or Concentrated Ambient PM2.5

Systemic inflammation is one of the putative mechanisms that mediate the development of extrapulmonary abnormalities due to exposure to ambient PM2.5 (Brook et al. 2010). In control mice, CAP vs. FA exposure was associated with higher plasma TNFα, IL1β, and IL-6 levels (significant for IL1β and IL-6 only) (Figure 3A–C). In CAP-exposed IKK2NeuKO mice, plasma TNFα, IL1β, and IL-6 were lower than in CAP-exposed controls (significant for TNFα and IL1β only). In addition, IL-6 was significantly higher in CAP- vs. FA-exposed IKK2NeuKO mice. (Figure 3C). In addition, exposure to ambient PM2.5 has been shown to induce marked inflammation in adipose tissues, and the latter is believed to be subsequent to systemic inflammation and play a role in the PM2.5 exposure–related abnormalities in energy and glucose metabolism (Brook et al. 2010). CAP vs. FA exposure was associated with significantly higher IL1β mRNA expression in epididymal adipose tissue from both IKK2NeuKO and control mice, but was not associated with significant differences in TNFα or IL-6 expression (Figure 4A–C). IL1β mRNA expression in adipose tissue was significantly lower in FA-exposed IKK2NeuKO mice vs. FA controls and in CAP-exposed IKK2NeuKO mice vs. CAP controls (Figure 4C). Macrophage infiltration in epididymal adipose tissue was significantly higher in CAP-exposed IKK2NeuKO and control mice relative to FA-exposed IKK2NeuKO and control mice, respectively, but was significantly lower in CAP-exposed IKK2NeuKO mice than in CAP-exposed controls (Figure 4D–E).

Figures 3A to 3C are stacked bar graphs with standard deviation plotting plasma TNF alpha in picograms per milliliter, plasma IL 1 beta in picograms per milliliter, and plasma IL 6 in picograms per milliliter, respectively (y-axis), from male control mice and IKK2 super Neu KO mice as exposed to CAP and FA (x-axis).

Figure 3. Plasma cytokine levels (BD™ Cytometric Bead Array Kit) following 4 wk of concentrated ambient particulate matter (PM)2.5 (CAP) or filtered air (FA) exposure in male NestincreIKK2flox/+ control and NestincreIKK2flox/flox (IKK2NeuKO) mice at 8–12 wk of age (n=67/group). (A) Tumor necrosis factor-α (TNFα), (B) interleukin (IL)1β, (C) IL-6. *p<0.05 vs. FA, #p<0.05 vs. control, analysis of variance (ANOVA).

Figures 4A to 4D are stacked bar graphs with standard deviation plotting percentage of macrophages, TNF alpha mRNA (fold), IL 6 mRNA (fold), IL 1 beta mRNA (fold), and macrophages (per 500 adipocytes), respectively (y-axis), from male control mice and IKK2 super Neu KO mice as exposed to CAP and FA (x-axis). Figure 4E shows four photomicrographs of male control mice and IKK2 super Neu KO mice as exposed to CAP and FA.

Figure 4. Cytokine mRNA expression in epididymal adipose tissue samples following 4 wk of filtered air (FA) or concentrated ambient particulate matter (PM)2.5 (CAP) exposure in male NestincreIKK2flox/+ control and NestincreIKK2flox/flox (IKK2NeuKO) mice at 8–12 wk of age (n=67/group). (AC) Fold difference in mRNA [quantitative real-time polymerase chain reaction (qPCR), relative to control gene] of tumor necrosis factor-α (TNFα), interleukin (IL)-6, and IL1β, respectively, according to exposure and genotype [mean±standarddeviation(SD)]. (D) Macrophage infiltration (number/500 adipocytes) according to exposure and genotype (mean±SD). (E) Representative histologic sections, scale bar, 25 μm. *p<0.05 vs. FA, #p<0.05 vs. control, analysis of variance (ANOVA).

Evaluation of Endothelial Dysfunction after Exposure to Filtered Air or Concentrated Ambient PM2.5

Endothelial dysfunction may contribute to PM2.5 exposure–related cardiometabolic abnormalities, such as insulin resistance and hypertension (Brook et al. 2010). CAP vs. FA exposure was associated with a significantly higher aortic contractile response to PE in both IKK2NeuKO and control mice, though the maximal PE response was significantly lower in CAP-exposed IKK2NeuKO mice than in CAP-exposed controls (Figure 5A and Table 2). As increased contractile response may result from increased contractility of vascular smooth muscle and/or decreased endothelium-dependent relaxation, we next assessed the aortic response to an endothelium-dependent vasodilator, Ach. CAP vs. FA exposure was associated with a significantly reduced Ach-induced aortic relaxation in both IKK2NeuKO and control mice, though the maximal effect was significantly lower in CAP-exposed IKK2NeuKO mice than in CAP-exposed controls (Figure 5B and Table 2). In contrast, we did not observe any significant effect of CAP exposure and neural IKK2 knockout on aortic contractile response to PE in the presence of a NOS inhibitor L-NAME (Figure 5C) and aortic relaxation by endothelium-independent vasodilator SNP (Figure 5D). Together, these findings support a role of central inflammation in the mediation of ambient PM2.5 exposure–induced endothelial dysfunction.

Figures 5A and 5B are line graphs with standard deviation respectively plotting force (percentage of 120 millimolar KCl) and force (percentage of precontraction) (y-axis) across log PE in molar and log ACh in molar (x-axis). Figures 5C and 5D are line graphs with standard deviation respectively plotting force (percentage of 120 millimolar KCl) in the presence of the NOS inhibitor L-NAME and force (percentage of precontraction) (y-axis) across log PE in molar and log SNP in molar (x-axis). All of these were associated with FA and CAP exposed male control mice and IKK2 super Neu KO mice.

Figure 5. Endothelial function following 4 wk of filtered air (FA) or concentrated ambient particulate matter (PM)2.5 (CAP) exposure in male NestincreIKK2flox/+ control and NestincreIKK2flox/flox (IKK2NeuKO) mice at 8–12 wk of age (n=67/group). After sacrifice, aorta was immediately isolated and mounted onto myograph. (A) Aortic contractile response to phenylephrine (PE). (B) Aortic rings precontracted by PE (1 μM) and then relaxed by acetylcholine (Ach). (C) Aortic contractile response to PE in the presence of the NOS inhibitor, L-NAME (100 μM). (D) Aortic rings precontracted by PE (1 μM) and then relaxed by the NO donor, sodium nitroprusside (SNP). Values are means±standarddeviation(SD). Numeric data for the maximum values are shown in Table 2.

Table 2. Maximal effects and logEC50s [means±standarddeviation(SD)] from myograph analyses of endothelial function following 4 mo of filtered air (FA) or concentrated ambient particulate matter (PM)2.5 (CAP) exposure in male NestincreIKK2flox/+ (control) and NestincreIKK2flox/flox (IKK2NeuKO) mice at 8–12 weeks of age (n=67/group).
Experiment FA/control CAP/control FA/IKK2NeuKO CAP/IKK2NeuKO
PE
 Maximal effect 53±3 81.8±4.2* 58.1±6.1 69±3.3*#
 logEC50 7.5±0.2 7.6±0.2 7.3±0.3 7.3±0.3
Ach
 Maximal effect 49.2±6.3 71.3±4.4* 45.5±8.1 57.5±4.8*#
 logEC50 7±0.3 7.2±0.4 6.9±0.3 7.3±0.4
L-NAME/PE
 Maximal effect 84.1±5.4 88.3±9 86.5±7.8 92.2±11.4
 logEC50 6.9±0.2 6.7±0.3 6.9±0.3 6.8±0.3
SNP
 Maximal effect 5.3±7.4 11.3±7.1 7.8±3.2 10.3±7.2
 logEC50 7±0.2 6.9±0.2 7.1±0.3 6.9±0.2

Note: Complete data are shown in Figure 5. Ach, acetylcholine; CAP, concentrated ambient particulate matter (PM)2.5; FA, filtered air; L-NAME, LNGNitroarginine methyl ester; PE, phenylephrine; SNP, sodium nitroprusside.

*p<0.05 vs. FA.

#p<0.05 vs. control, analysis of variance (ANOVA).

Discussion

Findings from experimental studies of mice suggest that exposure to ambient PM2.5 may induce central inflammation characterized by activation of the IKK2/NFκB signaling pathway in the hypothalamus (Liu et al. 2014; Ying et al. 2014). However, to our knowledge, the role of central inflammation in pathophysiologic responses to ambient PM2.5 exposure has not yet been systemically investigated. In the present study, we exploited a mouse model with reduced neuronal IKK2 expression to examine the role of central inflammation in PM2.5 exposure–induced peripheral inflammation and endothelial dysfunction. The main findings from the present study are that neural IKK2 reduction appears to inhibit CAP exposure–induced increases in a) hypothalamic TNFα and IL-6 mRNA expression; b) pulmonary, systemic, and adipose tissue inflammatory responses; and c) endothelial dysfunction. These findings provide further support for a role of central inflammation in the development of adverse cardiometabolic responses to ambient PM2.5 exposure.

Central inflammation, particularly in the hypothalamus, has been shown to play an important role in the pathogenesis of various cardiometabolic diseases such as obesity, hypertension, and diabetes (Han et al. 2016). We previously reported that exposure to ambient PM2.5 was associated with increased expression of pro-inflammatory cytokines and activation of IKK2 in the mouse hypothalamus (Ying et al. 2014). As in our previous study, we demonstrated that CAP exposure increased expression of pro-inflammatory cytokines in the hypothalamus. In addition, we demonstrated that these responses to CAP exposure were significantly inhibited or abolished in mice with reduced neural IKK2 expression. We found that, on average, IKK2 protein expression was only 31±12% lower in the brains of IKK2NeuKO mice compared with NestincreIKK2flox/+ controls, consistent with neuron-specific knockout by Nestin-cre (Tronche et al. 1999) and the presence of cells other than neurons in brain, including astrocytes and microglia that may play a role in local inflammatory responses (Douglass et al. 2017; Jassam et al. 2017). Nonetheless, the partial reduction in brain IKK2 expression appeared to be sufficient to reduce or eliminate CAP exposure–induced expression of TNFα and IL-6 mRNA in the hypothalamus. A recent study reported hypothalamic inflammation induced by a single-day exposure to an HFD was inhibited in vagotomized mice, suggesting that the vagal afferent nerve may be responsible for transmitting gut-derived inflammatory signals to the hypothalamus (Waise et al. 2015). Further studies are needed to determine whether a similar mechanism might be involved in CAP-induced hypothalamic inflammation.

In addition to systemic circulation, foreign materials may also gain access to the brain via the olfactory mucosa, known as olfactory bulb transmission (Tjälve et al. 1996). Given the previous demonstration of ambient particles in the brain (Maher et al. 2016) and increasing evidence that exposure to ambient PM2.5 correlates to various brain insults (Underwood 2017), olfactory bulb transmission has been suspected to be an important mechanism for the access of ambient PM2.5 to the brain (Underwood 2017). In the present study, CAP vs. FA exposure was associated with significantly higher expression of the pro-inflammatory cytokines TNFα and IL-6 in the hypothalamus, whereas in the olfactory bulb, TNFα expression was higher in CAP- vs. FA-exposed control mice, but not significantly, and there was a nonsignificant decrease in IL-6 expression. Additional research is needed to clarify the potential for olfactory bulb transmission of PM2.5 to the brain, and its influence, if any, on central inflammation.

Another finding from the present study was that increases in plasma TNFα, IL1β, and IL-6 in CAP- vs. FA-exposed control mice were reduced or eliminated in IKK2NeuKO mice. This suggests a potential role of the central nervous system in inflammatory responses to CAP, consistent with our previous finding that intracerebrovascular administration of an IKK2 inhibitor was associated with reduced systemic inflammatory responses to CAP exposure (Liu et al. 2014). Moreover, as hypothalamic inflammation has been shown to correlate to ANS dysfunction (Han et al. 2016), these present data are also consistent with increasing evidence that ANS regulates systemic inflammatory responses (Lambert et al. 2015; Pavlov and Tracey 2017). Our finding that CAP exposure–induced pulmonary inflammation was reduced in IKK2NeuKO mice compared with controls (Figure 2), consistent with involvement of a central neural mechanism in pathophysiologic responses to PM2.5 exposure, may seem inconsistent with compelling evidence that pulmonary inflammation in response to ambient PM2.5 is induced through toll-like receptors (TLRs)- and/or pattern recognition receptors (PRRs)-dependent mechanisms (Bekki et al. 2016; Fuertes et al. 2013; Shoenfelt et al. 2009; Zhao et al. 2012). However, previous studies have also reported evidence of close cross-talk between the brain and lung (Davison et al. 2012; Engel et al. 2015; Winklewski et al. 2014).

Exposure to ambient PM2.5 has been associated with endothelial dysfunction in humans (O’Neill et al. 2005; Wauters et al. 2013) and various animal models (Davel et al. 2012; Lei et al. 2005; Ying et al. 2013), and effects of PM2.5 on endothelial function have been implicated in the pathogenesis of cardiometabolic diseases associated with PM2.5 exposure, such as atherosclerosis, hypertension, and type 2 diabetes mellitus (Corban et al. 2017). As such, endothelial dysfunction is widely believed to be central in the pathophysiology due to exposure to PM2.5 (Brook et al. 2010). In the present study, although CAP exposure was associated with evidence of endothelial dysfunction in both control and IKK2NeuKO mice, the response was significantly less pronounced in the mice with reduced neural IKK2 expression compared with controls, consistent with a role of central inflammation in PM2.5 exposure–related effects. Many studies have demonstrated that systemic inflammation may lead to endothelial dysfunction (Piccardi et al. 2017).

Although the present study supports a role of neural IKK2 in pathophysiologic responses to ambient PM2.5 exposure, it has a number of important limitations. These include the fact that we did not evaluate whether specific types of neurons, or neurons in specific locations, are responsible for differences in responses to CAPs between control and IKK2NeuKO mice. This will require additional experiments targeting different neurons, particularly pro-opiomelanocortin and agouti-related peptide neurons in the hypothalamus and those in the nucleus ambiguus and dorsal nucleus of the brainstem, and concomitantly measuring effects of exposure to CAP. In addition, we did not investigate specific mechanisms through which neural IKK2 might mediate systemic inflammation or other effects of PM2.5 exposure, or directly assess neural mechanisms linking ambient PM2.5 with central inflammation.

Conclusion

In summary, our findings support a role of neural IKK2 in CAP exposure–induced local and systemic pro-inflammatory cytokine expression, pulmonary and adipose inflammation, and endothelial dysfunction, and thus advance knowledge our understanding of the contribution of central inflammation to the pathophysiologic effects of ambient PM2.5 exposure.

Acknowledgments

This work was supported by the National Institutes of Health (R01ES024516 to Z. Ying), the American Heart Association (13SDG17070131 to Z. Ying), and the National Natural Science Foundation of China (81770805 to Z. Ying, 81500216 to M. Chen, and 81302452 to L. Qiu).

References

Bekki K, Ito T, Yoshida Y, He C, Arashidani K, He M, et al. 2016. PM2.5 collected in China causes inflammatory and oxidative stress responses in macrophages through the multiple pathways. Environ Toxicol Pharmacol 45:362–369, PMID: 27393915, 10.1016/j.etap.2016.06.022.

Brook RD. 2005. You are what you breathe: evidence linking air pollution and blood pressure. Curr Hypertens Rep 7(6):427–434, PMID: 16386198, 10.1007/s11906-005-0037-9.

Brook RD. 2008. Cardiovascular effects of air pollution. Clin Sci 115(6):175–187, PMID: 18691154, 10.1042/CS20070444.

Brook RD, Rajagopalan S, Pope CA III, Brook JR, Bhatnagar A, Diez-Roux AV, et al. 2010. Particulate matter air pollution and cardiovascular disease: An update to the scientific statement from the American Heart Association. Circulation 121(21):2331–2378, PMID: 20458016, 10.1161/CIR.0b013e3181dbece1.

Chen LC, Hwang JS. 2005. Effects of subchronic exposures to concentrated ambient particles (CAPs) in mice. IV. Characterization of acute and chronic effects of ambient air fine particulate matter exposures on heart-rate variability. Inhal Toxicol 17(4-5):209–216, PMID: 15804938, 10.1080/08958370590912789.

Chen M, Liang S, Zhou H, Xu Y, Qin X, Hu Z, et al. 2017. Prenatal and postnatal mothering by diesel exhaust PM2.5-exposed dams differentially program mouse energy metabolism. Part Fibre Toxicol 14(1):3, PMID: 28100227, 10.1186/s12989-017-0183-7.

Corban MT, Duarte-Garcia A, McBane RD, Matteson EL, Lerman LO, Lerman A. 2017. Antiphospholipid syndrome: Role of vascular endothelial cells and implications for risk stratification and targeted therapeutics. J Am Coll Cardiol 69(18):2317–2330, PMID: 28473138, 10.1016/j.jacc.2017.02.058.

Davel AP, Lemos M, Pastro LM, Pedro SC, de André PA, Hebeda C, et al. 2012. Endothelial dysfunction in the pulmonary artery induced by concentrated fine particulate matter exposure is associated with local but not systemic inflammation. Toxicology 295(1-3):39–46, PMID: 22361244, 10.1016/j.tox.2012.02.004.

Davison DL, Terek M, Chawla LS. 2012. Neurogenic pulmonary edema. Crit Care 16(2):212, PMID: 22429697, 10.1186/cc11226.

DiBona GF. 2013. Sympathetic nervous system and hypertension. Hypertension 61(3):556–560, PMID: 23357181, 10.1161/HYPERTENSIONAHA.111.00633.

Douglass JD, Dorfman MD, Thaler JP. 2017. Glia: silent partners in energy homeostasis and obesity pathogenesis. Diabetologia 60(2):226–236, PMID: 27986987, 10.1007/s00125-016-4181-3.

Dunn AJ. 2000. Cytokine activation of the HPA axis. Ann N Y Acad Sci 917:608–617, PMID: 11268389, 10.1111/j.1749-6632.2000.tb05426.x.

Engel O, Akyuz L, da Costa Goncalves AC, Winek K, Dames C, Thielke M, et al. 2015. Cholinergic pathway suppresses pulmonary innate immunity facilitating pneumonia after stroke. Stroke 46(11):3232–3240, PMID: 26451017, 10.1161/STROKEAHA.115.008989.

Feng L, Yang X, Asweto CO, Wu J, Zhang Y, Hu H, et al. 2017. Genome-wide transcriptional analysis of cardiovascular-related genes and pathways induced by PM2.5 in human myocardial cells. Environ Sci Pollut Res Int 24(12):11683–11693, PMID: 28326484, 10.1007/s11356-017-8773-3.

Filep Á, Fodor GH, Kun-Szabó F, Tiszlavicz L, Razga Z, Bozso G, et al. 2016. Exposure to urban PM1 in rats: development of bronchial inflammation and airway hyperresponsiveness. Respir Res 17:26, PMID: 26966003, 10.1186/s12931-016-0332-9.

Froeschl M, Hadziomerovic A, Ruzicka M. 2013. Renal sympathetic denervation for resistant hypertension. Can J Cardiol 29(5):636–638, PMID: 23541665, 10.1016/j.cjca.2013.02.019.

Fuertes E, Brauer M, MacIntyre E, Bauer M, Bellander T, von Berg A, et al. 2013. Childhood allergic rhinitis, traffic-related air pollution, and variability in the GSTP1, TNF, TLR2, and TLR4 genes: results from the TAG study. J Allergy Clin Immunol 132(2):342–352.e2, PMID: 23639307, 10.1016/j.jaci.2013.03.007.

Geller MD, Biswas S, Fine PM, Sioutas C. 2005. A new compact aerosol concentrator for use in conjunction with low flow-rate continuous aerosol instrumentation. J Aerosol Sci 36(8):1006–1022, 10.1016/j.jaerosci.2004.11.015.

Godleski JJ, Verrier RL, Koutrakis P, Catalano P, Coull B, Reinisch U, et al. 2000. Mechanisms of morbidity and mortality from exposure to ambient air particles. Res Rep Health Eff Inst 5–88; discussion 89–103, PMID: 10817681.

Gong H Jr., Linn WS, Clark KW, Anderson KR, Sioutas C, Alexis NE, et al. 2008. Exposures of healthy and asthmatic volunteers to concentrated ambient ultrafine particles in Los Angeles. Inhal Toxicol 20(6):533–545, PMID: 18444007, 10.1080/08958370801911340.

Han C, Rice MW, Cai D. 2016. Neuroinflammatory and autonomic mechanisms in diabetes and hypertension. Am J Physiol Endocrinol Metab 311(1):E32–E41, PMID: 27166279, 10.1152/ajpendo.00012.2016.

Huang SL, Rahn KA, Arimoto R. 1999. Testing and optimizing two factor-analysis techniques on aerosol at Narragansett, Rhode Island. Atmos Environ 33:2169–2185, 10.1016/S1352-2310(98)00324-0.

Jassam YN, Izzy S, Whalen M, McGavern DB, El Khoury J. 2017. Neuroimmunology of traumatic brain injury: time for a paradigm shift. Neuron 95(6):1246–1265, PMID: 28910616, 10.1016/j.neuron.2017.07.010.

Jeong SC, Cho Y, Song MK, Lee E, Ryu JC. 2017. Epidermal growth factor receptor (EGFR)-MAPK-nuclear factor(NF)-κB-IL8: a possible mechanism of particulate matter(PM) 2.5-induced lung toxicity. Environ Toxicol 32(5):1628–1636, PMID: 28101945, 10.1002/tox.22390.

Kampfrath T, Maiseyeu A, Ying Z, Shah Z, Deiuliis JA, Xu X, et al. 2011. Chronic fine particulate matter exposure induces systemic vascular dysfunction via NADPH oxidase and TLR4 pathways. Circ Res 108(6):716–726, PMID: 21273555, 10.1161/CIRCRESAHA.110.237560.

Karemaker JM. 1999. Autonomic integration: the physiological basis of cardiovascular variability. J Physiol 517 (Pt 2):316, PMID: 10332083, 10.1111/j.1469-7793.1999.0316t.x.

Kennedy T, Ghio AJ, Reed W, Samet J, Zagorski J, Quay J, et al. 1998. Copper-dependent inflammation and nuclear factor-κB activation by particulate air pollution. Am J Respir Cell Mol Biol 19(3):366–378, PMID: 9730864, 10.1165/ajrcmb.19.3.3042.

Kreyling WG, Semmler-Behnke M, Seitz J, Scymczak W, Wenk A, Mayer P, et al. 2009. Size dependence of the translocation of inhaled iridium and carbon nanoparticle aggregates from the lung of rats to the blood and secondary target organs. Inhal Toxicol 21(suppl 1):55–60, PMID: 19558234, 10.1080/08958370902942517.

Lambert EA, Straznicky NE, Dixon JB, Lambert GW. 2015. Should the sympathetic nervous system be a target to improve cardiometabolic risk in obesity?. Am J Physiol Heart Circ Physiol 309(2):H244–H258, PMID: 25980020, 10.1152/ajpheart.00096.2015.

Lei YC, Hwang JS, Chan CC, Lee CT, Cheng TJ. 2005. Enhanced oxidative stress and endothelial dysfunction in streptozotocin-diabetic rats exposed to fine particles. Environ Res 99(3):335–343, PMID: 16307975, 10.1016/j.envres.2005.03.011.

Li ZW, Omori SA, Labuda T, Karin M, Rickert RC. 2003. IKK beta is required for peripheral B cell survival and proliferation. J Immunol 170(9):4630–4637, PMID: 12707341, 10.4049/jimmunol.170.9.4630.

Liu C, Fonken LK, Wang A, Maiseyeu A, Bai Y, Wang TY, et al. 2014. Central IKKβ inhibition prevents air pollution mediated peripheral inflammation and exaggeration of type II diabetes. Part Fibre Toxicol 11:53, PMID: 25358444, 10.1186/s12989-014-0053-5.

Maciejczyk P, Zhong M, Li Q, Xiong J, Nadziejko C, Chen LC. 2005. Effects of subchronic exposures to concentrated ambient particles (CAPs) in mice. Ii. The design of a caps exposure system for biometric telemetry monitoring. Inhal Toxicol 17(4-5):189–197, PMID: 15804936, 10.1080/08958370590912743.

Magari SR, Hauser R, Schwartz J, Williams PL, Smith TJ, Christiani DC. 2001. Association of heart rate variability with occupational and environmental exposure to particulate air pollution. Circulation 104(9):986–991, PMID: 11524390, 10.1161/hc3401.095038.

Maher BA, Ahmed IA, Karloukovski V, MacLaren DA, Foulds PG, Allsop D, et al. 2016. Magnetite pollution nanoparticles in the human brain. Proc Natl Acad Sci USA 113(39):10797–10801, PMID: 27601646, 10.1073/pnas.1605941113.

Mirowsky J, Hickey C, Horton L, Blaustein M, Galdanes K, Peltier RE, et al. 2013. The effect of particle size, location and season on the toxicity of urban and rural particulate matter. Inhal Toxicol 25(13):747–757, PMID: 24255952, 10.3109/08958378.2013.846443.

Oberdörster G, Sharp Z, Atudorei V, Elder A, Gelein R, Lunts A, et al. 2002. Extrapulmonary translocation of ultrafine carbon particles following whole-body inhalation exposure of rats. J Toxicol Environ Health A 65(20):1531–1543, PMID: 12396867, 10.1080/00984100290071658.

Oberdörster G, Utell MJ. 2002. Ultrafine particles in the urban air: to the respiratory tract–and beyond? Environ Health Perspect 110(8):A440–A441, PMID: 12153769, 10.1289/ehp.110-a440.

O’Neill MS, Veves A, Zanobetti A, Sarnat JA, Gold DR, Economides PA, et al. 2005. Diabetes enhances vulnerability to particulate air pollution-associated impairment in vascular reactivity and endothelial function. Circulation 111:2913–2920, PMID: 15927967, 10.1161/CIRCULATIONAHA.104.517110.

Pahl HL. 1999. Activators and target genes of Rel/NF-κB transcription factors. Oncogene 18(49):6853–6866, PMID: 10602461, 10.1038/sj.onc.1203239.

Pavlov VA, Tracey KJ. 2017. Neural regulation of immunity: molecular mechanisms and clinical translation. Nat Neurosci 20(2):156–166, PMID: 28092663, 10.1038/nn.4477.

Peters A, Frohlich M, Döring A, Immervoll T, Wichmann HE, Hutchinson WL, et al. 2001. Particulate air pollution is associated with an acute phase response in men; results from the MONICA-Augsburg study. Eur Heart J 22(14):1198–1204, PMID: 11440492, 10.1053/euhj.2000.2483.

Piccardi B, Giralt D, Bustamante A, Llombart V, Garcia-Berrocoso T, Inzitari D, et al. 2017. Blood markers of inflammation and endothelial dysfunction in cardioembolic stroke: Systematic review and meta-analysis. Biomarkers 22(3-4):200–209, PMID: 28117601, 10.1080/1354750X.2017.1286689.

Pope CA III. 2000. Epidemiology of fine particulate air pollution and human health: biologic mechanisms and who’s at risk? Environ Health Perspect 108(suppl 4):713–723, PMID: 10931790, 10.2307/3454408.

Purkayastha S, Zhang G, Cai D. 2011. Uncoupling the mechanisms of obesity and hypertension by targeting hypothalamic IKK-β and NF-κB. Nat Med 17(7):883–887, PMID: 21642978, 10.1038/nm.2372.

Quay JL, Reed W, Samet J, Devlin RB. 1998. Air pollution particles induce Il-6 gene expression in human airway epithelial cells via NF-κB activation. Am J Respir Cell Mol Biol 19(1):98–106, PMID: 9651185, 10.1165/ajrcmb.19.1.3132.

Semmler-Behnke M, Kreyling WG, Lipka J, Fertsch S, Wenk A, Takenaka S, et al. 2008. Biodistribution of 1.4- and 18-nm gold particles in rats. Small 4(12):2108–2111, PMID: 19031432, 10.1002/smll.200800922.

Shoenfelt J, Mitkus RJ, Zeisler R, Spatz RO, Powell J, Fenton MJ, et al. 2009. Involvement of TLR2 and TLR4 in inflammatory immune responses induced by fine and coarse ambient air particulate matter. J Leukoc Biol 86(2):303–312, PMID: 19406832, 10.1189/jlb.1008587.

Simonds SE, Cowley MA. 2013. Hypertension in obesity: is leptin the culprit? Trends Neurosci 36(2):121–132, PMID: 23333346, 10.1016/j.tins.2013.01.004.

Sivagangabalan G, Spears D, Masse S, Urch B, Brook RD, Silverman F, et al. 2011. The effect of air pollution on spatial dispersion of myocardial repolarization in healthy human volunteers. J Am Coll Cardiol 57(2):198–206, PMID: 21211691, 10.1016/j.jacc.2010.08.625.

Song L, Li D, Li X, Ma L, Bai X, Wen Z, et al. 2017. Exposure to PM2.5 induces aberrant activation of NF-κB in human airway epithelial cells by downregulating miR-331 expression. Environ Toxicol Pharmacol 50:192–199, PMID: 28192748, 10.1016/j.etap.2017.02.011.

Tankersley CG, Campen M, Bierman A, Flanders SE, Broman KW, Rabold R. 2004. Particle effects on heart-rate regulation in senescent mice. Inhal Toxicol 16(6-7):381–390, PMID: 15204754, 10.1080/08958370490439551.

Thaler JP, Yi CX, Schur EA, Guyenet SJ, Hwang BH, Dietrich MO, et al. 2012. Obesity is associated with hypothalamic injury in rodents and humans. J Clin Invest 122(1):153–162, PMID: 22201683, 10.1172/JCI59660.

Tjälve H, Henriksson J, Tallkvist J, Larsson BS, Lindquist NG. 1996. Uptake of manganese and cadmium from the nasal mucosa into the central nervous system via olfactory pathways in rats. Pharmacol Toxicol 79(6):347–356, PMID: 9000264, 10.1111/j.1600-0773.1996.tb00021.x.

Tronche F, Kellendonk C, Kretz O, Gass P, Anlag K, Orban PC, et al. 1999. Disruption of the glucocorticoid receptor gene in the nervous system results in reduced anxiety. Nat Genet 23(1):99–103, PMID: 10471508, 10.1038/12703.

Turnbull AV, Rivier CL. 1999. Regulation of the hypothalamic-pituitary-adrenal axis by cytokines: actions and mechanisms of action. Physiol Rev 79(1):1–71, PMID: 9922367, 10.1152/physrev.1999.79.1.1.

U.S. EPA (Environmental Protection Agency). 2009. Integrated Science Assessment (ISA) for Particulate Matter. Final Report. https://cfpub.epa.gov/ncea/isa/recordisplay.cfm?deid=216546 [accessed 5 May 2017].

Underwood E. 2017. The polluted brain. Science 355(6323):342–345, PMID: 28126768, 10.1126/science.355.6323.342.

Utell MJ, Frampton MW, Zareba W, Devlin RB, Cascio WE. 2002. Cardiovascular effects associated with air pollution: potential mechanisms and methods of testing. Inhal Toxicol 14(12):1231–1247, PMID: 12454788, 10.1080/08958370290084881.

van Eeden SF, Tan WC, Suwa T, Mukae H, Terashima T, Fujii T, et al. 2001. Cytokines involved in the systemic inflammatory response induced by exposure to particulate matter air pollutants (PM10). Am J Respir Crit Care Med 164(5):826–830, PMID: 11549540, 10.1164/ajrccm.164.5.2010160.

Vogel CF, Sciullo E, Wong P, Kuzmicky P, Kado N, Matsumura F. 2005. Induction of proinflammatory cytokines and C-reactive protein in human macrophage cell line U937 exposed to air pollution particulates. Environ Health Perspect 113(11):1536–1541, PMID: 16263508, 10.1289/ehp.8094.

Waise TM, Toshinai K, Naznin F, NamKoong C, Md Moin AS, Sakoda H, et al. 2015. One-day high-fat diet induces inflammation in the nodose ganglion and hypothalamus of mice. Biochem Biophys Res Commun 464(4):1157–1162, PMID: 26208455, 10.1016/j.bbrc.2015.07.097.

Wang X, Chen M, Zhong M, Hu Z, Qiu L, Rajagopalan S, et al. 2017. Exposure to concentrated ambient PM2.5 shortens lifespan and induces inflammation-associated signaling and oxidative stress in drosophila. Toxicol Sci 156(1):199–207, PMID: 28069988, 10.1093/toxsci/kfw240.

Wauters A, Dreyfuss C, Pochet S, Hendrick P, Berkenboom G, van de Borne P, et al. 2013. Acute exposure to diesel exhaust impairs nitric oxide-mediated endothelial vasomotor function by increasing endothelial oxidative stress. Hypertension 62(2):352–358, PMID: 23798345, 10.1161/HYPERTENSIONAHA.111.00991.

Winklewski PJ, Radkowski M, Demkow U. 2014. Cross-talk between the inflammatory response, sympathetic activation and pulmonary infection in the ischemic stroke. J Neuroinflammation 11:213, PMID: 25539803, 10.1186/s12974-014-0213-4.

Ying Z, Xu X, Bai Y, Zhong J, Chen M, Liang Y, et al. 2014. Long-term exposure to concentrated ambient PM2.5 increases mouse blood pressure through abnormal activation of the sympathetic nervous system: a role for hypothalamic inflammation. Environ Health Perspect 122(1):79–86, PMID: 24240275, 10.1289/ehp.1307151.

Ying Z, Xu X, Chen M, Liu D, Zhong M, Chen LC, et al. 2013. A synergistic vascular effect of airborne particulate matter and nickel in a mouse model. Toxicol Sci 135(1):72–80, PMID: 23788629, 10.1093/toxsci/kft136.

Yoshizaki K, Brito JM, Silva LF, Lino-Dos-Santos-Franco A, Frias DP, E Silva ES, et al. 2017. The effects of particulate matter on inflammation of respiratory system: Differences between male and female. Sci Total Environ 586:284–295, PMID: 28174048, 10.1016/j.scitotenv.2017.01.221.

Zhang Y, Zheng L, Tuo J, Liu Q, Zhang X, Xu Z, et al. 2017. Analysis of PM2.5-induced cytotoxicity in human HaCaT cells based on a microfluidic system. Toxicol In Vitro 43:1–8, PMID: 28431925, 10.1016/j.tiv.2017.04.018.

Zhang ZH, Wei SG, Francis J, Felder RB. 2003. Cardiovascular and renal sympathetic activation by blood-borne TNF-α in rat: the role of central prostaglandins. Am J Physiol Regul Integr Comp Physiol 284(4):R916–R927, PMID: 12626358, 10.1152/ajpregu.00406.2002.

Zhao C, Liao J, Chu W, Wang S, Yang T, Tao Y, et al. 2012. Involvement of TLR2 and TLR4 and TH1/TH2 shift in inflammatory responses induced by fine ambient particulate matter in mice. Inhal Toxicol 24(13):918–927, PMID: 23121301, 10.3109/08958378.2012.731093.

Community-Led Total Sanitation: A Mixed-Methods Systematic Review of Evidence and Its Quality

Author Affiliations open

1The Water Institute at University of North Carolina, Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA

PDF icon PDF Version (774 KB)

  • Background:
    Community-led total sanitation (CLTS) is a widely applied rural behavior change approach for ending open defecation. However, evidence of its impact is unclear.
    Objectives:
    We conducted a systematic review of journal-published and gray literature to a) assess evidence quality, b) summarize CLTS impacts, and c) identify factors affecting implementation and effectiveness.
    Methods:
    Eligible studies were systematically screened and selected for analysis from searches of seven databases and 16 websites. We developed a framework to appraise literature quality. We qualitatively analyzed factors enabling or constraining CLTS, and summarized results from quantitative evaluations.
    Discussion:
    We included 200 studies (14 quantitative evaluations, 29 qualitative studies, and 157 case studies). Journal-published literature was generally of higher quality than gray literature. Fourteen quantitative evaluations reported decreases in open defecation, but did not corroborate the widespread claims of open defecation–free (ODF) villages found in case studies. Over one-fourth of the literature overstated conclusions, attributing outcomes and impacts to interventions without an appropriate study design. We identified 43 implementation- and community-related factors reportedly affecting CLTS. This analysis revealed the importance of adaptability, structured posttriggering activities, appropriate community selection, and further research on combining and sequencing CLTS with other interventions.
    Conclusions:
    The evidence base on CLTS effectiveness available to practitioners, policy makers, and program managers to inform their actions is weak. Our results highlight the need for more rigorous research on CLTS impacts as well as applied research initiatives that bring researchers and practitioners together to address implementation challenges to improve rural sanitation efforts. https://doi.org/10.1289/EHP1965
  • Received: 30 March 2017
    Revised: 21 November 2017
    Accepted: 24 November 2017
    Published: 2 February 2018

    Address correspondence to V. Venkataramanan, 806 Sherman Ave., Apt. 2., Evanston IL 60202 USA. Telephone: (513) 652-1521. Email: vidyav@alumni.unc.edu

    Supplemental Material is available online (https://doi.org/10.1289/EHP1965).

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (296 KB)
    Zip icon Supplemental Code and Data Zip File (91 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (64 KB)

Introduction

An estimated 946 million people in the world practiced open defecation in 2015, 90% of whom lived in rural areas (UNICEF and WHO 2015). Open defecation adversely affects human health, contributing to diarrheal diseases and childhood stunting (Clasen et al. 2014; Spears et al. 2013; Vyas et al. 2016). Poor sanitation also has an adverse economic impact (DeFrancis 2011), and disproportionately affects the safety, health, and dignity of women (Hulland et al. 2015; Jadhav et al. 2016; Khanna and Das 2016; Kulkarni and O’Reilly 2014).

For decades, governments and nongovernmental organizations (NGOs) provided free or subsidized latrines to households, but practitioners widely believe that this approach was unable to guarantee regular latrine use. This recognition led to a focus on hygiene and health education programs, often combined with latrine subsidies, such as the Participatory Hygiene and Sanitation Transformation approach (WHO 1997). Lessons learned from implementing these programs led many sanitation professionals to conclude that while the infrastructure-heavy approach may have increased access to latrines and educational approaches may have increased awareness of health benefits, these strategies were largely insufficient to generate demand for latrines and change sanitation behavior (Jenkins and Sugden 2006).

As a response, the community-led total sanitation (CLTS) approach was developed, aiming to create open defecation–free (ODF) communities (Kar and Chambers 2008). This approach signified a fundamental shift from a focus on individual or household sanitation to a community-level concern for open defecation. CLTS facilitators attempt to trigger collective behavior change by encouraging and motivating people to confront the impact of community-wide open defecation. CLTS comprises three stages:

  • Pretriggering: selecting communities, training facilitators, collecting baseline information, and coordinating community entry.
  • Triggering: organizing a community-wide meeting where facilitators conduct participatory exercises intended to trigger shame and disgust. Attendees are expected to be motivated to change their sanitation situation.
  • Posttriggering: conducting routine follow-up visits, with the goal of verifying and certifying ODF status in communities.

Since the first pilot projects in Bangladesh in 2000, CLTS has been adopted by many international NGOs involved in rural sanitation and has been incorporated into national policy by many governments. It is arguably now the predominant rural sanitation behavior change approach.

Most literature on CLTS is contained on websites and knowledge bases in the form of gray literature, primarily produced by practitioners to share insights from their implementation experiences. It has often been noted that there is limited rigorous evidence on CLTS impacts. Governments and organizations implementing CLTS face the challenge of navigating a vast and cluttered body of literature to inform their decisions.

We identified 41 published systematic reviews relating to sanitation interventions. Most study the impact of sanitation on health outcomes. A handful look at behavior- or demand-related topics, such as factors affecting sustained adoption of water and sanitation technologies (Hulland et al. 2015), behavioral research relating to point-of-use-water treatment technologies (Fiebelkorn et al. 2012), behavioral models for water and sanitation (Dreibelbis et al. 2013; Dwipayanti et al. 2017), and water, sanitation, and hygiene (WaSH) social marketing approaches (Evans et al. 2014). No reviews were found that concentrate on CLTS or related interventions.

To address this gap, we conducted a mixed-methods systematic review of journal-published and gray literature on CLTS to characterize the state of the evidence. For the purpose of this review, journal-published refers only to peer-reviewed journals, with literature in other journals being classified as gray literature. The main objectives of the review were to: a) assess the quality of evidence, b) summarize evidence on the effectiveness and impact of CLTS on sanitation and health outcomes, and c) identify factors affecting CLTS implementation and effectiveness. We aimed to comprehensively document current understanding on CLTS from a variety of sources, including researchers, practitioners, and donors.

Methods

Search Strategy

We searched seven online peer-reviewed databases (Cochrane Library, Embase, Global Health, Web of Science, PubMed, Proquest, and Scopus) and the websites of 15 international organizations or sanitation knowledge hubs that document literature on CLTS. When expanding the scope of our search using Google Scholar, we observed considerable overlap between these search results and results from the 15 other websites already included in our search strategy. A previous study recommended focusing on the first 200 to 300 results from Google Scholar to find the most relevant results for systematic reviews of gray literature (Haddaway et al. 2015); therefore, we used this rationale to search the first 200 results from Google Scholar. A variety of search terms were employed to comprehensively search journal-published and gray literature, including combined keywords related to CLTS, open defecation, and demand-led and participatory approaches (see Table S1 for the full search strategy and list of databases). Documents were also reviewed from bibliographic hand searches and expert consultations. Searches were first conducted in December 2015 and updated in March 2017.

Document Selection and Eligibility Criteria

A multistep screening process was conducted for both journal-published and gray literature (Figure 1). Titles and abstracts of search results were screened independently by two authors (V.V. and A.K.), and they discussed any discrepancies in the selection of documents to reach a consensus on the final list before full-text review. Research studies, conference proceedings, evaluations, dissertations, reports, working papers, and organizational learning notes published between January 2000 (the year that CLTS was first piloted) and March 2017 were included. Guidelines, manuals, publicity material, news stories, slide presentations, workshop minutes, blog posts, reviews, and commentaries were excluded. In the case of multiple documents reporting data and findings from the same intervention or study, only the most recent document was included. Because the aim was to assess the quality of literature, no documents were excluded based on quality.

Flowchart showing the screening process.

Figure 1. Flow diagram of screening and selection process of community-led total sanitation literature. Note: In the first stage of screening, 4,187 records were excluded because the titles or abstracts indicated that they were guidelines, manuals, news stories, slide presentations, workshop minutes, blog posts, reviews, or commentaries.

Interventions labeled as CLTS are likely to contain a variety of adaptations, and several total sanitation strategies share characteristics with CLTS [e.g., School-led Total Sanitation, Community Approaches to Total Sanitation (CATS), Community-based Total Sanitation, and India’s variations on total sanitation campaigns (TSC)]. To be inclusive, interventions that met all of the following criteria were reviewed:

  • Mentioned sanitation behavior change as a key component of implementation.
  • Aimed to reduce or end open defecation.
  • Mobilized entire communities for sanitation rather than targeting households or specific populations.
  • Included participatory activities such as triggering, village mapping, or transect walks in the decision-making and data-gathering process.

Although the Handbook on CLTS describes it as a no-subsidy approach (Kar and Chambers 2008), there is considerable debate about the role of latrine subsidies as part of, or following CLTS activities (Papafilippou et al. 2011). Therefore, interventions that met the above criteria and provided subsidized latrine hardware were also considered for inclusion in this review.

We classified literature as quantitative evaluations, qualitative studies, and case studies and project reports, adapting an approach used in a systematic review of cook stoves (Rehfuess et al. 2014). For this review, quantitative evaluations were defined as studies designed to attribute outcomes to a CLTS or CLTS-like intervention. Studies had to include primary data collection of outcomes and an experimental comparison group (controlled trials, quasi-experimental designs, and before–after comparisons). Quantitative studies that did not meet these criteria or did not have a comparison group were classified as case studies. Qualitative studies were those that used qualitative data collection methods and analytical techniques. Case studies and project reports included mixed-methods studies, cross-sectional studies, and literature that described practitioner experiences or reports and evaluations of specific CLTS projects. Papers that shared general lessons or reflections without references to primary data were classified as commentaries and were excluded from the review.

Quality Appraisal

To characterize the quality of evidence on CLTS across journal-published and gray literature, we developed a quality appraisal framework for each of the three study types by reviewing and adapting questions from previously used protocols (Jack et al. 2010; Harden 2010; Heale and Twycross 2015; Loevinsohn 1990; Pluye et al. 2011; Puzzolo et al. 2013; Spencer et al. 2003; Thomas et al. 2004). Our framework comprises three categories: quality of reporting, minimizing risk of bias, and appropriateness of conclusions. Table 1 presents our quality appraisal framework, with questions to assess each criterion for the three study types. Most criteria in the quality appraisal could receive a score of 0, 0.5, or 1; the remaining, such as whether there was evidence of external peer review, could only receive a score of zero or one (see Table S2 for detailed scoring guidelines). Similar questions were asked about quality of reporting (with a maximum possible score of 6) and appropriateness of conclusions (with a maximum possible score of 3) across all three study types. However, because of differences in study design and intent, questions to assess the risk of bias differed by study type; in this category, case studies and projects reports could receive a maximum possible score of 4, qualitative studies a maximum possible score of 5, and quantitative evaluations a maximum possible score of 7 (see Table S2). Scores within each of the three categories were converted into percentages. Documents with the maximum possible score for each category were assigned a value of 100% for that category. The percentage scores for all documents within a particular subgroup (e.g., the 10 quantitative evaluations from gray literature) were then averaged to derive the mean percentage to assess differences by type of study and type of literature (journal-published vs. gray literature). An aggregate quality score was not computed for each document, as this would not allow for a nuanced discussion of quality by category, and could lead to misinterpretation of scoring. All documents were scored by the first author (V.V.), and 20% of the documents were subjected to independent quality control by the second or third authors (J.C. and A.K.).

Table 1. Quality appraisal framework for literature.
Category and criteria Questions to guide quality appraisala,b,c
Quality of reporting
  Objectives For all study types: Were the objectives and purpose of the study, program, or intervention described?
  Context For all study types: Was sufficient detail provided on the context and setting of the study, program, or intervention?
  Process For all study types: Was the process of the program or intervention described thoroughly?
  Study design For quantitative evaluations: Was sufficient detail provided on how households or individuals were assigned to interventions?
For qualitative studies: Was sufficient detail provided on the sampling approach?
For case studies/project reports: Is there evidence of a sampling approach?
  Data collection For quantitative evaluations and qualitative studies: Was sufficient detail provided on data collection methods and procedures?
For case studies/project reports: Is there evidence of a systematic data collection process?
  Analysis For quantitative evaluations and qualitative studies: Was sufficient detail provided on analytical methods used in the study?
For case studies/project reports: Is there description of data analysis?
Minimizing risk of bias
  Assignment to intervention For quantitative evaluations: Indicate the study design (randomized controlled trial, quasi-experimental design, natural experiment, or pre+post in single group).
For qualitative studies and case studies/project reports: Not scored.
  Appropriateness of sampling For quantitative evaluations: Was sampling representative at the household level (did the survey represent the study population?)
For qualitative studies and case studies/project reports: Was sampling appropriate given stated objectives?
  Independence of data collection For all study types: Was data collection conducted by an independent and trained source?
For quantitative evaluations: Auditing procedures are also considered.
  Rigor in data collection For quantitative evaluations: Were the indicators measured in the study relevant to the research question, and were they consistent with prior work and/or thoroughly justified? Was validity of data collection tools (testing/piloting) reported?
For qualitative studies: Were there attempts to establish the credibility, neutrality, consistency, and/or transferability of data collection tools?
For case studies/project reports: Were appropriate measures taken to provide rigor to the execution of the study (includes data collection and analysis)?
  Analytical rigor For quantitative evaluations: Were appropriate analytical methods used (RCTs and non-RCTs)?
For qualitative studies: Were there attempts to establish the credibility, neutrality, consistency, and/or transferability of data analysis methods?
For case studies/project reports: Scored alongside data collection (see previous question).
  External peer-review For all study types: Is there evidence of the document being subjected to external/independent review?
Appropriateness of conclusions
  Interpretation For all study types: Is there a discussion and interpretation of the main findings?
  Limitations For all study types: Were study limitations described?
  Conclusions For all study types: Were stated conclusions and implications within the scope of the study design and data collection methods?

Note: RCT, randomized controlled trial.

aWe developed this framework after reviewing and adapting questions from several previously used protocols (Jack et al. 2010; Harden 2010; Heale and Twycross 2015; Loevinsohn 1990; Pluye et al. 2011; Puzzolo et al. 2013; Spencer et al. 2003; Thomas et al. 2004).

bSee Table S2 for quality appraisal tools by study design and detailed scoring guidelines.

cDefinitions: Quantitative evaluations were defined as studies with an experimental comparison group designed to attribute outcomes to a community-led total sanitation (CLTS) or CLTS-like intervention; qualitative studies were defined as those that used qualitative data collection methods and analytical techniques; and case studies and project reports included mixed-methods studies, cross-sectional studies, and practitioner experiences, reports, or evaluations of CLTS projects.

Data Extraction and Analysis

Descriptive data on study type, author, project year, study design, countries of focus, country of publication, and methods were entered into a Microsoft Excel database. The main outcomes from quantitative evaluations were extracted and summarized. Qualitative content analysis was conducted for all included literature, regardless of study type, using Atlas.ti (version 7.0; Atlas.ti Scientific Softare Development). Documents were coded in two cycles (each cycle by one author) for the following: enablers and barriers to successful implementation in different stages of CLTS, key themes discussed, and indicators of success measured by programs and researchers. The first cycle of coding identified 150 factors reported as enabling or constraining CLTS. By combining similar factors, we narrowed this list to 43 factors in the second cycle. A similar process was conducted for indicators of success. Based on this inductive analysis, factors that enabled or constrained CLTS were grouped under three implementation-related domains (policy environment, implementation quality, and administrative context) and four community-related domains (environment, capacity, participation patterns, and behavior). Indicators of success were grouped under the following domains: WaSH outcomes, CLTS process, behavioral outcomes, extended impact, and motivators for behavior change.

Results

We identified 5,884 documents from databases, websites, Google Scholar, and hand searches (Figure 1). After screening for duplicates and excluding documents that did not meet our inclusion criteria, we reviewed the full texts of 855 documents for further assessment. Of these, we further excluded documents that did not describe and analyze primary data, documents that referenced identical data and findings, and documents in which the interventions were not described sufficiently to determine their similarity to CLTS. In total, 200 out of the 5,884 documents were included in this review.

The results are organized into four sections: broad characteristics of the literature, quality appraisal, summary of quantitative outcomes and indicators from the literature, and qualitative analysis of factors influencing CLTS implementation.

Characteristics of Included Literature

Table 2 presents broad characteristics of the 200 documents included in the review (see Excel Tables S1–S5 for detailed study-level information and Table S3 for a full list of included documents). One hundred and sixty-two (81%) documents were gray literature. Ten of the 14 (71%) quantitative evaluations were journal-published literature. One hundred and twenty-seven (64%) documents were based exclusively on CLTS interventions, 47 (23%) included CLTS interventions as part of a larger WaSH project, and 26 (13%) documents were based on CLTS-like interventions. The top three publishers of gray literature were the United Nations Children’s Fund (UNICEF); the Water, Engineering, and Development Centre at the University of Loughborough; and the Institute of Development Studies at the University of Sussex.

Table 2. Characteristics of 200 included documents.
Characteristica (n=200) No. of documents (%)
Literature type
  Journal-published literature 38 (19%)
  Gray literature 162 (81%)
Study type
  Quantitative evaluation 14 (7%)
  Qualitative study 29 (14%)
  Case study/project report 157 (79%)
Intervention topic
  Community-led total sanitation (CLTS) only 127 (64%)
  CLTS+other water, sanitation, and hygiene intervention 47 (23%)
  CLTS-like interventions (e.g. Community Approaches to    Total Sanitation, Total Sanitation Campaign) 26 (13%)
World regions representedb
  Africa 125 (63%)
  South Asia 60 (30%)
  East and Southeast Asia 33 (17%)
  Pacific Islands 3 (2%)
  Latin America and the Caribbean 3 (2%)

aSee Excel Tables S1–S3 for individual study-level information.

bThe sum of documents for world regions is greater than 200 because some documents covered multiple world regions (two world regions, n=6; three world regions, n=6; four world regions, n=2).

Fifty-three countries were represented across the literature. Twenty-seven documents reported experiences from more than one country and often multiple regions of the world. Most documents reported experiences from Africa (n=125, 63%), followed by South Asia (n=60, 30%), and Southeast Asia (n=33, 17%) (Table 2). The most represented countries were India (n=29, 15%), Kenya (n=26, 13%), Nepal (n=19, 10%), Indonesia (n=19, 10%), Ethiopia (n=15, 8%), and Bangladesh (n=15, 8%). More than three-fourths of the literature was published after 2010.

Quality of the Literature

Figure 2 shows quality scores in each category by literature type (journal-published vs. gray) and study type (quantitative evaluations vs. qualitative studies vs. case studies and project reports). Scores are presented as a percentage.

Figures 2A, 2B, and 2C are bar graphs with standard error of the mean plotting mean scores (percentage of maximum) (y-axis) by literature type and study design (x-axis) for the quality of reporting, minimizing risk of bias, and appropriateness of conclusions, respectively.

Figure 2. Quality appraisal scores (mean percentage) by literature type and study design for (A) quality of reporting, (B) minimizing risk of bias, and (C) appropriateness of conclusions. Raw scores in each category were converted into percentages, and documents that received the maximum possible score in a category were assigned a value of 100%. Error bars represent standard error of the mean.

The average score for quality of reporting was higher for journal-published literature than for gray literature overall (80% vs. 58%, respectively), as well as for each study type (Figure 2A). No document scored zero on quality of reporting, and 13 of the 200 documents received a full score (6 points, 100%). There was greater variability in quality of reporting scores for gray literature than for journal-published literature. Case studies and project reports received the lowest average score (57% for both literature types combined).

On average, journal-published literature scored better than gray literature in minimizing risk of bias across all study types (Figure 2B). Quantitative evaluations had the highest average score of 69% across all literature types (average values of 85% for journal-published vs. 55% for gray literature). Three quantitative evaluations scored below 50% (i.e., <3.5 points out of 7 possible), and one study scored 100% in this category. Qualitative studies had an average score of 48% across all literature types (average values of 60% for journal-published vs. 41% for gray literature). Case studies and project reports had an average score of 20% (average values of 53% for journal-published vs. 15% for gray literature). Seventy-two of the 157 case studies and projects reports (71 of which were gray literature) scored zero points on minimizing risk of bias, 22 (13 of which were gray literature) scored ≥50% (raw score ≥2), and one received the maximum score (4, 100%). Sixty-two percent of the 200 documents received a score of zero for the independence of data collection criterion.

We used the appropriateness of conclusions category to assess interpretation of findings, description of limitations, and whether conclusions were within the scope of the study design. Appropriateness of conclusion scores were generally higher across all study types than the other two categories, and there was less variation between journal-published and gray literature within each study type (Figure 2C). Case studies and project reports had the lowest average score in this category of 61% for journal-published and 62% for gray literature. Seven of the 14 quantitative evaluations (two of which were gray literature) received a maximum score of 3 (100%), as did 13 of the 29 qualitative studies (10 of which were gray literature), and 27 of the 157 case studies and project reports (23 of which were gray literature). No quantitative evaluations received a score of zero, whereas two qualitative studies and three case studies and project reports received a score of zero. Study limitations were not described in 128 of the 200 documents (64%), including 17 of 38 documents from journal-published literature (45%) and 111 of 162 gray literature documents (69%). However, 63% of all 200 documents received a full score for the conclusions criterion, indicating that stated conclusions and implications were within the scope of study design and data collection methods.

Measuring the Effectiveness of Community-Led Total Sanitation

The main characteristics of the 14 quantitative evaluations and the main outcomes of the interventions they evaluate are presented in Table 3. Through qualitative coding of all documents, we also aggregated a list of commonly used indicators grouped under WaSH outcomes, CLTS process, behavioral outcomes, extended impact, and motivators for behavior change (Table 4). The following section reports quantitative outcomes from the 14 evaluations as well as indicators used across all 200 documents. Because of differences in study designs, it was not possible or appropriate to calculate pooled estimates of health impacts.

Table 3. Summary of quantitative evaluations with a comparison group (n=14).
Country Reference Intervention description Intervention time frame Main outcomes
Latrine ownership Latrine quality Open defecation practice Health impacts
Study design: longitudinal randomized controlled trial
  Ghana Crocker et al. 2016a CLTS+training natural leaders vs. CLTS 2012–2014 +18.3 pp vs. conventional CLTS Latrines in both CLTS groups less durable than preexisting latrines −19.9 pp in CLTS+training natural leaders group
  India Pattanayak et al. 2009 TSC with IEC vs. control 2006 +29%; +34% for subsidy group and +21% for no-subsidy group vs. control
  India Patil et al. 2014 TSC with IEC vs. control Not reported +19 pp vs control (41% vs. 23% control) −9 to 10 pp among adults No significant difference in diarrhea, HCGI, anemia, or growth outcomes
  India Dickinson et al. 2015 TSC with IEC vs. control 2006 +27 pp (35% vs. 15% control) Reduction in self-reported diarrhea not significant
  Indonesia Cameron et al. 2013 CLTS+sanitation marketing vs. control 2008–2011 +4.0 pp vs. baseline; +31% vs. control +6 pp (−17%) in nonpoor households −1.4 pp (−30%) in diarrhea; decrease in parasitic infection; increases in height and weight among nonpoor households without sanitation at baseline
  Indonesia Borja-Vega 2014 CLTS+sanitation marketing (subgroup analysis) vs. control 2008–2011 −20.6% in some ethnic groups; not significant among female-headed households −5.8% diarrhea (children <5 y) in female-headed households; significant increase, height for age, head circumference in Madurese ethnic group; other impacts not significant
  Mali Pickering et al. 2015 CLTS vs. control Not reported +32 pp vs. control group CLTS latrines >2 times more likely to have cover, less likely to have flies −23 pp among adult women (−71%); −24 pp (−71%) among adult men; −43 pp (−49%) among children 5–10 y; −43 pp (−51%) among children <5 y No difference in diarrheal prevalence; +0.18 height-for-age z-score; lower likelihood of childhood stunting (35% vs. 41%); 22% children under-weight vs. 26% control
  Mozambique Godfrey et al. 2014 CLTS+safe water vs. control 2008–2013 1 million new users −29% Self-reported water-related diseases between 2008–2010
  Tanzania Briceño et al. 2015 CLTS vs. CLTS+handwashing vs. control 2009–2011 +12.4 pp (+33%) vs. control No difference across groups −12 pp (−52%) No significant impacts
Study design: longitudinal quasi-experimental design
  Ethiopia Crocker et al. 2016b Teacher-facilitated vs. health extension worker–facilitated CLTS 2012–2014 +9.0 pp in conventional CLTS vs. teacher-facilitated CLTS Both interventions improved floors, superstructure, cleanliness, handwashing materials −9.2 pp in teacher-facilitated CLTS
Study design: single group, baseline vs. end line
  Kenya Schlegelmilch et al. 2016 CLTS+other WaSH components 2007–2010 +24 pp (43% vs. 19% baseline)
  Philippines UNICEF 2016 PhATS 2014–2016 +12.6 pp (76.3% end line vs. 63.7% baseline) using improved nonshared facility No significant change vs. baseline (15.2%)
Study design: comparative cross-sectional
  Ethiopia BDS-Center for Development Research 2016 CLTSH vs. control 2012–2015 60.8% latrine use in intervention vs. 58% in control 27.4% in intervention vs. 33.0% in control 24.8% self-reported diarrhea prevalence in children in intervention vs. 30% in control
  Kenya Makotsi et al. 2016 CLTS vs. control Not reported 6.7% in intervention vs. 74.6% in control 11.1% two-week diarrhea prevalence in intervention vs. 21.6% in control

Abbreviations: CLTS, community-led total sanitation; CLTSH, CLTS+hygiene; IEC, information education and communication; pp, percentage point; PhATS, Philippines Approach to Total Sanitation; TSC, Total Sanitation Campaign; WaSH, Water, sanitation, and hygiene.

Table 4. Indicators of progress and outcomes measured in community-led total sanitation programs.
Indicator (n=200) No. of documents (%)
Water, Sanitation, and Hygiene (WaSH) outcomes
  No. (%) of people with access to latrines 124 (62%)
  No. of communities declared/certified open    defecation–free (ODF) 113 (57%)
  No. (%) of people using latrines 52 (26%)
  Quality of latrine (various measures) 52 (26%)
  Health outcomes/impact (various measures) 51 (26%)
  Type of latrine constructed 44 (22%)
  Change in environmental sanitation (various measures) 44 (22%)
  Presence of handwashing station 41 (21%)
  No. (%) of people with access to water 34 (17%)
  No. of beneficiaries affected by intervention 31 (16%)
  No. (%) of people practicing open defecation 31 (16%)
  No. (%) of people reverting to open defecation 14 (7%)
  Presence of cleaning materials near latrine (soap or ash) 10 (5%)
  Distance from latrine to water source 3 (2%)
  Water quality 1 (1%)
  Distance from latrine to home 1 (1%)
CLTS process
  No. of training events held/people trained 63 (32%)
  No. of communities triggered 52 (26%)
  Costs of CLTS activities and/or latrine hardware 36 (18%)
  Presence of WaSH/CLTS Committee 35 (18%)
  Collection of baseline data 33 (17%)
  Provision of community rewards for ODF 19 (10%)
  Presence of government champions 17 (9%)
  No. of follow-up visits 17 (9%)
  Attendance at triggering events 16 (8%)
  Presence of sanctions/enforcement mechanisms 14 (7%)
  Observation of latrine upgrading during posttriggering 13 (7%)
  No. of natural leaders identified 10 (5%)
  Sustainability of ODF status 11 (6%)
  Provision of incentives or rewards to volunteers 9 (5%)
Behavioral outcomes
  Awareness of consequences of open defecation 38 (19%)
  Change in handwashing behavior 35 (18%)
  Satisfaction with latrine (including time savings) 17 (9%)
  Change in social norms 14 (7%)
  Child feces disposal practices 4 (2%)
Extended impact
  Positive nonsanitation outcomes resulting from CLTS 27 (14%)
  Diffusion of CLTS message to neighboring    communities 24 (12%)
  Influence of intervention on sanitation policy 15 (8%)
  Intervention influence on women and girls 13 (7%)
  No. of natural leaders that became CLTS facilitators 8 (4%)
  Sense of ownership 5 (3%)
Motivators for behavior change
  Improved health 35 (18%)
  Dignity or pride 29 (15%)
  Shame or embarrassment 16 (8%)
  Safety 14 (7%)
  Privacy 12 (6%)
  Empowerment 11 (6%)
  Convenience 11 (6%)
  Upgraded social status 5 (3%)

Abbreviations: CLTS, community-led total sanitation; ODF, open defecation–free.

Nine evaluations were randomized controlled trials (RCTs), one used a quasi-experimental design, two used cross-sectional designs with a comparison group, and two were baseline to end-line evaluations of a single intervention group (Table 3). The RCTs comprised evaluations of the following interventions and comparison groups: CLTS vs. a control group in Mali (Pickering et al. 2015); a four-arm intervention of CLTS, CLTS plus handwashing, handwashing only, and a control group in Tanzania (Briceño et al. 2015); CLTS with sanitation marketing vs. a control group in Indonesia (Cameron et al. 2013; Borja-Vega 2014); TSC in India combining behavior change activities with the option of subsidies vs. control groups (Patil et al. 2014; Pattanayak et al. 2009; Dickinson et al. 2015); the One Million Initiative in Mozambique that included CLTS vs. a control group (Godfrey et al. 2014); and a comparison of conventional CLTS to CLTS plus training natural leaders in Ghana (Crocker et al. 2016a). The quasi-experimental study compared conventional CLTS in Ethiopia facilitated by Health Extension Workers (HEWs) to teacher-facilitated CLTS (Crocker et al. 2016b). Of the two single-group evaluations, one was a baseline to end-line evaluation of the Philippines Phased Approach to Total Sanitation (UNICEF 2016), and the other was a baseline to end-line evaluation of CLTS and other WaSH components in Kenya (Schlegelmilch et al. 2016). Finally, of the two comparative cross-sectional studies, one evaluated CLTS and Hygiene in Ethiopia to a control group (BDS-Center for Development Research 2016), and the other assessed health outcomes in a CLTS group vs. a control group in Kenya (Makotsi et al. 2016).

Latrine ownership, use, and quality indicators were identified in most of the 200 documents, but diverse measures were used (Table 4). A few documents contained definitions of the types of latrine that would be acceptable, and others outlined latrine quality indicators such as a cover, superstructure, handwashing facility with soap, and evidence of use. Of the 14 quantitative evaluations, all but two measured either private or household latrine ownership or latrine use after CLTS, and four evaluations also reported some measure of latrine quality, such as the presence of a cover, concrete slab, superstructure, or availability of handwashing materials (Table 3). Overall, the quantitative evaluations reported a statistically significant increase in private or shared latrine construction in intervention groups compared to comparison groups. The Mali CLTS evaluation reported a 32-percentage-point (pp) increase in latrine use in intervention villages and no statistically significant increase in the control group (Pickering et al. 2015) 18 mo after the intervention. In Orissa, India, a 29-pp increase in ownership 4 to 6 mo after the intervention was attributed to the subsidy and behavior change intervention, with two-thirds of the overall treatment effect due to the CLTS-like component compared to the subsidy component of the intervention (Pattanayak et al. 2009). The CLTS studies from Ghana and Ethiopia differed slightly from other included studies in that they compared changes in sanitation outcomes between conventional CLTS in that country and a modification of the approach. The Ghana study reported an increase in private latrine ownership of 18.3 pp from training natural leaders (Crocker et al. 2016a), and the Ethiopia study reported that where HEWs facilitated CLTS, latrine ownership increased by 9 pp more than where teachers facilitated CLTS soon after the interventions (Crocker et al. 2016b).

Declaration or certification of ODF status was the second most common indicator in the literature after latrine access, but no consistent definition was reported for ODF (Table 4). Latrine coverage was the most widely used proxy measure of ODF status. Status was measured at the community level, most often by CLTS facilitators or local government. Most documents that attempted to define ODF wrote about the absence of open defecation in the environment, but few described criteria or frequency for verifying this observation. There was some recognition that “the process of maintaining an open defecation free community is not static and communities cannot simply be checked off and assumed to be ODF, without systems in place that monitor and assist households to repair/replace/rebuild their latrines” (Haq and Bode 2009). In one program in the Philippines, ODF included the “enactment of local legislation at the village level supporting CLTS activities” and the “implementation of other local government activities that supported the maintenance of ODF status” (Belizario et al. 2015). Of the 14 quantitative evaluations summarized in Table 3, two reported the status of ODF certification conducted by the government. In Ethiopia, four of the six kebeles (subgovernmental units) were certified as ODF in the first year, and the remaining two were certified during the follow-up stage (Crocker et al. 2016b). In Mali, 97% of 60 villages in the treatment group were certified as ODF (Pickering et al. 2015).

Two other indicators relating to open defecation identified in the literature were the number of people practicing open defecation (n=31, 16%) in a community, and the number reverting to open defecation after a community had achieved ODF status (n=14, 7%) (Table 4). Change in open defecation practice was reported in ten quantitative evaluations, measured at the household or individual level. As shown in Table 3, results varied considerably, from no statistically significant change between baseline and end line (15.2%) in the Philippines (UNICEF 2016), to a statistically significant 6-pp (17%) decrease in open defecation by nonpoor households in Indonesia (Cameron et al. 2013), to a 23- to 24-pp (71%) decrease in open defecation by adults in Mali (Pickering et al. 2015). The study from Maharashtra reported a 9- to 10-pp decrease in open defecation by adults; reductions were greater in below-poverty-line households or households that did not have latrines at baseline compared to wealthier households or those with latrines at baseline (Patil et al. 2014).

Fifty-one (26%) documents reported some measure of change in health status in communities after CLTS, often comprising anecdotal reports of changes in diarrhea after achieving ODF status (Table 4). Nine quantitative evaluations measured potential health impacts through self-reported changes in diarrhea prevalence or anthropometric measures in children (Table 3). The study from Mali reported no differences in childhood diarrheal prevalence between CLTS and comparison villages, but reported modest statistically significant improvements in child height, stunting, and weight (Pickering et al 2015). One study from Indonesia reported a 1.4-pp (30%) reduction in diarrhea in CLTS communities, decreases in the intensity of parasitic infection, and increases in height and weight among nonpoor households that had no sanitation at baseline (Cameron et al. 2013). A comparative cross-sectional study from Kenya reported lower diarrhea prevalence in the CLTS group vs. a control group (Makotsi et al. 2016). Four other studies did not find statistically significant evidence of health impacts due to the intervention (Borja-Vega 2014; Briceño et al. 2015; Patil et al. 2014; Dickinson et al. 2015), and one evaluation did not report statistically significant health outcomes (BDS-Center for Development Research 2016).

In addition to quantitative sanitation and health indicators, some documents described qualitative measures of CLTS success, such as influence on sanitation policy (n=15, 8%), influence of the intervention on women and girls, particularly with regard to participation and leadership in CLTS activities (n=13, 7%), diffusion of CLTS messages to other communities (n=24, 13%), and positive nonsanitation outcomes resulting from CLTS, such as community mobilization for other development activities (n=27, 14%) (Table 4).

Factors Affecting Community-Led Total Sanitation Implementation and Outcomes

Table 5 presents enabling and constraining factors that emerged inductively from a qualitative content analysis of themes discussed in the 200 documents, as well as the stage of CLTS in which they occur. The 21 implementation-related factors fall under three domains: policy environment, implementation quality, and administrative context. The 22 community-related factors fall under four domains: environment, capacity, participation patterns, and behavior.

Table 5. Factors that facilitated or constrained implementation by stage of community-led total sanitation.
Implementation and community-related factors (n=200)a No. of documents (%) Stage of CLTSb
Pretriggering Triggering Posttriggering
Policy environment
  National government awareness and buy-in for CLTS 41 (21%) X X X
  National sanitation policy vis-à-vis CLTS implementation 37 (19%) X X
  Ambitious national ODF and/or sanitation targets 26 (13%) X X
  History of latrine subsidy provision in the country 21 (11%) X X X
  Ongoing latrine subsidy programs near triggered communities 20 (10%) X
Implementation quality
  Triggering quality 80 (40%) X
  Frequency and effectiveness of follow-up activities in villages 54 (27%) X
  Facilitator skill 45 (23%) X X
  Provision of technical support on latrine construction 44 (22%) X
  Community enforcement measures for noncompliance 39 (20%) X
  Provision of incentives or rewards to villages for ODF status 32 (16%) X
  Planning 30 (15%) X X X
  Provision of latrine subsidies in triggered communities 25 (13%) X
  Provision of incentives to community volunteers 13 (7%) X X
  Presence of exchange visits between community leaders 12 (6%) X
Administrative context
  Local government ownership of CLTS 84 (42%) X X
  Institutional capacity of implementers 66 (33%) X X X
  Administrative and financial arrangements 60 (30%) X X X
  Presence and functioning of M&E system 42 (21%) X
  Coordination between implementing organizations 37 (19%) X X X
  Presence/functioning of sanitation working groups 14 (7%) X
Community environment
  Climate conditions 33 (17%) X X
  Soil or groundwater conditions 28 (14%) X X
  Access to water in community 23 (12%) X X
  Remoteness of community 13 (7%) X X
Community capacity
  Access to supply of latrine hardware 62 (31%) X
  Availability of financial resources 54 (27%) X
  Technical knowledge of latrine construction 24 (12%) X
  Availability of land or land ownership 18 (9%) X X
  Availability of time to construct latrines 11 (6%) X
  Awareness of benefits of stopping open defecation 10 (5%) X X X
Community participation
  Community participation in CLTS 82 (41%) X X
  Presence of village-level leadership 50 (25%) X X X
  Initiative of “natural leaders” 29 (15%) X
  Social cohesion 27 (14%) X X X
  Sense of community responsibility 25 (13%) X X
  Traditional beliefs about women and children’s role in society 9 (5%) X X
Community behavior
  Expectation of subsidy for latrines 29 (15%) X
  Preference for open defecation 20 (10%) X X
  Traditional beliefs regarding open defecation 19 (10%) X X
  Alternative priorities (other than sanitation) 14 (7%) X X X
  Community’s trust in implementers’ motives 11 (6%) X X
  Preference for a better latrine 10 (5%) X

Abbreviations: CLTS, community-led total sanitation; M&E, monitoring and evaluation; ODF, open defecation–free.

aThe factors listed in this table emerged inductively from qualitative coding and analysis of all included literature. Percentages provided are out of all 200 documents, and are meant to illustrate how frequently the respective factor was mentioned in the CLTS literature that was reviewed.

bThe pretriggering stage comprises community selection, facilitator training, baseline information, and community entry; the triggering stage comprises a community-wide meeting with participatory exercises to trigger shame and disgust; and the posttriggering stage includes routine follow-up visits to verify and certify ODF status in communities.

Policy environment.

Eighty-four (42%) documents referred to the influence of the policy environment on CLTS activities, often about policies regarding latrine subsidies and latrine quality (Table 5). The national sanitation policy (n=37, 19%) was reported more often as a constraint than as an enabler. Policy that promoted specific national latrine standards was perceived to conflict with the CLTS message of building a latrine with whatever means available. Policy that encouraged hardware subsidies—most often targeted to the poor—often conflicted with the no-subsidy approach of many CLTS projects. A history of latrine subsidies in communities that were to be triggered with CLTS, or current provision of subsidies near CLTS communities, were cited as constraints, and were often a result of policy decisions beyond the control of CLTS implementers.

A few documents mentioned latrine subsidy policy being used to the advantage of CLTS, such as in Nigeria, where WaterAid Nigeria prioritized follow-up activities on households that had hardware available from previous subsidy programs, “as these households are easiest targets” (Bawa and Ziyok 2013). They reported that this approach led to faster latrine construction because households did not need to think about technology options or financing. A study in Cambodia also reported that “ODF has been reached regularly in the dry season” in villages with subsidy programs, although they also reported that “proximity to on-going subsidized programmes erodes the effectiveness of CLTS” (Kunthy and Catalla 2009).

As shown in Table 5, setting national targets for sanitation was described in 26 (13%) documents and was a constraint in all but one case, where it created an incentive for local government officers in Kenya (Musyoki 2010). Most documents that referred to this factor noted that setting targets created a sense of top-down policy making that conflicted with the community-led nature of CLTS (Crocker et al. 2016c; Davis 2012; De Silva 2013; UNICEF WCARO 2011), led to a focus on rapid latrine construction rather than behavior change (AAN Associates 2013; Dyalchand et al. 2009; Haq and Bode 2009; Jha 2007; Kar and Bongartz 2006; Pardeshi et al. 2008; USAID 2014), led to community triggering outpacing capacity to follow-up (Toft and Onabolu 2012; UNICEF 2014), and created an incentive for implementers to misrepresent data (Mukherjee et al. 2012).

Implementation quality.

Factors relating to implementation quality were reported in 149 (75%) documents (Table 5). Adequate preparation and planning in the pretriggering stage, including the importance of systematic community selection, was emphasized in 30 (15%) documents. Some mentioned the need to target specific types of communities rather than using CLTS everywhere (Burton 2007; Crocker et al. 2016b; Evans et al. 2009; Global Sanitation Fund 2015; Kunthy and Catalla 2009), but one suggested that targeting certain communities leaves behind those communities with unfavorable conditions (Bawa and Ziyok 2013).

The importance of the facilitators’ skills (n=45, 23%) and quality of triggering events (n=80, 40%) were identified as determinants of CLTS success, with an underlying theme of adaptation (Table 5). A practitioner account from Zimbabwe emphasized the “need to be culturally insensitive during facilitation” by not being afraid to use bold terminology and to prioritize creative adaptations of triggering tools based on the context, with the aim to “create a sense of shame, fear and disgust” without “teaching, preaching or prescribing” (Chimhowa 2010). On the other hand, a study critical of CLTS in Indonesia concludes that “the use of shaming and taunting both disqualifies it as an empowerment approach and is likely to undermine its effectiveness in promoting long-term behaviour change. Even if shaming were shown to be effective, the morality of punishing the poor for their circumstances requires deeper consideration” (Engel and Susilo 2014).

Eight of the ten implementation quality factors presented in Table 5 were referred to primarily in the context of posttriggering activities. Fifty-four (27%) documents provided examples of frequent follow-up activities by NGOs or local government helping, or poor follow-up hurting CLTS outcomes. As part of follow-up, the theme of improving monitoring and evaluation of programs was mentioned in 48% of the literature. Many expressed a need for more systematic evaluations of CLTS projects and better use of data that are already being collected by practitioners. In one study, authors observe: “Several nongovernment organizations in the WASH sector worldwide have developed different protocols for defining, declaring, and certifying ODF status in communities, yet no protocol has been recognized as the global standard” (Belizario et al. 2015). Based on their experience with CLTS in Indonesia, authors from the Water and Sanitation Program of the World Bank recommended that “post-triggering processes should be given a verifiable structure by establishing and periodically checking for desired progress quality indicators/milestones for success in triggered communities in order to improve institutional accountability for and the quality of follow-up” (Mukherjee et al. 2012). This need for a reporting structure was echoed in reports from Kenya (Tiwari 2011) and Ghana (Magala and Roberts 2009), among others.

Technical support (n=44, 22%) was often cited as an enabler in projects that provided guidance directly to communities on latrine construction or trained masons to improve toilet design (Evans et al. 2009; Huda 2009; Kalimuthu 2008; Magala and Roberts 2009; SEED Madagascar 2016; Shayamal et al. 2008; WaterAid India 2008). Technical support and subsidies were both contentious in the literature, with several practitioner accounts defending a strictly no-subsidy implementation of CLTS, whereas others advocated for greater flexibility, such as this evaluation of a UNICEF program: “Many implementers share the opinion that more work on the technical standards together with targeted subsidies are unavoidable to help reach the households build latrines and reach the ODF status in such areas” (Hydroconseil et al. 2014).

The presence of enforcement mechanisms or sanctions on open defecation or latrine construction was described in 39 (20%) documents (Table 5). One study in Nepal reported that “coercive methods … did not always bring out tangible results” (Jha 2007), and in Bangladesh, enforcement led some people to “construct toilets out of fear of being fined without understanding the reasoning for doing so or the best methods for construction. This in turn leads to poor use of the latrines” (Kar and Bongartz 2006). An evaluation of CLTS in West Africa found that “such punitive measures seem out of line with the CLTS spirit of self-help and dignity. However, community enforcement may be considered as an appropriate additional measure [if] it is implemented in a real participatory and community-based way, with a collective decision” (UNICEF WCARO 2011). Local by-laws were described as effective in several settings; an evaluation of CATS reported that “in many countries, the strongest evidence of a change in social norms is the genuine adoption and the enforcement of formal and informal rules/bylaws at the level of the community, accepted by all the community members and recognized as collective rules which cannot be transgressed without consequences” (Hydroconseil et al. 2014).

Administrative context.

Over half of the literature (n=112, 56%) contained factors relating to the administrative context (Table 5). This domain included institutional capacity to implement CLTS (n=66, 33%), administrative or financial arrangements (n=60, 30%), and coordination between implementing organizations (n=37, 19%). Concerns were documented relating to time availability, technical experience, skilled human resources, and the capacity to plan, budget, and allocate resources for CLTS. Some adaptations were also documented, such as “peer-to-peer accountability mechanisms” for government health surveillance assistants in Malawi (Kennedy and Meek 2013) and village-level microplanning exercises for local government in Kenya (Singh and Balfour 2015). More documents described existing administrative and financial arrangements as constraints rather than enablers of CLTS activities.

Eighty-four (42%) documents cited local government ownership of CLTS as an important factor for success, scale-up, or sustainability (Table 5). The most effective level and type of local government involvement was unclear. In an evaluation from Zambia, authors state: “The level of support given to CLTS in certain districts is obvious, with a high level of involvement from everyone from Town Clerks and Chiefs to government representatives across sectors, knowing and understanding what the CLTS approach means. This level of understanding surely forms the basis for sustainability in an institutional sense” (Morris-Iveson and Siantumbu 2011). On the other hand, in Moroto, Uganda, “support for better sanitation and hygiene from political leadership was reported as lacking or weak in most respects … and served to undermine efforts of the extension staff as the latter strive to promote ODF villages” (Asingwire 2012). Local government ownership was often related to decentralization as well. For example, in Cambodia, decentralization helped transfer financial responsibility to the local government, creating a local source of funds for district and commune activities (Kunthy and Catalla 2009). On the other hand, it created uncertainty in Kenya (Crocker et al. 2016c), and was not matched with sufficient institutional capacity in Indonesia (Engel and Susilo 2014).

Community environment.

Across 66 (33%) documents, environmental, geographical, and climate-related factors were cited, such as poor soil conditions and floods destroying latrines (Table 5). However, remoteness of a village (n=13, 7%) was sometimes an enabler, as remote villages were less likely to have been exposed to subsidy projects and might therefore be more receptive to the CLTS message. Access to clean water in triggered communities (n=23, 12%) was described as an enabler in several documents. For example: “one of the key entry processes is access to water. In the project communities water points were rehabilitated and in few cases new ones were installed. Communities clearly associated the effectiveness of CLTS to availability of water” (Burton 2007). Adeyeye also noted that WaterAid Nigeria “holds that access to water is a necessary prerequisite to access to adequate sanitation” (Adeyeye 2011).

Community capacity.

Ninety-seven documents (49%) cited at least one community capacity factor (Table 5). The most frequently identified factors related to building latrines were access to supply of latrine hardware (n=62, 31%), availability of financial resources (n=54, 27%), and technical knowledge of latrine construction (n=24, 12%). Studies often reported community members’ desire for more guidance from implementers on how to build a high-quality latrine to avoid costly maintenance and repairs that could result in reversion to open defecation. There were examples of local mechanisms to address financial constraints, such as creating access to credit through village savings and loans associations (Adhikari et al. 2008; De Silva 2013; Global Sanitation Fund 2015; Magala and Roberts 2009; Mwanzia and Misati 2013; Tremolet et al. 2010) or collective community efforts to build latrines (Mukherjee et al. 2012; Priyono 2009). Nevertheless, poor latrine quality and resulting sustainability challenges emerged as important themes.

Community participation patterns.

A key participant in CLTS is the natural leader, typically a community member who emerges in the triggering process as someone particularly motivated to improve sanitation. Eighty-two (41%) documents broadly referred to natural leaders (Table 5). Twenty-nine (15%) documents specifically noted the initiative of natural leaders as an enabler or barrier to CLTS, but only a few gave concrete examples, such as the challenge in identifying natural leaders or how training natural leaders in latrine construction or mobilization techniques proved to be effective. One study reported that training led to greater participation and better sanitation outcomes in Ghana (Crocker et al. 2016a), and practitioners in Madagascar reported that training helped motivate natural leaders to be more active in their communities (SEED Madagascar 2016).

Broader mobilization, participation, and motivation of community members in triggering and posttriggering activities was reported in 82 (41%) documents as an important reason for success or failure of CLTS (Table 5). A sense of community responsibility (n=25, 13%) and social cohesion (n=27, 14%) emerged in several documents. Smaller, homogeneous communities tended to be more successful (e.g., Evans et al. 2009; Haq and Bode 2009; Mukherjee et al. 2012; Tyndale-Biscoe et al. 2013; USAID 2014; Venkataramanan 2016), and greater cohesion was also connected with greater likelihood of self-help initiatives, for example, gotong rayong in Indonesia (Mukherjee et al. 2012).

Community behavior.

Expectation of subsidies for latrine construction was the most commonly cited behavioral constraint in the literature (n=29, 15%) (Table 5). Preference for open defecation was also an important behavioral factor (n=20, 10%) that related to slow progress or no change in sanitation behavior after triggering. Part of this was reported to be due to cultural or religious beliefs regarding open defecation or latrine use, which were often cited as either enablers and barriers to CLTS. For example, speaking about defecation was considered to be a private matter in several settings (Dittmer 2009; Evans et al. 2009; Shayamal et al. 2008). There are also taboos around different members of the household using or sharing a latrine (Bulaya et al. 2015; Burton 2007; Kappauf 2011; Mukherjee et al. 2012; Zombo 2010) or superstitions around latrine use (Dittmer 2009). Dittmer (2009) gave an example of an ethnic group in Burkina Faso with the tradition that “if someone gives you food, you are expected to defecate in his field (and fertilize the crops), as the act of giving entitles the giver to receive something in return.” These traditional beliefs were used to adapt triggering in several programs (n=19, 10%) by engaging religious leaders or using passages from the Bible or Quran during triggering events.

Fourteen (7%) documents noted that priorities other than sanitation can affect the response to CLTS (Table 5). For example, Engel and Susilo document one village chief’s observation in Indonesia that “despite the supposedly participatory approach of the CLTS, the villagers did not want a sanitation project and would have preferred an adequate irrigation system for their farmland and a programme for replanting an area of cleared forest located near their farmland” (Engel and Susilo 2014).

Discussion

In our systematic review of CLTS and related interventions, we comprehensively characterize the state of evidence through a detailed quality appraisal, summary of quantitative outcomes, and qualitative analysis of factors affecting CLTS implementation and outcomes.

Quality of the Literature

Our quality appraisal indicates that evidence available to practitioners and policy makers is of variable quality, particularly regarding the ability to estimate the impact of CLTS on sanitation, health, or other community outcomes. Overall, we found that the journal-published literature was of higher quality than gray literature. We show that case studies and project reports—which were primarily found in the gray literature—did not adequately describe their study design, data collection, or data analysis. Poor reporting of study characteristics makes it difficult to judge the objectivity and quality of information presented. Minimizing risk of bias was the weakest link in the quality of CLTS literature across all study types, but particularly in the case of qualitative studies, and case studies and project reports, which rarely described sampling methods, quality control in data collection, or analysis appropriate for the respective study design. Nearly two-thirds of all literature lacked independent data collection; improving this metric alone would minimize the risk of bias of all study types. Although nearly all documents gave context for their findings, there were large gaps in the description of limitations, preventing the reader from understanding the extent to which findings may be generalized. Furthermore, more than one-fourth of the literature overstated conclusions by attributing outcomes to their intervention without an appropriate study design or by making claims about impact using unverified data sources or anecdotes.

CLTS is one of the most common rural sanitation behavior change approaches. It is increasingly being tested in urban settings as well (Murigi et al. 2015; Mwanzia and Misati 2013; Myers 2016; Prabhakaran et al. 2016). Therefore, there is an urgent need to better understand its effectiveness by improving the rigor of the evidence base. By reviewing the literature as a whole, we are able to compare the quality of different study types and identify specific areas for improvement. Case studies and project reports—which have the potential to detail processes and share lessons learned—can be improved through more systematic data collection and analysis, and more thorough reporting to determine the extent of transferability of findings. Quantitative cross-sectional designs—a subset of studies we classified as case studies and project reports—can describe outcomes on a large scale, but can be improved through more detailed descriptions of context and intervention processes. Qualitative studies can provide rich contextual descriptions, perceptions of different stakeholders, and reasons for success or failure, but researchers and implementers using these methods must improve the rigor of data collection and analytical techniques in order for findings to have sound policy and practice implications. Finally, well-designed quantitative evaluations have the potential to attribute outcomes and impact to interventions, but the quality of quantitative evaluations can be improved through more rigorous data collection methods and better descriptions of context, process, and study limitations.

Measuring the Effectiveness of Community-Led Total Sanitation

We found few rigorous quantitative evaluations of sanitation and health impacts of CLTS and related interventions. The 14 that we included evaluated interventions from nine countries, whereas CLTS is now practiced in at least 53 countries. These studies reported increases in latrine ownership and decreases in open defecation, but did not corroborate the widespread claims of ODF villages reported in case studies and project reports. As Evans et al. note: “Like many terms in development, [ODF] has become de-linked from its true semantic meaning and become more of a milestone or marker in programme development” (Evans et al. 2009). The term ODF may serve as a motivator for communities to improve sanitation behavior, but is a poor indicator to compare across studies, programs, or countries, given the variety of definitions (see Thomas and Bevan 2014 for a review of various ODF protocols in sub-Saharan Africa). Household-level latrine outcomes, while imperfect, are a better measure of sanitation progress. For research purposes, pairing these with more robust measures of the defecation practices of individual people is a further improvement. To avoid the pitfall of simply counting latrines, programs could add routine measures of open defecation behavior, latrine use, and cleanliness through community monitoring initiatives (Coombes 2011).

One of the primary aims of sanitation is improved health, but measuring these changes is difficult if the sanitation intervention did not result in a sufficient reduction in open defecation or exposure to fecal contamination. Our synthesis of health outcomes from CLTS and related interventions supports findings of previous reviews (Garn et al. 2016; Sclar et al. 2016; Taylor et al. 2015; Wolf et al. 2014) and underscores the challenge of attributing health impacts to sanitation, particularly over a short follow-up. Furthermore, our review supports prior observations that self-reported diarrhea is an unreliable measure of impact (Schmidt 2014), even though one-fourth of the literature contained anecdotal reports of perceived health impact through self-reported reduction in diarrhea. While more studies that consider a variety of sanitation-related health outcomes, including measures of nutritional status, may be beneficial, they tend to be expensive, require the intervention itself to be sufficiently successful to change sanitation outcomes, and require a long enough follow-up period to observe a noticeable change in health outcomes (Cairncross et al. 2010; Gertler et al.2015).

Many programs are unlikely to have the resources or technical expertise to incorporate health impact into monitoring and evaluation systems or to commission such studies. Nevertheless, our review reveals an opportunity for researchers and practitioners to work together to address more immediate implementation and operational research questions by leveraging a variety of study designs. Given the participatory nature of CLTS activities and emphasis on sustained behavior change, such research would be strengthened using mixed methods, including qualitative indicators of participation and perceptions and better measures of social norms for a thorough picture of CLTS effectiveness and impact.

Factors Affecting Community-Led Total Sanitation Implementation and Outcomes

We identified 43 implementation and community-related factors from the literature affecting CLTS. Many were context-specific enablers or constraints to CLTS implementation and outcomes. Other factors, such as local government ownership of CLTS, institutional capacity, importance of facilitators’ skills, and community participation in CLTS, were described in a similar manner across much of the literature. We suggest four important considerations from this qualitative analysis of the literature.

First, our review confirms the narrative of CLTS as a highly adaptable approach. Like Sigler et al. (2015)’s finding that multiple behavior change frameworks are employed in CLTS, we found that shame and disgust, although popular, were not reported as universal motivators that triggered communities; instead, improved health, dignity, and pride were cited more often. Skilled facilitators adapted their triggering techniques based on cultural considerations. However, finding such facilitators was described in the literature as an important constraint. Less-skilled facilitators resorted to either lecturing communities on health benefits or falling back on conventional shaming or disgust-inducing triggering techniques, regardless of their appropriateness in that context (Venkataramanan 2016). We did not find any studies in our review that evaluated the relative effectiveness of different triggering adaptations, despite calls for a closer analysis of the potential human rights implications of CLTS and related techniques (Bardosh 2015; Bartram et al. 2012; Engel and Susilo 2014; Galvin 2015).

Second, although a high degree of flexibility is expected during triggering, lack of structure in posttriggering activities may be less beneficial. The Handbook on CLTS—the de facto manual that most CLTS programs use as a starting point—does not detail the structure of the posttriggering stage, acknowledging that activities are likely to depend on the context and characteristics of the specific community (Kar and Chambers 2008). Our analysis suggests, however, that certain elements of posttriggering activities routinely challenge programs around the world. For example, there was no clear evidence on the effectiveness or appropriateness of providing incentives or subsidies to some communities or on the role of enforcement and sanctions for noncompliance. Our review confirms Bartram et al.’s observation that there continues to be minimal debate or critical review of the effectiveness or humans rights consequences of posttriggering punitive measures (Bartram et al. 2012).

Another set of posttriggering challenges relates to the supply of durable and affordable latrine hardware and technical support on latrine construction. Notably, we identified a debate over the nature of technical support that should be provided to communities for latrine construction. CLTS programs do not follow uniform guidance on technical support, as communities are supposed to identify their own solutions to stop open defecation. Whereas some programs provided detailed technical support on latrine options, trained masons, or attempted to improve the supply chain for hardware, follow-up in other programs simply meant monitoring latrine construction. However, our analysis of the literature suggests a need for additional guidance, as substantive concerns were expressed from both community and implementer perspectives about the quality of latrines built because of CLTS, potentially discouraging sustained behavior change and possibly explaining the minimal effects seen in health impact studies (Papafilippou et al. 2011). We argue that programs should routinely incorporate technical support into the posttriggering stage, particularly when communities prefer durable latrines and express a need for this kind of support. An eight-country evaluation of CLTS in Africa similarly recommended that in the absence of a sanitation marketing program, “the post-ODF approach should include a set of ‘second-phase’ interventions designed to provide advice on how to upgrade and improve sanitation and handwashing facilities using local materials” (Robinson 2016).

Third, we suggest that communities should be selected for triggering based on community characteristics and resources available to maintain routine follow-up activities, including local government ownership. We reveal conflicting views on the scope for application of CLTS, with practitioners often suggesting that it is appropriate in all rural settings, whereas evaluations and studies of CLTS point to more deliberate targeting. This is particularly relevant when community members express priorities other than sanitation.

What interventions should be implemented, then, in places where CLTS is not likely to be successful? While there are some settings where CLTS is never going to be an appropriate intervention, there are also likely to be settings where CLTS may not be successful on its own, but can result in sustained changes when combined or sequenced with other demand-generating or demand-fulfilling approaches, such as sanitation marketing or other WaSH interventions. Further research is needed to understand the most effective combination and sequencing of WaSH interventions. Our review revealed that several programs install water supply projects simultaneously with or following CLTS projects to try to ensure that anticipated gains from sanitation behavior change were not lost due to limited water supply. Several programs also measured total sanitation practices in their CLTS programs such as handwashing, water and food safety, and garbage disposal as opposed to focusing solely on open defecation. Some practitioners consider this lack of standardization of CLTS as a problem for scalability and sustainability (SNV Uganda 2014), but we suggest that it can instead be viewed as an opportunity to expand the conversation to consider CLTS as part of a total WaSH strategy to achieve the WaSH Sustainable Development Goals by 2030 (United Nations General Assembly 2015).

Limitations

Although we present findings from 53 countries, we did not specifically search for non-English documents and may have missed experiences from some countries. Because gray literature is, by definition, not published in peer-reviewed journals, and because it is produced so rapidly, we may not have captured all the available literature. However, we believe we reached saturation and have captured the vast majority of the CLTS evidence base by scanning 5,884 documents from diverse sources and reviewing 200 in detail. Finally, although the content analysis was conducted systematically in two stages by two authors, it is possible that our frequency counts were slightly underestimated, and we were unable to capture every factor and indicator presented across all documents.

Conclusions

This is the first comprehensive systematic review, to our knowledge, of the state of the CLTS evidence base. Most literature on CLTS is on websites and knowledge bases rather than in peer-reviewed journals; this gray literature is more extensive and more accessible. Therefore, the large and inclusive scope of this review offers one of the first aggregate views of the evidence currently available for decision-makers as they consider whether and how to test, adopt, or scale-up CLTS worldwide. By including a variety of literature types (journal-published and gray) and study designs (quantitative, qualitative, and case studies and project reports), we were also able to identify their strengths or weaknesses and compare their relative contribution to the evidence base.

The quality appraisal framework we developed serves as a practical tool for assessing the quality of evidence from sources as varied as NGO reports, qualitative studies, and RCTs. To our knowledge, this is the first tool of its kind that enables a combined assessment of such literature on water and sanitation to develop specific recommendations for improving the evidence base.

The mixed-methods analysis of the quality and content of literature enabled us to pool findings in a much richer way than a meta-analysis of one particular study type would have allowed. By and large, there is substantive room for improvement in the quality of evidence on CLTS. We found that CLTS has been rolled out with minimal rigorous evidence on its effectiveness and impact on sanitation and health outcomes. While quantitative evaluations show reductions in open defecation and increases in latrine coverage, they do not mirror practitioner accounts of widespread elimination of open defecation. There is little evidence for sustained sanitation behavior change as a result of CLTS, and there has been minimal systematic research of the CLTS implementation process and its adaptations. We also provide evidence for the need to improve the structure of CLTS activities and the need to consider CLTS as part of a larger WaSH strategy rather than as a singular solution to changing sanitation behavior.

The research–practice gap can be narrowed if researchers work more closely with implementers to design implementation and operational research studies to address specific challenges relating to sustainable behavior change and change in social norms, as well combining and sequencing of different sanitation or WaSH approaches. Donor agencies and national governments should support researcher–practitioner initiatives to improve the evidence base and provide policy makers opportunities to make more informed decisions to improve sanitation outcomes.


Acknowledgments

This research is a result of a subagreement to the University of North Carolina at Chapel Hill from Plan International USA, which was a recipient of a grant from the Bill & Melinda Gates Foundation (OPP1028953). Additional funding from a National Institute of Environmental Health Sciences Training grant (T32ES007018) supported the primary author and second author’s time. Neither funder had any role in study design, execution, or article preparation. We thank E. Rehfuess for advice on quality appraisal, and D. Saywell, P. Kolksy, and E. Ojomo for comments on the paper.

References

AAN Associates. 2013. UNICEF WASH Pakistan End of Programme Evaluation for Rural Sanitation in Flood-Affected Districts (RuSFAD Phase III) 2011–12. New York:UNICEF. http://www.unicef.org/evaldatabase/index_73521.html [accessed 10 January 2016].

Adeyeye A. 2011. Gender and Community-Led Total Sanitation: a Case Study of Ekiti State, Nigeria. Trop. Resour. Bull. Yale Trop. Resour. Inst. 30:1–27. https://environment.yale.edu/tri/uploads/AAdeyeye2011.pdf [accessed 10 May 2016].

Adhikari S, Shrestha NL, Malla M, Shrestha GR. 2008. Nepal School-Led Total Sanitation Seems Unstoppable. Sanitation and Hygiene Case Study #7. New York:UNICEF. http://www.communityledtotalsanitation.org/resource/school-led-total-sanitation-seems-unstoppable-nepal [accessed 15 December 2015].

Asingwire N. 2012. WASH Initiative for the Rural Poor in 21 Districts in Uganda. “End of Term Evaluation Final Report.” Uganda:SWSA Dept. Makerere University, European Commission, and UNICEF. https://www.unicef.org/evaldatabase/files/30-08-12_WASH_Project_Evaluation_Report_Revised.pdf [accessed 10 January 2016].

Bardosh K. 2015. Achieving “total sanitation” in rural African geographies: poverty, participation and pit latrines in eastern Zambia. Geoforum 66:53–63, 10.1016/j.geoforum.2015.09.004.

Bartram J, Charles K, Evans B, O’Hanlon L, Pedley S. 2012. Commentary on community-led total sanitation and human rights: should the right to community-wide health be won at the cost of individual rights?. J Water Health 10(4):499–503, PMID: 23165706, 10.2166/wh.2012.205.

Bawa K, Ziyok I. 2013. Implementing CLTS in areas with a history of subsidised sanitation programmes. Briefing Paper 1674. Proceedings of the 36th WEDC International Conference, Nakuru, Kenya. Loughborough, UK:Water, Engineering, and Development Centre, https://wedc-knowledge.lboro.ac.uk/details.html?id=20597 [accessed 10 October 2016].

BDS-Center for Development Research. 2016. Outcome Evaluation of Community-Led Total Sanitation and Hygiene (CLTSH) Program in Ethiopia from 2012-2015. Addis Ababa, Ethiopia:UNICEF, https://www.unicef.org/evaldatabase/files/Final_Report_of_CLTSH_Outcome_Evaluation-Designed_Ethiopia_2016-012.pdf [accessed 15 March 2017].

Belizario VY, Liwanag HJC, Naig JRA, Chua PLC, Madamba MI, Dahildahil RO. 2015. Parasitological and nutritional status of school-age and preschool-age children in four villages in Southern Leyte, Philippines: Lessons for monitoring the outcome of Community-Led Total Sanitation. Acta Trop 141(Pt A):16–24, PMID: 25255966, 10.1016/j.actatropica.2014.09.008.

Borja-Vega C. 2014. The effects of the Total Sanitation and Sanitation Marketing programme on gender and ethnic groups in Indonesia. Waterlines 33(1):55–70, 10.3362/1756-3488.2014.007.

Briceño B, Coville A, Martinez S. 2015. Promoting Handwashing and Sanitation: Evidence from a Large-Scale Randomized Trial in Rural Tanzania. World Bank Policy Res. Work. Pap. 7164. Washington, DC:World Bank. http://ssrn.com/abstract=2550526 [accessed 15 December 2016].

Bulaya C, Mwape KE, Michelo C, Sikasunge CS, Makungu C, Gabriel S, et al. 2015. Preliminary evaluation of Community-Led Total Sanitation for the control of Taenia solium cysticercosis in Katete District of Zambia. Vet Parasitol 207(3-4):241–248, PMID: 25591408, 10.1016/j.vetpar.2014.12.030.

Burton S. 2007. “An Evaluation of the WaterAid’s CLTS Programme in Nigeria.” Abuja, Nigeria:WaterAid. http://www.communityledtotalsanitation.org/resource/community-led-total-sanitation-clts-evaluation-wateraid-s-clts-programme-nigeria [accessed 10 January 2016].

Cairncross S, Hunt C, Boisson S, Bostoen K, Curtis V, Fung IC, et al. 2010. Water, sanitation and hygiene for the prevention of diarrhoea. Int. J. Epidemiol 39(suppl 1):i193–i205, PMID: 20348121, 10.1093/ije/dyq035.

Cameron L, Shah M, Olivia S. 2013. Impact Evaluation of a Large-Scale Rural Sanitation Project in Indonesia. World Bank Policy Res. Work. Pap. 6360. Washington, DC:World Bank. 10.1596/1813-9450-6360.

Chimhowa HK. 2010. Freeing the imagination: innovations in CLTS facilitation in Zimbabwe. In: Participatory Learning and Action 61: Tales of Shit: Community-Led Total Sanitation in Africa. Bongartz P , Musyoki SM , Milligan A , Ashley H , eds. London, UK:IEED, 65–72. https://www.iied.org/pla-61-tales-shit-community-led-total-sanitation-africa [accessed 10 October 2016].

Clasen T, Boisson S, Routray P, Torondel B, Bell M, Cumming O, et al. 2014. Effectiveness of a rural sanitation programme on diarrhoea, soil-transmitted helminth infection, and child malnutrition in Odisha, India: a cluster-randomised trial. Lancet Glob. Heal 2(11):e645–e653, PMID: 25442689, 10.1016/S2214-109X(14)70307-9.

Coombes Y. 2011. Utilizing Community-Based Registers to Monitor Improved Access to Sanitation and Hygiene in Tanzania. Washington, DC:The World Bank Water and Sanitation Program. https://openknowledge.worldbank.org/handle/10986/11688 [accessed 10 October 2016].

Crocker J, Abodoo E, Asamani D, Domapielle W, Gyapong B, Bartram J. 2016a. Impact evaluation of training natural leaders during a community-led total sanitation intervention: a cluster-randomized field trial in Ghana. Environ Sci Technol 50(16):8867–8875, PMID: 27428399, 10.1021/acs.est.6b01557.

Crocker J, Geremew A, Atalie F, Yetie M, Bartram J. 2016b. Teachers and sanitation promotion: an assessment of community-led total sanitation in Ethiopia. Environ Sci Technol 50(12):6517–6525, PMID: 27211881, 10.1021/acs.est.6b01021.

Crocker J, Shields KF, Venkataramanan V, Saywell D, Bartram J. 2016c. Building capacity for water, sanitation, and hygiene programming: training evaluation theory applied to CLTS management training in Kenya. Soc Sci Med 166:66–76, PMID: 27543683, 10.1016/j.socscimed.2016.08.008.

Davis B. 2012. Sanitation in Cambodia: a Review. Phnom Penh:ADRA Cambodia. http://genderandwater.org/en/other-resources/reports/rural-sanitation-evaluation-cambodia/ [accessed 10 October 2016].

De Silva P. 2013. Ex-post Evaluation of Water, Sanitation and Hygiene Programme in Eastern Indonesia. Evaluation Report. Indonesia:Government of Indonesia and UNICEF. https://www.unicef.org/evaldatabase/index_72934.html [accessed 10 January 2016].

DeFrancis MP , ed. 2011. Economic Impacts of Inadequate Sanitation in India. Washington, DC:The World Bank Water and Sanitation Program. http://www.wsp.org/sites/wsp.org/files/publications/wsp-esi-india.pdf [accessed 11 November 2016].

Dickinson KL, Patil SR, Pattanayak SK, Poulos C, Yang J-H. 2015. Nature’s Call: Impacts of Sanitation Choices in Orissa, India. Econ Dev Cult Change 64(1):1–29, 10.1086/682958.

Dittmer A. 2009. Towards Total Sanitation: Socio-Cultural Barriers and Triggers to Total Sanitation in West Africa. London, UK:WaterAid. http://www.wateraid.org/~/media/Publications/total-sanitation-socio-cultural-barriers-triggers-west-africa.pdf [accessed 10 October 2016].

Dreibelbis R, Winch PJ, Leontsini E, Hulland KR, Ram PK, Unicomb L, et al. 2013. The Integrated Behavioural Model for Water, Sanitation, and Hygiene: a systematic review of behavioural models and a framework for designing and evaluating behaviour change interventions in infrastructure-restricted settings. BMC Public Health 13:1015, PMID: 24160869, 10.1186/1471-2458-13-1015.

Dwipayanti NMU, Phung TD, Rutherford S, Chu C. 2017. Towards sustained sanitation services: a review of existing frameworks and an alternative framework combining ecological and sanitation life stage approaches. J Water Sanit Hyg Dev 7(1):25, 10.2166/washdev.2017.086.

Dyalchand A, Kale M, Vasudevan S. 2009. What Communication and Institutional Arrangements Influence Sanitation Related Social Norms in Rural India? http://www.communityledtotalsanitation.org/resource/what-communication-and-institutional-arrangements-influence-sanitation-related-social-norms [accessed 1 Dec 2015].

Engel S, Susilo A. 2014. Shaming and sanitation in Indonesia: a return to colonial public health practices? Dev Change 45(1):157–178, 10.1111/dech.12075.

Evans B, Colin J, Jones H, Robinson A. 2009. Sustainability and equity aspects of total sanitation programmes: A Study of Recent WaterAid-Supported Programmes in Three Countries. Global Synthesis Report. London, UK:WaterAid. www.wateraid.org/~/media/Publications/community-led-total-sanitation.pdf [accessed 1 September 2016].

Evans WD, Pattanayak SK, Young S, Buszin J, Rai S, Bihm JW. 2014. Social marketing of water and sanitation products: a systematic review of peer-reviewed literature. Soc Sci Med 110:18–25, PMID: 24704890, 10.1016/j.socscimed.2014.03.011.

Fiebelkorn AP, Person B, Quick RE, Vindigni SM, Jhung M, Bowen A, et al. 2012. Systematic review of behavior change research on point-of-use water treatment interventions in countries categorized as low- to medium-development on the human development index. Soc Sci Med 75(4):622–633, PMID: 22497845, 10.1016/j.socscimed.2012.02.011.

Galvin M. 2015. Talking shit: is Community-Led Total Sanitation a radical and revolutionary approach to sanitation? WIREs Water 2(1):9–20, 10.1002/wat2.1055.

Garn JV, Sclar GD, Freeman MC, Penakalapati G, Alexander KT, Brooks P, et al. 2016. The impact of sanitation interventions on latrine coverage and latrine use: a systematic review and meta-analysis. Int J Hyg Environ Health 220(2 Pt B):329–340, PMID: 27825597, 10.1016/j.ijheh.2016.10.001.

Gertler P, Shah M, Alzua M, Cameron L. 2015. How Does Health Promotion Work? Evidence from the Dirty Business of Eliminating Dirty Defecation. NBER Working Paper Series. Cambridge, MA:National Bureau of Economic Research. 10.3386/w20997.

Global Sanitation Fund. 2015. Sanitation and Hygiene Promotion in Madagascar. Geneva, Switzerland:WSSCC. http://wsscc.org/wp-content/uploads/2015/10/Mada-1-3.pdf [accessed 15 March 2017].

Godfrey S, Van Der Velden M, Muianga A, Vigh M, Gunning JW, Elbers C. 2014. Impact study of the One Million Initiative rural water and sanitation programme in Mozambique. Waterlines 33(1):35–44, 10.3362/1756-3488.2014.005.

Haddaway NR, Collins AM, Coughlin D, Kirk S. 2015. The role of Google Scholar in evidence reviews and its applicability to grey literature searching. PLoS One 10(9):1–17, PMID: 26379270, 10.1371/journal.pone.0138237.

Haq A, Bode B. 2009. Hunger, Subsidies and Process Facilitation: Challenges for Community Led Total Sanitation in Bangladesh. A paper prepared for the Institute of Development Studies CLTS Write Shop held on May 19-23, 2008. Brighton, UK:Institute of Development Studies, University of Sussex. http://www.communityledtotalsanitation.org/resource/hunger-subsidies-and-process-facilitation-challenges-community-led-total-sanitation [accessed 15 December 2015].

Harden A. 2010. Mixed-Methods Systematic Reviews: Integrating Quantitative and Qualitative Findings. Focus Technical Brief No. 25. Austin, TX:National Center for the Dissemination of Disability Research. http://ktdrr.org/ktlibrary/articles_pubs/ncddrwork/focus/focus25/Focus25.pdf [accessed 15 October 2016].

Heale R, Twycross A. 2015. Validity and reliability in quantitative studies. Evid Based Nurs 18(3):66–67, PMID: 25979629, 10.1136/eb-2015-102129.

Huda E. 2009. Natural Leaders Emerged through Community Led Total Sanitation (CLTS) Approach in Bangladesh (Profile and Market Promotion). Sussex, UK:CLTS Knowledge Hub, Institute for Development Studies (IDS). http://www.communityledtotalsanitation.org/resource/natural-leaders-emerged-through-clts-approach-bangladesh-profile-and-market-promotion [accessed 15 December 2015].

Hulland K, Martin N, Dreibelbis R, DeBruicker Valliant J, Winch P. 2015. What Factors Affect Sustained Adoption of Safe Water, Hygiene and Sanitation Technologies? A Systematic Review of Literature. London, UK:EPPI-Centre, Social Science Research Unit, UCL Institute of Fducation, University College London. 10.13140/RG.2.1.5031.4329

Hulland KR, Chase RP, Caruso BA, Swain R, Biswal B, Sahoo KC, et al. 2015. Sanitation, stress, and life stage: a systematic data collection study among women in Odisha, India. PLoS One 10(11):e0141883, PMID: 26551866, 10.1371/journal.pone.0141883.

Hydroconseil, Ecopsis, WEDC (Water, Engineering, and Development Center of Loughborough University). 2014. Evaluation of the WASH Sector Strategy: Community Approaches to Total Sanitation (CATS). Final Evaluation Report. New York NY:UNICEF. https://www.unicef.org/evaluation/files/Evaluation_of_the_WASH_Sector_Strategy_FINAL_VERSION_March_2014.pdf [accessed 10 January 2016].

Jack L, Hayes SC, Scharalda JG, Stetson B, Jones-Jack NH, Valliere M, et al. 2010. Appraising quantitative research in health education: guidelines for public health educators. Health Promot Pract 11(2):161–165, PMID: 20400654, 10.1177/1524839909353023.

Jadhav A, Weitzman A, Smith-Greenaway E. 2016. Household sanitation facilities and women’s risk of non-partner sexual violence in India. BMC Public Health 16(1):1139, PMID: 27821158, 10.1186/s12889-016-3797-z.

Jenkins M, Sugden S. 2006. Human Development Report 2006. Rethinking Sanitation: Lessons and Innovations for Sustainability in the New Millennium. Human Development Report Office Occasional Paper. New York:UNDP. http://hdr.undp.org/en/content/rethinking-sanitation-lessons-and-innovation-sustainability-and-success-new-millennium [accessed 1 November 2016].

Jha HB. 2007. An assessment of CLTS projects and formulation of the strategy on sanitation promotion. Dhobighat, Lalitpur, Nepal:Centre for Economic and Technical Studies Pvt. Ltd. http://www.communityledtotalsanitation.org/sites/communityledtotalsanitation.org/files/FINAL_CLTS_Report.doc [accessed 10 October 2016].

Kalimuthu A. 2008. Chapter 13: Sustainable Community Owned Total Sanitation. In: Beyond Construction: Use by All. Wickens J , Verhagen J , Sijbesma C , Da Silva C , Ryan P , eds. London and Delft, UK:WaterAid and IRC International Water and Sanitation Centre, 169–180. http://www.ircwash.org/resources/beyond-construction-use-all [accessed 10 January 2016].

Kappauf L. 2011. Opportunities and Constraints for more Sustainable Sanitation through Sanitation Marketing in Malawi: Case study from Mzimba and Lilongwe Districts [MS Thesis]. 8: 137. Loughborough, UK:Water, Engineering, and Development Centre, Loughborough University. http://wedc.lboro.ac.uk/resources/pubs/KAPPAUF_Leonie_-_Opportunities_and_Constraints_for_more_Sustainable_Sanitation_through_Sanitation_Marketing_in_Malawi.pdf [accessed 1 June 2016].

Kar K, Bongartz P. 2006. Update on Some Recent Developments in CLTS. Brighton, UK:Institute of Development Studies, University of Sussex. http://www.communityledtotalsanitation.org/sites/communityledtotalsanitation.org/files/CLTS_update_for_WP257_0.pdf [accessed 10 October 2016].

Kar K, Chambers R. 2008. Handbook on Community-Led Total Sanitation. Brighton, UK:Institute of Development Studies, University of Sussex. http://www.communityledtotalsanitation.org/resource/handbook-community-led-total-sanitation [acccessed 1 December 2015].

Kennedy M , Meek A . 2013. Extension agent reorganization into a ‘block’ system for CLTS implementation and monitoring in Salima and Zomba Districts, Malawi. Toronto, ON, Canada:Engineers Without Borders. https://www.ircwash.org/sites/default/files/kennedy_and_meek_block_system_for_clts_implementation_and_monitoring.pdf [accessed 15 December 2015].

Khanna T, Das M. 2016. Why gender matters in the solution towards safe sanitation? Reflections from rural India. Glob Public Health 11(10):1185–1201, PMID: 26278418, 10.1080/17441692.2015.1062905.

Kulkarni S, O’Reilly K. 2014. Sanitation Vulnerability: Women’s Stress and Struggles for Violence-free Sanitation. https://www.gov.uk/dfid-research-outputs/sanitation-vulnerability-women-s-stress-and-struggles-for-violence-free-sanitation [accessed 1 November 2016].

Kunthy S, Catalla RNF. 2009. 2009 Cambodia: Evaluation of Community-Led Total Sanitation (CLTS). Draft Final Evaluation Report. New York, NY:UNICEF. https://www.unicef.org/evaldatabase/index_57963.html [accessed 10 January 2016].

Loevinsohn BP. 1990. Health education Interventions in developing countries: a methodological review of published articles. Int J Epidemiol 19(4):788–794, PMID: 2084004, 10.1093/ije/19.4.788.

Magala JM, Roberts L. 2009. 2009 Ghana: 2nd Evaluation of Strategy For Scaling Up Community Led Total Sanitation in Ghana. Final Report. New York, NY:UNICEF. https://www.unicef.org/evaldatabase/index_67844.html [accessed 10 January 2016].

Makotsi N, Kaseje D, Mumma J, Opiyo J, Lukorito L. 2016. Association of community led total sanitation to reduced household morbidity in Nyando District. Int J Sci Basic Appl Res 28(1):220–230. https://gssrr.org/index.php?journal=JournalOfBasicAndApplied&page=article&op=view&path%5B%5D=5825 [accessed 15 March 2017].

Morris-Iveson L, Siantumbu B. 2011. Community Led Total Sanitation in Zambia: An Evaluation of Experiences and Approaches to Date. Zambia:Government of the Republic of Zambia and UNICEF. https://www.unicef.org/evaldatabase/index_66788.html [accessed 10 January 2016].

Mukherjee N, Robiarto A, Saputra E, Wartono D. 2012. Achieving and Sustaining Open Defecation Free Communities: Learning from East Java. Washington, DC:The World Bank Water and Sanitation Program. https://www.wsp.org/sites/wsp.org/files/publications/WSP_Indonesia_Action_Research_Report.pdf [accessed 1 June 2016].

Murigi P, Stevens L, Mwanzia P, Pasteur K. 2015. Lessons learnt in urban community-led total sanitation from Nakuru, Kenya. Proceedings of the 38th WEDC International Conference. Loughborough, UK:Water, Engineering, and Development Centre. https://wedc-knowledge.lboro.ac.uk/details.html?id=22213 [accessed 15 March 2017].

Musyoki S. 2010. Scaling up CLTS in Kenya: opportunities, challenges and lessons. In: Participatory Learning and Action 61: Tales of Shit: Community-Led Total Sanitation in Africa. Bongartz P , Musyoki SM , Milligan A , Ashley H , eds. London, UK:IEED, 149–156. https://www.iied.org/pla-61-tales-shit-community-led-total-sanitation-africa [accessed 10 October 2016].

Mwanzia P, Misati W. 2013. Urban community-led total sanitation: a case study of Nakuru. Briefing Paper 1813. Proceedings of the 36th WEDC International Conference. Loughborough, UK:Water, Engineering, and Development Centre, Loughborough University. https://wedc-knowledge.lboro.ac.uk/details.html?id=20797 [accessed 10 October 2016].

Myers, J. 2016. Plan Netherlands’ Experience of Using a CLTS Approach in Urban Environments. Amsterdam: Plan Nederland. http://www.communityledtotalsanitation.org/sites/communityledtotalsanitation.org/files/Urban_CLTS_Plan.pdf [accessed 15 March 2017].

Papafilippou N, Templeton MR, Ali M. 2011. Technical note: is there a role for external technical support in the Community-Led Total Sanitation (CLTS) approach?. Int Dev Plan Rev 33(1):81–94, 10.3828/idpr.2011.5.

Pardeshi G, Shirke A, Jagtap M. 2008. SWOT analysis of total sanitation campaign in Yavatmal district of Maharashtra. Indian J Community Med 33(4):255–259, PMID: 19876501, 10.4103/0970-0218.43233.

Patil SR, Arnold BF, Salvatore AL, Briceno B, Ganguly S, Colford JM Jr, et al. 2014. The effect of India’s total sanitation campaign on defecation behaviors and child health in rural Madhya Pradesh: a cluster randomized controlled trial. PLoS Med 11(8):e1001709, PMID: 25157929, 10.1371/journal.pmed.1001709.

Pattanayak SK, Yang JC, Dickinson KL, Poulos C, Patil SR, Mallick RK, et al. 2009. Shame or subsidy revisited: social mobilization for sanitation in Orissa, India. Bull World Health Organ 87(8):580–587, PMID: 19705007, 10.2471/BLT.08.057422.

Pickering AJ, Djebbari H, Lopez C, Coulibaly M, Alzua ML. 2015. Effect of a community-led sanitation intervention on child diarrhoea and child growth in rural Mali: a cluster-randomised controlled trial. Lancet Glob. Heal 3(11):e701–e711, 10.1016/S2214-109X(15)00144-8.

Pluye P, Robert E, Cargo M, Bartlett G, O’Cathain A, Griffiths F, et al. 2011. Proposal: A Mixed Methods Appraisal Tool for Systematic Mixed Studies Reviews. Montréal, Canada:McGill University. http://mixedmethodsappraisaltoolpublic.pbworks.com [accessed 15 October 2016].

Prabhakaran P, Kar K, Mehta L, Chowdhury SR. 2016. Impact of Community-led Total Sanitation on women’s health in urban slums: a case study from Kalyani municipality. Evidence Report No 194. Sussex, UK and Kolkata, India:Institute for Development Studies (IDS) and CLTS Foundation. http://www.ids.ac.uk/publication/impact-of-community-led-total-sanitation-on-women-s-health-in-urban-slums-a-case-study-from-kalyani-municipality [accessed 15 March 2017].

Priyono E. 2009. Institutional Dimensions of Scaling up of CLTS in Indonesia. Bekasi, Indonesia:AKADEMIKA-Center for Public Policy Analysis. http://waterwiki.net/images/0/0c/InstitutionalDimensionsScalingUpCLTSIndonesia.pdf [accessed 1 June 2016].

Puzzolo E, Stanistreet D, Pope D, Bruce N, Rehfuess E. 2013. Systematic review: Factors influencing the large-scale uptake by households of cleaner and more efficient household energy technologies. London UK:EPPI-Centre, Social Science Research Unit, Institute of Education, University of London. http://eppi.ioe.ac.uk/cms/Default.aspx?tabid=3426 [accessed 15 October 2016].

Rehfuess EA, Puzzolo E, Stanistreet D, Pope D, Bruce NG. 2014. Enablers and barriers to large-scale uptake of improved solid fuel stoves: a systematic review. Environ Health Perspect 122(2):120–130, PMID: 24300100, 10.1289/ehp.1306639.

Robinson A. 2016. Final Evaluation: Pan African CLTS program 2010–2015. Amsterdam, Netherlands:Plan Netherlands. http://www.communityledtotalsanitation.org/resource/final-evaluation-plans-pan-africa-programme [accessed 15 March 2017].

Schlegelmilch MP, Lakhani A, Saunders LD, Jhangri GS. 2016. Evaluation of water, sanitation and hygiene program outcomes shows knowledge-behavior gaps in Coast Province, Kenya. Pan Afr Med J 23:145, PMID: 27279970, 10.11604/pamj.2016.23.145.7546.

Schmidt W-P. 2014. The elusive effect of water and sanitation on the global burden of disease. Trop Med Int Health 19(5):522–527, PMID: 24576060, 10.1111/tmi.12286.

Sclar GD, Penakalapati G, Amato HK, Garn JV, Alexander K, Freeman MC, et al. 2016. Assessing the impact of sanitation on indicators of fecal exposure along principal transmission pathways: a systematic review. Int J Hyg Environ Health 219(8):709–723, PMID: 27720133, 10.1016/j.ijheh.2016.09.021.

SEED Madagascar. 2016. Learning from rural CLTS: SEED Madagascar’s experiences in south east Madagascar. Antananarivo, Madagascar:SEED Madagascar and ONG Azafady, http://www.communityledtotalsanitation.org/sites/communityledtotalsanitation.org/files/SEED_Madagas-car_Lessons_learned_rural_CLTS.pdf [accessed 15 October 2016].

Shayamal S , Kashem MA , Rafi SM. 2008. Moving up the sanitation ladder: A participatory study of the drivers of sustainability and progress in Community Led Total Sanitation. In: Beyond Construction: Use by All. Wickens J , Verhagen J , Sijbesma C , Da Silva C , Ryan P , eds. London and Delft, UK:WaterAid and IRC International Water and Sanitation Centre, 263–280. http://www.ircwash.org/resources/beyond-construction-use-all [accessed 10 January 2016].

Sigler R, Mahmoudi L, Graham JP. 2015. Analysis of behavioral change techniques in community-led total sanitation programs. Health Promot Int 30(1):16–28, PMID: 25209916, 10.1093/heapro/dau073.

Singh S , Balfour N . 2015. Micro-Planning for CLTS: Experience from Kenya. WASH Field Note. New York, NY:UNICEF. https://www.unicef.org/esaro/WASH-Field-Microplanning-low-res.pdf [accessed 15 March 2017].

SNV Uganda. 2014. Documenting Our Process: Facilitating Community-Driven Development. Kampala, Uganda:SNV Uganda.

Spears D, Ghosh A, Cumming O. 2013. Open defecation and childhood stunting in India: an ecological analysis of new data from 112 districts. PLoS ONE 8(9):e73784, PMID: 24066070, 10.1371/journal.pone.0073784.

Spencer L, Ritchie J, Lewis J, Dillon L. 2003. Quality in Qualitative Evaluation: A Framework for Assessing Research Evidence. London, UK:National Center for Social Research. http://dera.ioe.ac.uk/21069/2/a-quality-framework-tcm6-38740.pdf [accessed 15 October 2016].

Taylor DL, Kahawita TM, Cairncross S, Ensink JHJ. 2015. The impact of water, sanitation and hygiene interventions to control cholera: a systematic review. PLoS One 10(8):e0135676, PMID: 26284367, 10.1371/journal.pone.0135676.

Thomas A, Bevan J. 2014. Developing and Monitoring Protocol for the Elimination of Open Defecation in Sub-Saharan Africa. http://www.communityledtotalsanitation.org/sites/communityledtotalsanitation.org/files/Thomas_and_Bevan_Elimination_of_open_defecation_SSA.pdf [accessed 15 March 2017].

Thomas BH, Ciliska D, Dobbins M, Micucci S. 2004. A process for systematically reviewing the literature: providing the research evidence for public health nursing interventions. Worldviews Evid Based Nurs 1(3):176–184, PMID: 17163895, 10.1111/j.1524-475X.2004.04006.x.

Tiwari C. 2011. Role of District Level Reflection and Government Leadership in Scaling Up CLTS: Lessons on Process Monitoring in Kenya. Netherlands:SNV Netherlands Development Organisation.

Toft E, Onabolu B. 2012. Evaluation of the Government of Ghana-UNICEF Integrated Approach to Guinea Worm Eradication through Water Supply, Sanitation and Hygiene in Northern Region, Ghana (I-WASH). Final Report. New York, NY:UNICEF. https://www.unicef.org/evaldatabase/files/I-WASH_evaluation_report_2012_-_Final_31102012.pdf [accessed 10 January 2016].

Tremolet S, Kolsky P, Perez E. 2010. Financing On-Site Sanitation for the Poor: A Six Country Comparative Review and Analysis. Washington, DC:The World Bank Water and Sanitation Program. http://documents.worldbank.org/curated/en/165231468341112439/Financing-on-site-sanitation-for-the-poor-a-six-country-comparative-review-and-analysis [accessed 15 December 2015].

Tyndale-Biscoe P, Bond M, Kidd R. 2013. ODF Sustainability Study. London:Plan International. http://www.communityledtotalsanitation.org/sites/communityledtotalsanitation.org/files/Plan_International_ODF_Sustainability_Study.pdf [accessed 10 October 2016].

UNICEF (United Nations Children’s Fund). 2014. Evaluation of the “Community Led Total Sanitation” (CLTS) Approach. Final Report. Madagascar:UNICEF. http://wsscc.org/wp-content/uploads/2016/04/Fonds-Mondial-pour-l%E2%80%99Assainissement-Rapport-d%E2%80%99avancement-2014-WSSCC.pdf [accessed 15 December 2015].

UNICEF WCARO. 2011. Roll-Out Evaluation of “Community Led Total Sanitation” in West and Central Africa. Final Report. Dakar, Senegal:UNICEF. https://www.unicef.org/evaluation/files/WCARO_2010-005_CLTS_Final_Report_03_11.pdf [accessed 10 January 2016].

UNICEF, WHO (World Health Organization). 2015. 25 Years Progress on Sanitation and Drinking Water: 2015 Update and MDG Assessment. Geneva, Switzerland:WHO Press. files.unicef.org/publications/files/Progress_on_Sanitation_and_Drinking_Water_2015_Update_.pdf [accessed 26 February 2017].

UNICEF. 2016. Phased Approach to Total Sanitation (PhATS) in Haiyan-Affected Areas – End-line Assessment. New York, NY:UNICEF. http://www.reachresourcecentre.info/system/files/resource-documents/reach_phl_report_phats_endline_report_march_2016_0.pdf [accessed 15 March 2017].

United Nations General Assembly. 2015. Draft Resolution a/69/l.85: Transforming Our World: The 2030 Agenda for Sustainable Development. Gen. Assem. 70 Sess. 16301. New York:United Nations. 10.1007/s13398-014-0173-7.2.

USAID (U.S. Agency for International Development). 2014. Ghana WASH Project – Lessons Learned: Hybrid CLTS Approach to Improving Sanitation. Washington, DC:USAID. http://pdf.usaid.gov/pdf_docs/PA00JR3J.pdf [accessed 10 October 2016].

Venkataramanan V. 2016. CLTS Learning Series: Lessons from CLTS Implementation in Seven Countries. Chapel Hill, NC:The University of North Carolina at Chapel Hill. https://waterinstitute.unc.edu/files/2016/01/CLTS-Learning-Series-Final-Report_011416.pdf [accessed 1 March 2016].

Vyas S, Kov P, Smets S, Spears D. 2016. Disease externalities and net nutrition: Evidence from changes in sanitation and child height in Cambodia, 2005–2010. Econ Hum Biol 23:235–245, PMID: 27776300, 10.1016/j.ehb.2016.10.002.

WaterAid India. 2008. Feeling the Pulse: A Study of the Total Sanitation Campaign in Five States. New Delhi, India:India. http://www.wateraid.org/~/media/Publications/study-total-sanitation-campaign-india.pdf [accessed 1 February 2016].

WHO. 1997. Participatory Hygiene and Sanitation Transformation: a new approach to working with communities. Geneva, Switzerland:WHO. http://apps.who.int/iris/bitstream/10665/63260/1/WHO_EOS_96.11.pdf [accessed 1 February 2017].

Wolf J, Prüss-Ustün A, Cumming O, Bartram J, Bonjour S, Cairncross S, et al. 2014. Systematic review: Assessing the impact of drinking water and sanitation on diarrhoeal disease in low- and middle-income settings: systematic review and meta-regression. Trop Med Int Health 19(8):928–942, PMID: 24811732, 10.1111/tmi.12331.

Zombo MM. 2010. Walking down the forbidden lane: ‘shit talk’ promotes sanitation. In: Participatory Learning and Action 61: Tales of Shit: Community-Led Total Sanitation in Africa. Bongartz P , Musyoki SM , Milligan A , Ashley H , eds. London, UK:IEED, 73–80. https://www.iied.org/pla-61-tales-shit-community-led-total-sanitation-africa [accessed 10 October 2016].

Prenatal Exposure to Phthalates and the Development of Eczema Phenotypes in Male Children: Results from the EDEN Mother–Child Cohort Study

Author Affiliations open

1Department of Epidemiology of Allergic and Respiratory Diseases (EPAR), Saint-Antoine Medical School, Pierre Louis Institute of Epidemiology and Public Health (IPLESP UMRS 1136), Institut national de la santé et de la recherche médicale (Inserm), UPMC Université Paris 06, Sorbonne Universités, Paris, France

2Department of community Medicine, Al Nafees Medical College and Hospital, Isra University–Islamabad Campus, Islamabad, Pakistan

3Team of Environmental Epidemiology applied to Reproduction and Respiratory Health, Institute for Advanced Biosciences (IAB), Inserm U1209, Université Grenoble Alpes, Grenoble, France

4Department of Health Sciences, Karlstad University, Karlstad, Sweden

5Department of Preventive Medicine, Icahn School of Medicine at Mount Sinai, New York, New York, USA

PDF icon PDF Version (270 KB)

  • Background:
    Contradictory results exist regarding the importance of early-life exposure to phthalates for development of childhood eczema.
    Objectives:
    We evaluated the association between maternal urinary concentrations of phthalate metabolites between the 24th and 28th week of gestation and occurrence of eczema in their sons up to 5 y of age, according to allergic sensitization as assessed by total immunoglobulin E (IgE) in a subsample of individuals.
    Methods:
    Data on health outcomes and background factors were collected using five standardized annual questionnaires completed by parents at the children’s ages of 1–5 y, and their associations with phthalate metabolite urinary concentrations were assessed in 604 mother–son pairs with adjusted multiple logistic regression and Cox’s survival model. Several eczema phenotypes were considered. Atopic status was assessed at 5 y of age in 293 boys through total IgE assessment.
    Results:
    At 5 y of age, the prevalence of ever eczema was 30.4%. Metabolites of di-isobutyl phthalate (DiBP) and di-isononyl phthalate (DiNP) were positively associated with early-onset (0–24 mo of age) eczema (15.7%) and late-onset (24–60 mo of age) eczema (14.7%). Applying the Cox’s model showed a significant association of occurrence of eczema in the first 5 y of life with DiBP and DiNP metabolites. Among IgE-sensitized boys, metabolites of di-n-butyl phthalate (DBP) and DiBP were significantly associated with ever eczema {hazard ratio HR=1.67 [95% confidence interval (CI): 1.10, 2.54], p=0.01 and HR=1.87 (95% CI: 1.01, 3.48), p=0.04, respectively}.
    Conclusions:
    Occurrence of eczema in early childhood may be influenced by prenatal exposure to certain phthalates in boys. Further investigations are needed to confirm this observation. https://doi.org/10.1289/EHP1829
  • Received: 1 March 2017
    Revised: 21 December 2017
    Accepted: 21 December 2017
    Published: 2 February 2018

    Address correspondence to M.H. Soomro, Epidemiology of Allergic and Respiratory Diseases Department (EPAR), Pierre Louis Institute of Epidemiology and Public Health (IPLESP UMRS 1136), UPMC and INSERM, Saint-Antoine Medical School, 27 rue Chaligny, 75571 Paris CEDEX 12, France. Telephone: 33 144738657. Email: munawar.soomro@iplesp.upmc.fr

    Supplemental Material is available online (https://doi.org/10.1289/EHP1829).

    The authors declare they have no actual or potential competing financial interests.

    Note to readers with disabilities: EHP strives to ensure that all journal content is accessible to all readers. However, some figures and Supplemental Material published in EHP articles may not conform to 508 standards due to the complexity of the information being presented. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.

  • PDF icon Supplemental Material PDF (308 KB)


    Note to readers with disabilities: EHP has provided a 508-conformant table of contents summarizing the Supplemental Material for this article (see below) so readers with disabilities may determine whether they wish to access the full, nonconformant Supplemental Material. If you need assistance accessing journal content, please contact ehponline@niehs.nih.gov. Our staff will work with you to assess and meet your accessibility needs within 3 working days.
    PDF icon Supplemental Table of Contents PDF (114 KB)

Introduction

Phthalates are man-made chemicals widely used in our daily life [e.g., food wrappings, cosmetic products, children’s toys, cleaning products, medical devices, pharmaceuticals, polyvinyl chloride (PVC) flooring, building materials] because of their properties of transparency, durability, and flexibility (Meeker et al. 2009; Bornehag et al. 2005; Arbuckle et al. 2014). People can be exposed to phthalates through inhalation, dermal absorption (including air-to-skin transport), contact with contaminated surfaces, and use of personal care products and ingestion (including dietary ingestion and incidental ingestion) (Swan 2008). Urinary concentrations of phthalate metabolite are the most reliable measures for exposure assessment (Swan 2008). The ubiquitous presence of phthalates in the environment and the potential consequences of human exposure to phthalates have raised concerns, particularly in vulnerable populations such as pregnant women and infants. Prenatal and early life are critical in the development of the immune system; exposure to toxic pollutants during this period can result in an increased risk of adverse health outcomes later in life.

Animal as well as epidemiological studies in infants and children have found various health effects of specific phthalates, including abnormal reproductive outcomes (Meeker et al. 2009; Abdel-Maksoud et al. 2015; Niermann et al. 2015; Aydoğan Ahbab and Barlas 2015; Jurewicz and Hanke 2011), children’s neurodevelopmental and behavioral problems (Whyatt et al. 2012; Ejaredar et al. 2015; Braun et al. 2013), and asthma and allergies (Bornehag et al. 2004b; Whyatt et al. 2014; Stelmach et al. 2015; Just et al. 2012; Hsu et al. 2012), although not consistently.

The prevalence of eczema in childhood, including its allergic-related phenotype, varies globally from 3% to 37% and is steadily increasing worldwide (Asher et al. 2006; Nutten 2015; Eichenfield et al. 2014; Deckers et al. 2012). The causes of eczema and of its increased prevalence are still unclear but are likely to be multifactorial in nature and depend on both genetic and environmental factors (Nutten 2015; Pyun 2015). During in utero life, the skin of the fetus may be exposed to various products absorbed by the mothers, including phthalates given that some phthalates have been detected in amniotic fluid (Jensen et al. 2015). Later in life, children’s dermal exposure can occur through the use of emollients, personal care products, and dermal contact with plastic products, soil, and dust, which can add to the total intake of certain phthalates through other routes such as inhalation and ingestion (Wormuth et al. 2006; Overgaard et al. 2017).

Eczema pathogenesis involves immunologic dysfunction and skin barrier defects (Pyun 2015). Phthalates might influence the epidermal barrier development processes, which can take several years to occur, and this influence could be stronger in allergic individuals due to the fragility of their mucosae. Although dermal absorption of phthalates and its impact on child health is established, contradictory results exist regarding eczema development after phthalate exposure in utero (Just et al. 2012; Smit et al. 2015; Weschler et al. 2015; Gong et al. 2015; Pan et al. 2014).

The objective of the present study was to evaluate the potential associations between maternal exposure to phthalates during the second trimester (between the 24th and 28th weeks) of pregnancy and occurrence of eczema phenotypes in male children in their first 5 y of life using data collected in the French EDEN mother–child prospective cohort study. The association was explored according to atopic status as defined by total immunoglobulin E (IgE) in a subsample of boys for which this assessment was available.

Methods

Study Population and Data Collection

The population in this study is a subgroup of the French EDEN (Etude des Déterminants pré et post natals du développement de la santé de l’Enfant) mother–child prospective birth cohort restricted to boys (https://eden.vjf.inserm.fr). The mothers’ recruitment and follow-up procedures and methodological assumptions have been published in earlier studies (Heude et al. 2015; Philippat et al. 2012). Briefly, women were given an appointment with a study midwife between the 24th and 28th gestational weeks, during which an interview on lifestyle factors was conducted and biological samples were collected. The information on the mothers and their newborns, including parity, mode of delivery, birth weight, gestational age, season of birth, behavioral, and environmental data were collected by the use of questionnaires, interviews, and obstetric and pediatric records. Boys were followed from birth to 5 y of age. The present study includes all male offspring for whom maternal urine samples had been analyzed for metabolites of phthalates as a part of previous studies (Chevrier et al. 2012; Mortamais et al. 2012; Botton et al. 2016), which explains why we had only boys in the study. Restricting to one sex could be an advantage in the context where sex-specific effects of exposures are expected (Harley et al. 2013; Wolff et al. 2008); a study restricted to one sex likely has a higher statistical power than a study including both.

Urine Collection and Analysis

Women were invited for a clinical examination and to provide a first morning urine at home before the hospital study visit between the 24th and 28th gestational weeks (second trimester of the pregnancy). If forgotten, the urine sample was collected at the hospital. A polypropylene container [FP40VPS; Centre Européen de Biotechnologie SA (CEB), Angers, France] was used. The urine samples were aliquoted and then frozen at −80°C. The analyses of the samples were conducted at the National Center for Environment Health Laboratory at the Centers for Disease Control and Prevention (CDC) in Atlanta, Georgia (USA) for 11 metabolites of 8 phthalates: diethyl phthalate (DEP), di-n-butyl phthalate (DBP), di-isobutyl phthalate (diBP), di(2-ethylhexyl) phthalate (DEHP), benzylbutyl phthalate (BBzP), di-isononyl phthalate (diNP), di-n-octyl phthalate (DOP), and di-isodecyl phthalate (DiDP).

The targeted phthalate metabolites were monoethyl phthalate (MEP), mono-n-butyl phthalate (MBP), mono-isobutyl phthalate (MiBP), mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono(2-ethyl-5-oxohexyl) phthalate (MEOHP), mono(2-ethylhexyl) phthalate (MEHP), monobenzyl phthalate (MBzP), monocarboxy-isooctyl phthalate (MCOP), mono(3-carboxypropyl) phthalate (MCPP), and monocarboxy-isononyl phthalate (MCNP).

Quantification of the phthalate metabolites concentrations was performed using on-line solid phase extraction–high-performance liquid chromatography–electrospray ionization isotope dilution–tandem mass spectrometry (Silva et al. 2007). Creatinine concentrations were also quantified following the same analytical procedure described previously (Mortamais et al. 2012). The concentrations of metabolites below the limit of detection (LOD) were replaced by instrumental reading values or by the compound-specific lowest instrumental reading value divided by the square root of 2 when the instrumental value was missing (Hornung and Reed 1990). The vast majority of the samples/metabolites (>97%) were above LOD. Two-step standardization was performed to reduce variability in the urinary concentrations of biomarkers due to sampling conditions, as described previously (Mortamais et al. 2012). There was no significant difference in the concentrations of phthalates in the urine samples collected at homes compared with those collected in the hospital.

Eczema Phenotypes

Eczema was assessed by administration of an enriched version of the International Study of Asthma and Allergies in Childhood (ISAAC) standardized and validated questionnaire (ISAAC Steering Committee 1998; Williams et al. 1999) to the parents at the children’s ages of 1, 2, 3, 4, and 5 y by appointment. At each year of follow-up, eczema was defined as eczema diagnosed by a doctor in the last 12 mo. Ever eczema was then defined as having ever reported a diagnosis of eczema at 1, 2, 3, 4, or 5 y of age. Additionally, two eczema phenotypes were distinguished: early-onset eczema if eczema was reported on any questionnaire in the first 2 y of life, and late-onset eczema if it was reported only between 24 and 60 mo of age.

IgE and Atopic Status

Total serum IgE was assessed in 293 boys at 5 y of age by using the CAP assay according to the manufacturers’ instructions (Pharmacia CAP System™, Pharmacia and Upjohn Diagnostics AB, Uppsala, Sweden). Boys were classified as atopic if they had IgE levels of ≥60 IU/mL (Baiz et al. 2017).

Other Variables

Information on potential confounders related to the children’s eczema included birth weight, gestational age (two-class variable: <37 gestational week, ≥37 gestational week), season of birth, parity, number of siblings (0, 1, 2, ≥3), exclusive breast-feeding for ≥4 mo (yes/no), maternal age at delivery (<25 y, 25–29 y, 30–34 y, ≥35 y), pre-pregnancy maternal body mass index (BMI, <18.5, 18.5–24.9, 25.0–30.0, >30 kg/m2), maternal and paternal history of allergies, maternal and paternal educational level (<2 y after high school, high school+2 y, ≥high school +3 y), household income, city of residence (Nancy/Poitiers), mode of delivery (vaginal, cesarean section), smoking during pregnancy (yes/no), maternal alcohol use during pregnancy (yes/no), maternal and paternal physician-diagnosed cases of asthma, rhinitis, eczema, and food allergies (yes/no).

Statistical Analysis

Urinary concentrations of phthalate metabolites were log-transformed because of nonnormality of the distribution. For the purpose of present analysis, MECPP, MEHHP, MEOHP, and MEHP were grouped as the sum of ΣDEHP (DEHP being their parent compound). We calculated the molar sum of these DEHP metabolites (nanomoles per milliliter) by dividing each metabolite concentration by its molecular weight and then summing the individual concentrations.

The prevalence and the incidence of eczema were computed. We applied multiple logistic regression analysis to examine the association between each phthalate metabolite and ΣDEHP and eczema outcomes adjusted for potential confounders. The potential confounders were determined a priori from literature review. Then we used Directed Acyclic Graph (http://www.dagitty.net/dags.html) to show the hypothesis relations between biomarkers concentration, confounders, and eczema outcomes (see Figure S1). The confounders included in our final model were parental asthma/rhinitis/eczema, maternal smoking, maternal age, maternal BMI, maternal education level, gestational age, and number of siblings. Models were additionally adjusted for recruitment center.

We also applied the Cox proportional hazard discrete time survival model to estimate the association between in utero exposure to phthalates and eczema occurrence in the first 5 y of life after adjustment for the confounding factors. Results were further stratified in order to see whether phthalate biomarkers were associated with atopic (presence of allergic sensitization according to IgE) or nonatopic (absence of allergic sensitization to IgE) eczema. We also took IgE as an additional variable to see its association with phthalate biomarkers by applying multiple logistic regression analysis. Because our statistical analysis involved multiple comparisons so that it became increasingly likely that at least one comparison would have been statistically different due to random sampling error alone, the multiple testing hypothesis was considered using the Bonferroni test. All the statistical analysis was conducted using Stata statistical software (version 14; Stata Corporation), where p<0.05 was considered statistically significant.

Ethics Approval

The study was approved by the relevant ethical committees (Comité Consultatif pour la Protection des Personnes dans la Recherche Biomédicale, Le Kremlin Bicêtre University Hospital, and Commission Nationale de l’Informatique et des Libertés), and all participating women gave informed written consent for their own participation, and both parents for a newborn child after delivery.

Results

Characteristics of the Population

There were 998 male offspring in the cohort, among whom 604 had phthalates assessment. The participants included in this study analysis did not differ in major demographic characteristics from the other EDEN boys (Table 1) except for maternal age. However, results did not change when adjusting for maternal age in the final model. Average maternal age was 29.9 y, 26.3% women were overweight or obese (BMI>25 kg/m2), and 85.8% did not smoke during pregnancy. There were 24 (4.0%) boys born before the 37th week of gestation, and 15 (3.0%) had a birth weight <2,500 g. Ninety-five (15.7%) mothers delivered through cesarean section. Urine was reported as being collected before 0800 hours 51.3% of the time, and between 0800 and 1200 hours 24.7% of the time. We also observed that during pregnancy 14.4% of the mothers had eczema, 10.9% had asthma, and 16.7% had rhinitis.

Table 1. Characteristics of the study population and comparison with the entire EDEN cohort of boys.
Characteristics All male EDEN cohort (n=998) Present study (n=604) p-Value
n % n %
Maternal age (y)
 <25 200 20.0 79 13.0
 25–29 378 37.9 218 36.1
 30–34 285 28.6 205 34.0
 ≥35 135 13.5 102 16.9 0.0006
Parity
 0 436 43.7 269 44.5
 1 374 37.5 222 36.8
 ≥2 187 18.7 112 18.6
 Missing 1 0.1 1 0.1 0.9
BMI (kg/m2)
 <18.5 92 9.2 57 9.4
 18.5–24.9 634 63.6 379 62.9
 25–30 171 17.1 110 18.3
 >30 80 8.0 49 8.1
 Missing 21 2.1 9 1.5 0.8
Maternal education
 <2 y after high school 459 46.0 263 43.5
 High school+2 y 219 21.9 132 21.9
 ≥High school +3 y 297 29.8 199 32.9
 Missing 23 2.3 10 1.7 0.4
Active smoking (cigarettes/day)
 0 827 82.9 518 86.0
 1–5 82 8.2 45 7.5
 >5 86 8.6 37 6.2
 Missing 3 0.3 2 0.3 0.2
Passive smoking
 Yes 278 27.9 192 31.8
 No 711 71.2 405 67.1
 Missing 9 0.9 7 1.1 0.2
Alcohol during pregnancy
 Yes 278 27.9 151 25.0
 No 714 71.5 450 74.5
 Missing 6 0.6 3 0.5 0.4
Center
 Poitiers 533 53.4 347 57.4
 Nancy 465 46.6 257 42.6 0.1
Gestational age
 <37 gestational week 63 6.3 24 4.0
 ≥37 gestational week 935 93.7 580 96.0 0.04
Birth weight
 <2,500 g 47 4.7 15 2.5
 ≥2,500 g 951 95.3 589 97.5 0.02
Year of birth
 2003–2004 560 56.1 330 54.6
 2005–2006 438 43.9 274 45.4 0.5
Breast-feeding
 Yes 720 72.2 432 71.5
 No 270 27.0 172 28.5
 Missing 8 0.8 0 0 0.07
Employment
 Yes 727 72.9 458 75.9
 No 262 26.2 142 23.5
 Missing 9 0.90 4 0.6 0.3
Cesarean section
 Yes 175 17.5 95 15.7
 No 822 82.4 509 84.3
 Missing 1 0.1 0 0 0.4
Marital status
 Married 534 53.6 329 54.5
 Unmarried/single 435 43.5 257 42.6
 Divorced 19 1.9 12 1.9
 Separated 2 0.2 2 0.3
 Widowed 2 0.2 1 0.2
 Missing 6 0.6 3 0.5 0.9
Season of urine sample collection
 Winter 262 26.2 189 31.3
 Spring 301 30.2 146 24.2
 Summer 249 24.9 129 21.3
 Fall 186 18.7 140 23.2 0.002
Time of sampling
 2400–0759 hours 500 50.1 310 51.3
 0800–1159 hours 272 27.3 149 24.7
 1200–1900 hours 7 0.7 5 0.8
 Missing 219 21.9 140 23.2 0.7
Maternal history of asthma
 Yes 110 11.0 66 10.9
 No 884 88.6 536 88.7
 Missing 4 0.4 2 0.3 0.9
Maternal history of eczema
 Yes 142 14.2 87 14.4
 No 852 85.4 515 85.3
 Missing 4 0.4 2 0.3 0.9
Maternal history of rhinitis
 Yes 167 16.7 101 16.7
 No 827 82.9 501 83.0
 Missing 4 0.4 2 0.3 0.9
Paternal history of asthma
 Yes 99 9.9 58 9.6
 No 895 89.7 544 90.1
 Missing 4 0.4 2 0.3 0.9
Paternal history of eczema
 Yes 48 4.8 24 4.0
 No 946 94.8 578 95.7
 Missing 4 0.4 2 0.3 0.7
Paternal history of rhinitis
 Yes 100 10.0 60 9.9
 No 894 89.6 542 89.8
 Missing 4 0.4 2 0.3 0.3

Note: BMI; body mass index.

Eczema Prevalence and Incidence

The prevalence of ever eczema was 9.7%, 15.7%, 21.0%, 26.6%, and 30.4% at 1, 2, 3, 4, and 5 y of age, respectively. The incidence decreased from 9.7% at 1 y of age to 5.2% at 5 y of age (Figure 1). The prevalence of early-onset and late-onset eczema were 15.7% and 14.7%, respectively.

 Bar graph plotting percentage incidence of eczema (y-axis) across age in years (x-axis).

Figure 1: The incidence of eczema with age of the boys. The y-axis shows the percentage, and the numbers above each bar indicate the number of boys diagnosed at each year of age.

Prenatal Exposure to Phthalates

Phthalate metabolites were detected in the urine of more than 97% of the mothers. MEP had the highest geometric mean (110.7 μg/L), whereas MCNP had the lowest (1.5 μg/L). The LOD ranged from 0.2 to 0.6 μg/L. A summary of detection frequency as well as geometric means and select percentiles of urinary concentrations of phthalate metabolites are outlined in Table 2.

Table 2. Distribution of phthalate metabolites concentration (μg/L) in maternal prenatal urine (n=604).
Parent compound Abbrev Phthalate metabolites (μg/L) Abbrev LOD (μg/L) %>LOD GM Minimum Percentile Maximum
5th 25th 50th 75th 95th
Diethyl phthalate DEP Monoethyl phthalate MEP 0.6 100 110.7 7.9 21 51 96 195 714 4093.4
Di-n-butyl phthalate DBP Mono-n-butyl phthalate MBP 0.2 100 51.6 2.8 12 28 43 73 438 2418.8
Di-isobutyl phthalate DiBP Mono-isobutyl phthalate MiBP 0.2 100 42.7 1.5 12 25 39 69 166 689.9
Di(2-ethylhexyl) phthalate DEHP Mono(2-ethyl-5-carboxypentyl) phthalate MECPP 0.2 100 40.5 2.6 12 25 38 62 160 2640.5
Di(2-ethylhexyl) phthalate DEHP Mono(2-ethyl-5-hydroxyhexyl) phthalate MEHHP 0.2 100 26.9 1.7 6.8 16 27 44 103 2103.8
Di(2-ethylhexyl) phthalate DEHP Mono(2-ethyl-5-oxohexyl) phthalate MEOHP 0.2 99.7 22.1 0.6 5.6 14 23 36 84 1441.4
Di(2-ethylhexyl) phthalate DEHP Mono(2-ethylhexyl) phthalate MEHP 0.5 98 7.2 0.1 1.6 4.4 7.7 15 35 292.3
Benzylbutyl phthalate BBzP Monobenzyl phthalate MBzP 0.3 100 19.7 1.6 4.5 11 18 33 104 970.4
Di-isononyl phthalate DiNP Monocarboxy-isooctyl phthalate MCOP 0.2 99 4.1 0.1 1.2 2.4 3.9 6.5 19 375.3
Di-n-octyl phthalate DOP Mono(3-carboxypropyl) phthalate MCPP 0.2 100 2.1 0.1 0.7 1.3 1.9 3.4 9.1 78.7
Di-isodecyl phthalate DiDP Monocarboxy-isononyl phthalate MCNP 0.2 97.5 1.5 0.1 0.5 0.8 1.2 2.2 9.5 686.8

Note: Abbrev, abbreviation; GM, geometric mean; LOD, limit of detection.

Spearman correlation coefficients between ln-transformed phthalate metabolites urinary concentrations are shown in Table S1. The majority of biomarkers were positively correlated except for the correlation between MEP and MiBP, MECPP, MEHHP, MCOP, MCPP, or MCNP. All the other biomarkers were significantly correlated with each other (with correlation coefficients ranging from r=0.08, p<0.05 to r=0.97, p<0.05). As expected, DEHP metabolites were highly correlated with each other (r≥0.77, p<0.05).

Associations between Phthalates Metabolites and Eczema Phenotypes

Table 3 presents the relationship of prenatal phthalate exposure (based on phthalate metabolite urinary concentrations) to the occurrence of ever eczema outcomes in boys between 1 and 5 y of age. Based on a multiple logistic regression model that was adjusted for confounding factors, we observed an increased risk of eczema at 1 y of age in association with increased MEP concentrations [odds ratio =1.69()OR; 95% confidence interval (CI): 1.05, 2.73; p<0.05] and at 1, 3, 4, and 5 y of age for MiBP.

Table 3. Associations {adjusted odds ratio and 95% confidence interval [OR (95% CI)]} between maternal urinary concentrations of phthalate metabolites and occurrence of eczema phenotypes in boys (n=604).
Year, eczema MEP MBP MiBP MECPP MEHHP MEOHP MEHP MBzP MCOP MCPP MCNP ΣDEHP
1 1.69