Statistical strategies for constructing health risk models with multiple pollutants and their interactions: possible choices and comparisons

Exposure to Environmental Pollutants and Their Association with Biomarkers of Aging: A Multipollutant Approach

Mitochondrial DNA (mtDNA) content and telomere length are putative aging biomarkers and are sensitive to environmental stressors, including pollutants. Our objective was to identify, from a set of environmental exposures, which exposure is associated with leukocyte mtDNA content and telomere length in adults. This study includes 175 adults from 50 to 65 years old from the cross-sectional Flemish Environment and Health study, of whom leukocyte telomere length and mtDNA content were determined using qPCR. The levels of exposure of seven metals, 11 organohalogens, and four perfluorinated compounds (PFHxS, PFNA, PFOA, PFOS) were measured. We performed sparse partial least-squares regression analyses followed by ordinary least-squares regression to assess the multipollutant associations. While accounting for possible confounders and coexposures, we identified that urinary cadmium (6.52%, 95% confidence interval, 1.06, 12.28), serum hexachlorobenzene (2.89%, 018, 5.68), and perfluorooctanesulfonic acid (11.38%, 5.97, 17.08) exposure were positively associated ( p < 0.05) with mtDNA content, while urinary copper (-9.88%, -14.82, -4.66) and serum perfluorohexanesulfonic acid (-4.75%, -8.79, -0.54) exposure were inversely associated with mtDNA content. Urinary antimony (2.69%, 0.45, 4.99) and mercury (1.91%, 0.42, 3.43) exposure were positively associated with leukocyte telomere length, while urinary copper (-3.52%, -6.60, -0.34) and serum perfluorooctanesulfonic acid (-3.64%, -6.60, -0.60) showed an inverse association. Our findings support the hypothesis that environmental pollutants interact with molecular hallmarks of aging.

Two-step approach for assessing the health effects of environmental chemical mixtures: application to simulated datasets and real data from the Navajo Birth Cohort Study

BACKGROUND

There is increasing interest in examining the consequences of simultaneous exposures to chemical mixtures. However, a consensus or recommendations on how to appropriately select the statistical approach analyzing the health effects of mixture exposures which best aligns with study goals has not been well established. We recognize the limitations that existing methods have in effectively reducing data dimension and detecting interaction effects when analyzing chemical mixture exposures collected in high dimensional datasets with varying degrees of variable intercorrelations. In this research, we aim to examine the performance of a two-step statistical approach in addressing the analytical challenges of chemical mixture exposures using two simulated data sets, and an existing data set from the Navajo Birth Cohort Study as a representative case study.

METHODS

We propose to use a two-step approach: a robust variable selection step using the random forest approach followed by adaptive lasso methods that incorporate both dimensionality reduction and quantification of the degree of association between the chemical exposures and the outcome of interest, including interaction terms. We compared the proposed method with other approaches including (1) single step adaptive lasso; and (2) two-step Classification and regression trees (CART) followed by adaptive lasso method.

RESULTS

Utilizing simulated data sets and applying the method to a real-life dataset from the Navajo Birth Cohort Study, we have demonstrated good performance of the proposed two-step approach. Results from the simulation datasets indicated the effectiveness of variable dimension reduction and reliable identification of a parsimonious model compared to other methods: single-step adaptive lasso or two-step CART followed by adaptive lasso method.

CONCLUSIONS

Our proposed two-step approach provides a robust way of analyzing the effects of high-throughput chemical mixture exposures on health outcomes by combining the strengths of variable selection and adaptive shrinkage strategies.

Selection of nonlinear interactions by a forward stepwise algorithm: Application to identifying environmental chemical mixtures affecting health outcomes

In this paper, we propose a stepwise forward selection algorithm for detecting the effects of a set of correlated exposures and their interactions on a health outcome of interest when the underlying relationship could potentially be nonlinear. Though the proposed method is very general, our application in this paper remains to be on analysis of multiple pollutants and their interactions. Simultaneous exposure to multiple environmental pollutants could affect human health in a multitude of complex ways. For understanding the health effects of multiple environmental exposures, it is often important to identify and estimate complex interactions among exposures. However, this issue becomes analytically challenging in the presence of potential nonlinearity in the outcome-exposure response surface and a set of correlated exposures. Through simulation studies and analyses of test datasets that were simulated as a part of a data challenge in multipollutant modeling organized by the National Institute of Environmental Health Sciences (http://www.niehs.nih.gov/about/events/pastmtg/2015/statistical/), we illustrate the advantages of our proposed method in comparison with existing alternative approaches. A particular strength of our method is that it demonstrates very low false positives across empirical studies. Our method is also used to analyze a dataset that was released from the Health Outcomes and Measurement of the Environment Study as a benchmark beta-tester dataset as a part of the same workshop.

A tensor product quasi-Poisson model for estimating health effects of multiple ambient pollutants on mortality

BACKGROUND

People are exposed to mixtures of highly correlated gaseous, liquid and solid pollutants. However, in previous studies, the assessment of air pollution effects was mainly based on single-pollutant models or was simultaneously included as multiple pollutants in a model. It is essential to develop appropriate methods to accurately estimate the health effects of multiple pollutants in the presence of a high correlation between pollutants.

METHODS

The flexible tensor product smooths of multiple pollutants was applied for the first time in a quasi-Poisson model to estimate the health effects of SO₂, NO₂ and PM₁₀ on daily all-cause deaths during 2005-2012 in Guangzhou, China. The results were compared with those from three other conventional models, including the single-pollutant model and the three-pollutant model with and without first-order interactions.

RESULTS

The tensor product model revealed a complex interaction among three pollutants and significant combined effects of PM₁₀, NO₂ and SO₂, which revealed a 2.53% (95%CI: 1.03-4.01%) increase in mortality associated with an interquartile-range (IQR) increase in the concentrations of all three pollutants. The combined effect estimated by the single-pollutant model was 5.63% (95% CI: 3.96-7.34%). Although the conventional three-pollutant models produced combined effect estimates (2.20, 95%CI, 1.18-3.23%; 2.78, 95%CI: 1.35-4.23%) similar to those of the tensor product model, they distorted the estimates and inflated the variances of the estimates when attributing the combined health effects to individual pollutants.

CONCLUSIONS

The single-pollutant model or conventional multi-pollutant model may yield misleading results in the presence of collinearity. The tensor product quasi-Poisson regression provides a novel approach to the assessment of the health impacts of multiple pollutants by flexibly fitting the interaction effects and avoiding the collinearity problem.

Early-life exposure to persistent organic pollutants (OCPs, PBDEs, PCBs, PFASs) and attention-deficit/hyperactivity disorder: A multi-pollutant analysis of a Norwegian birth cohort

BACKGROUND

Numerous ubiquitous environmental chemicals are established or suspected neurotoxicants, and infants are exposed to a mixture of these during the critical period of brain maturation. However, evidence for associations with the risk of attention-deficit/hyperactivity disorder (ADHD) is sparse. We investigated early-life chemical exposures in relation to ADHD.

METHODS

We used a birth cohort of 2606 Norwegian mother-child pairs enrolled 2002-2009 (HUMIS), and studied a subset of 1199 pairs oversampled for child neurodevelopmental outcomes. Concentrations of 27 persistent organic pollutants (14 polychlorinated biphenyls, 5 organochlorine pesticides, 6 brominated flame retardants, and 2 perfluoroalkyl substances) were measured in breast milk, reflecting the child's early-life exposures. We estimated postnatal exposures in the first 2 years of life using a pharmacokinetic model. Fifty-five children had a clinical diagnosis of ADHD (hyperkinetic disorder) by 2016, at a median age of 13 years. We used elastic net penalized logistic regression models to identify associations while adjusting for co-exposure confounding, and subsequently used multivariable logistic regression models to obtain effect estimates for the selected exposures.

RESULTS

Breast milk concentrations of perfluorooctane sulfonate (PFOS) and β‑hexachlorocyclohexane (β-HCH) were associated with increased odds of ADHD: odds ratio (OR) = 1.77, 95% confidence interval (CI): 1.16, 2.72 and OR = 1.75, 95% CI: 1.22, 2.53, per interquartile range increase in ln-transformed concentrations, respectively. Stronger associations were observed among girls than boys for PFOS (p_interaction = 0.025). p,p'‑Dichlorodiphenyltrichloroethane (p,p'-DDT) levels were associated with lower odds of ADHD (OR = 0.64, 95% CI: 0.42, 0.97). Hexachlorobenzene (HCB) had a non-linear association with ADHD, with increasing risk in the low-level exposure range that switched to a decreasing risk at concentrations above 8 ng/g lipid. Postnatal exposures showed similar results, whereas effect estimates for other chemicals were weaker and imprecise.

CONCLUSIONS

In a multi-pollutant analysis of four classes of chemicals, early-life exposure to β-HCH and PFOS was associated with increased risk of ADHD, with suggestion of sex-specific effects for PFOS. The unexpected inverse associations between p,p'-DDT and higher HCB levels and ADHD could be due to live birth bias; alternatively, results may be due to chance findings.

Co-exposure to environmental endocrine disruptors in the US population

Exposure to environmental endocrine disruptors (EEDs) has been linked to adverse health outcomes. The vast majority of studies examined one class of EEDs at a time but humans often are exposed to multiple EEDs at the same time. It is, therefore, important to know the co-exposure status of multiple EEDs in an individual, to preclude and control for potential confounding effects posed by co-exposed EEDs. This study examined the concentrations of seven classes of EEDs in the US population utilizing the data from the National Health and Nutrition Examination Survey (NHANES), 2009-2014 survey cycles. We applied linear correlation and cluster analysis to characterize the correlation profile and cluster patterns of these EEDs. We found that EEDs with a similar structure are often highly correlated. Among between-class correlations, mercury and perfluoroalkyl substances (PFAS) and cadmium and polycyclic aromatic hydrocarbons (PAHs) were two significantly correlated EEDs. In epidemiologic studies, measurement and control for co-exposure to pollutants, especially those with similar biological effects, are critical when attempting to make causal inferences. Appropriate statistical methods to handle within- and between-class correlations are needed.

Statistical Methodology in Studies of Prenatal Exposure to Mixtures of Endocrine-Disrupting Chemicals: A Review of Existing Approaches and New Alternatives

BACKGROUND

Prenatal exposures to endocrine-disrupting chemicals (EDCs) during critical developmental windows have been implicated in the etiologies of a wide array of adverse perinatal and pediatric outcomes. Epidemiological studies have concentrated on the health effects of individual chemicals, despite the understanding that EDCs act together via common mechanisms, that pregnant women are exposed to multiple EDCs simultaneously, and that substantial toxicological evidence of adverse developmental effects has been documented. There is a move toward multipollutant models in environmental epidemiology; however, there is no current consensus on appropriate statistical methods.

OBJECTIVES

We aimed to review the statistical methods used in these studies, to identify additional applicable methods, and to determine the strengths and weaknesses of each method for addressing the salient statistical and epidemiological challenges.

METHODS

We searched Embase, MEDLINE, and Web of Science for epidemiological studies of endocrine-sensitive outcomes in the children of mothers exposed to EDC mixtures during pregnancy and identified alternative statistical methods from the wider literature.

DISCUSSION

We identified 74 studies and analyzed the methods used to estimate mixture health effects, identify important mixture components, account for nonmonotonicity in exposure–response relationships, assess interactions, and identify windows of exposure susceptibility. We identified both frequentist and Bayesian methods that are robust to multicollinearity, performing shrinkage, variable selection, dimension reduction, statistical learning, or smoothing, including methods that were not used by the studies included in our review.

CONCLUSIONS

Compelling motivation exists for analyzing EDCs as mixtures, yet many studies make simplifying assumptions about EDC additivity, relative potency, and linearity, or overlook the potential for bias due to asymmetries in chemical persistence. We discuss the potential impacts of these choices and suggest alternative methods to improve analyses of prenatal exposure to EDC mixtures. https://doi.org/10.1289/EHP2207.

The Short-Term Effects of Ambient Air Pollutants on Childhood Asthma Hospitalization in Taiwan: A National Study

This investigation determined the effects of air pollution on childhood asthma hospitalization in regions with differing air pollution levels in Taiwan over a long time period. Data of childhood hospital admissions for asthma in patients aged 0–18 years and air quality in eight regions for the period 2001–2012 in Taiwan were collected. Poisson generalized linear regression analysis was employed to identify the relative risks of hospitalization due to asthma in children associated with exposure to varying levels of air pollutants with a change in the interquartile range after adjusting for temperature and relative humidity. Particulate matter ≤2.5 μm (PM_2.5), particulate matter ≤10 μm (PM₁₀), ozone (O₃), sulfur dioxide (SO₂), and nitrogen dioxide (NO₂), were positively associated with childhood asthma hospitalization, while O₃ was negatively associated with childhood asthma hospitalization. SO₂ was identified as the most significant risk factor. The relative risks for asthma hospitalization associated with air pollutants were higher among children aged 0–5 years than aged 6–18 years and were higher among males than females. The effects of air pollution on childhood asthma were greater in the higher-level air pollution regions, while no association was observed in the lower-level air pollution regions. These findings may prove important for policymakers involved in implementing policies to reduce air pollution.

The Exposome: Molecules to Populations

Derived from the term exposure, the exposome is an omic-scale characterization of the nongenetic drivers of health and disease. With the genome, it defines the phenome of an individual. The measurement of complex environmental factors that exert pressure on our health has not kept pace with genomics and historically has not provided a similar level of resolution. Emerging technologies make it possible to obtain detailed information on drugs, toxicants, pollutants, nutrients, and physical and psychological stressors on an omic scale. These forces can also be assessed at systems and network levels, providing a framework for advances in pharmacology and toxicology. The exposome paradigm can improve the analysis of drug interactions and detection of adverse effects of drugs and toxicants and provide data on biological responses to exposures. The comprehensive model can provide data at the individual level for precision medicine, group level for clinical trials, and population level for public health.

Prenatal exposure to phthalates and autism spectrum disorder in the MARBLES study

BACKGROUND

Evidence from experimental and observational studies suggests that prenatal phthalate exposures may be associated with autism spectrum disorder (ASD). We examined whether prenatal phthalate exposures were associated with an increased risk of ASD.

METHODS

We quantified 14 metabolites of eight phthalates in 636 multiple maternal urine samples collected during 2nd and 3rd trimesters of pregnancy from 201 mother-child pairs in MARBLES (Markers of Autism Risk in Babies - Learning Early Signs), a high-risk ASD longitudinal cohort. At 3 years old, children were clinically assessed for ASD and classified into three diagnostic categories: ASD (n = 46), non-typical development (Non-TD, n = 55), and typical development (TD, n = 100). We used multinomial logistic regression to evaluate the association of phthalate metabolite concentrations with ASD and Non-TD.

RESULTS

Most associations of phthalate biomarkers with both ASD and Non-TD were null, with the exception that monoethyl phthalate (MEP) was significantly associated with an increased risk of Non-TD (per 2.72-fold relative increase in concentration: Relative risk ratio (RRR) = 1.38; 95% confidence interval (CI): 1.01, 1.90). When stratified by prenatal vitamin use during the first month of pregnancy, among mothers who took vitamins, ASD risk was inversely associated with mono-isobutyl phthalate (MiBP, RRR = 0.44; 95% CI: 0.21, 0.88), mono(3-carboxypropyl) phthalate (MCPP, RRR = 0.41; 95% CI: 0.20, 0.83) and mono-carboxyisooctyl phthalate (MCOP, RRR = 0.49; 95% CI: 0.27, 0.88), but among mothers who did not take prenatal vitamins, Non-TD risk was positively associated with MCPP (RRR = 5.09; 95% CI: 2.05, 12.6), MCOP (RRR = 1.86; 95% CI: 1.01, 3.39), and mono-carboxyisononyl phthalate (MCNP, RRR = 3.67; 95% CI: 1.80, 7.48). When stratified by sex, among boys, MEP, monobenzyl phthalate, MCPP, MCNP, and sum of di(2-ethylhexyl) phthalate metabolites (ΣDEHP) were positively associated with Non-TD risk, but associations with ASD were null. Among girls, associations with both ASD and Non-TD were null.

CONCLUSIONS

Our study showed that phthalate exposures in mid- to late pregnancy were not associated with ASD in children from this high-risk ASD cohort. Further studies should be conducted in the general population without high-risk genes to confirm our findings.

Statistical software for analyzing the health effects of multiple concurrent exposures via Bayesian kernel machine regression

BACKGROUND

Estimating the health effects of multi-pollutant mixtures is of increasing interest in environmental epidemiology. Recently, a new approach for estimating the health effects of mixtures, Bayesian kernel machine regression (BKMR), has been developed. This method estimates the multivariable exposure-response function in a flexible and parsimonious way, conducts variable selection on the (potentially high-dimensional) vector of exposures, and allows for a grouped variable selection approach that can accommodate highly correlated exposures. However, the application of this novel method has been limited by a lack of available software, the need to derive interpretable output in a computationally efficient manner, and the inability to apply the method to non-continuous outcome variables.

METHODS

This paper addresses these limitations by (i) introducing an open-source software package in the R programming language, the bkmr R package, (ii) demonstrating methods for visualizing high-dimensional exposure-response functions, and for estimating scientifically relevant summaries, (iii) illustrating a probit regression implementation of BKMR for binary outcomes, and (iv) describing a fast version of BKMR that utilizes a Gaussian predictive process approach. All of the methods are illustrated using fully reproducible examples with the provided R code.

RESULTS

Applying the methods to a continuous outcome example illustrated the ability of the BKMR implementation to estimate the health effects of multi-pollutant mixtures in the context of a highly nonlinear, biologically-based dose-response function, and to estimate overall, single-exposure, and interactive health effects. The Gaussian predictive process method led to a substantial reduction in the runtime, without a major decrease in accuracy. In the setting of a larger number of exposures and a dichotomous outcome, the probit BKMR implementation was able to correctly identify the variables included in the exposure-response function and yielded interpretable quantities on the scale of a latent continuous outcome or on the scale of the outcome probability.

CONCLUSIONS

This newly developed software, integrated suite of tools, and extended methodology makes BKMR accessible for use across a broad range of epidemiological applications in which multiple risk factors have complex effects on health.

Assessing the cumulative health effect following short term exposure to multiple pollutants: An evaluation of methodological approaches using simulations and real data

BACKGROUND

Assessment of the cumulative effect of correlated exposures is an open methodological issue in environmental epidemiology. Most previous studies have applied regression models with interaction terms or dimension reduction methods. The combined effect of pollutants has been also evaluated through the use of exposure scores that incorporate weights based on the strength of the component-specific associations with health outcomes.

METHODS

We compared three approaches addressing multi-pollutant exposures in epidemiological models: main effects models, the adaptive least absolute shrinkage and selection operator (LASSO) and a weighted exposure score. We assessed the performance of the methods by simulations under various scenarios for the pollutants' correlations. We further applied these methods to time series data from Athens, Greece in 2007-12 to investigate the combined effect of short-term exposure to six regulated pollutants on all-cause and respiratory mortality.

RESULTS

The exposure score provided the least biased estimate under all correlation scenarios for both mortality outcomes. The adaptive LASSO performed well in the case of low and medium correlation between exposures while the main effect model resulted in severe bias. In the real data application, the cumulative effect estimate was similar between approaches for all-cause mortality ranging from 0.7% increase per interquartile range (IQR) (score) to 1.1% (main effects), while for respiratory mortality conclusions were contradictive and ranged from - 0.6% (adaptive LASSO) to 2.8% (score).

CONCLUSIONS

Τhe use of a weighted exposure score to address cumulative effects of correlated metrics may perform well under different exposure correlation and variability in the health outcomes.

Distributed Lag Interaction Models with Two Pollutants

Distributed lag models (DLMs) have been widely used in environmental epidemiology to quantify the lagged effects of air pollution on a health outcome of interest such as mortality and morbidity. Most previous DLM approaches only consider one pollutant at a time. In this article, we propose distributed lag interaction model (DLIM) to characterize the joint lagged effect of two pollutants. One natural way to model the interaction surface is by assuming that the underlying basis functions are tensor products of the basis functions that generate the main-effect distributed lag functions. We extend Tukey's one-degree-of-freedom interaction structure to the two-dimensional DLM context. We also consider shrinkage versions of the two to allow departure from the specified Tukey's interaction structure and achieve bias-variance tradeoff. We derive the marginal lag effects of one pollutant when the other pollutant is fixed at certain quantiles. In a simulation study, we show that the shrinkage methods have better average performance in terms of mean squared error (MSE) across different scenarios. We illustrate the proposed methods by using the National Morbidity, Mortality, and Air Pollution Study (NMMAPS) data to model the joint effects of PM₁₀ and O₃ on mortality count in Chicago, Illinois, from 1987 to 2000.

Exposure to polybrominated diphenyl ethers (PBDEs) and child behavior: Current findings and future directions

Polybrominated diphenyl ethers (PBDEs) are recognized neurotoxicants, but the extent to which PBDEs influence various domains of behavior in children is not fully understood. As such, we reviewed epidemiologic studies published to date to provide an overview of the current state of knowledge on PBDEs' potential role in behavioral development. We identified 19 epidemiologic studies reporting on associations of prenatal and childhood concentrations of PBDEs with behaviors assessed in children from 1 to 12years, including executive function, attention, externalizing and internalizing behaviors, adaptive skills, and social behaviors/Autism Spectrum Disorder (ASD). While the mechanisms of PBDE neurotoxicity in humans are still not clearly elucidated, findings from this review indicate that PBDE exposure during fetal development is associated with impairments in executive function and poorer attentional control in children. Results from large prospective cohorts demonstrate that prenatal and postnatal PBDE exposure adversely impacts externalizing behavior (e.g., hyperactivity and conduct problems). Additional studies are needed to determine whether PBDEs are associated with internalizing problems, adaptive skills, and social behaviors/ASD in children. Future studies will help better understand the potential neurotoxic effects of PBDE exposures during adolescence, possible sex-dependent effects, and the impact of exposure to BDE-209 and alternative flame retardants. Future studies should also examine chemical mixtures to capture real-world exposures when examining PBDEs and their impact on various behavioral domains in the context of multiple chemical exposures.

Evaluating effects of prenatal exposure to phthalate mixtures on birth weight: A comparison of three statistical approaches

OBJECTIVES

We applied three statistical approaches for evaluating associations between prenatal urinary concentrations of a mixture of phthalate metabolites and birth weight.

METHODS

We included 300 women who provided 732 urine samples during pregnancy and delivered a singleton infant. We measured urinary concentrations of metabolites of di(2-ethylhexyl)-phthalate, di-isobutyl-, di-n-butyl-, butylbenzyl-, and diethyl phthalates. We applied 1) linear regressions; 2) classification methods [principal component analysis (PCA) and structural equation models (SEM)]; and 3) Bayesian kernel machine regression (BKMR), to evaluate associations between phthalate metabolite mixtures and birth weight adjusting for potential confounders. Data were presented as mean differences (95% CI) in birth weight (grams) as each phthalate increased from the 10th to the 90th percentile.

RESULTS

When analyzing individual phthalate metabolites using linear regressions, each metabolite demonstrated a modest inverse association with birth weight [from -93 (-206, 21) to -49 (-164, 65)]. When simultaneously including all metabolites in a multivariable model, inflation of the estimates and standard errors were noted. PCA identified two principal components, both inversely associated with birth weight [-23 (-68, 22), -27 (-71, 17), respectively]. These inverse associations were confirmed when applying SEM. BKMR further identified that monoethyl and mono(2-ethylhexyl) phthalate and phthalate concentrations were linearly related to lower birth weight [-51(-164, 63) and -122 (-311, 67), respectively], and suggested no evidence of interaction between metabolites.

CONCLUSIONS

While none of the methods produced significant results, we demonstrated the potential issues arising using linear regression models in the context of correlated exposures. Among the other selected approaches, classification techniques identified common sources of exposures with implications for interventions, while BKMR further identified specific contributions of individual metabolites.

A multivariate approach to investigate the combined biological effects of multiple exposures

Epidemiological studies provide evidence that environmental exposures may affect health through complex mixtures. Formal investigation of the effect of exposure mixtures is usually achieved by modelling interactions, which relies on strong assumptions relating to the identity and the number of the exposures involved in such interactions, and on the order and parametric form of these interactions. These hypotheses become difficult to formulate and justify in an exposome context, where influential exposures are numerous and heterogeneous. To capture both the complexity of the exposome and its possibly pleiotropic effects, models handling multivariate predictors and responses, such as partial least squares (PLS) algorithms, can prove useful. As an illustrative example, we applied PLS models to data from a study investigating the inflammatory response (blood concentration of 13 immune markers) to the exposure to four disinfection by-products (one brominated and three chlorinated compounds), while swimming in a pool. To accommodate the multiple observations per participant (n=60; before and after the swim), we adopted a multilevel extension of PLS algorithms, including sparse PLS models shrinking loadings coefficients of unimportant predictors (exposures) and/or responses (protein levels). Despite the strong correlation among co-occurring exposures, our approach identified a subset of exposures (n=3/4) affecting the exhaled levels of 8 (out of 13) immune markers. PLS algorithms can easily scale to high-dimensional exposures and responses, and prove useful for exposome research to identify sparse sets of exposures jointly affecting a set of (selected) biological markers. Our descriptive work may guide these extensions for higher dimensional data.

The Exposome Research Paradigm: an Opportunity to Understand the Environmental Basis for Human Health and Disease

PURPOSE OF REVIEW

This paper presents an overview of the exposome research paradigm with particular application to understanding human reproduction and development and its implications for health across a lifespan.

RECENT FINDINGS

The exposome research paradigm has generated considerable discussion about its feasibility and utility for delineating the impact of environmental exposures on human health. Early initiatives are underway, including smaller proof-of-principle studies and larger concerted efforts. Despite the notable challenges underlying the exposome paradigm, analytic techniques are being developed to handle its untargeted approach and correlated and multi-level or hierarchical data structures such initiatives generate, while considering multiple comparisons. The relatively short intervals for critical and sensitive windows of human reproduction and development seem well suited for exposome research and may revolutionize our understanding of later onset diseases. Early initiatives suggest that the exposome paradigm is feasible, but its utility remains to be established with applications to population human health research.

Statistical Approaches to Address Multi-Pollutant Mixtures and Multiple Exposures: the State of the Science

PURPOSE OF REVIEW

The purpose of this review is to describe the most recent statistical approaches to estimate the effect of multi-pollutant mixtures or multiple correlated exposures on human health.

RECENT FINDINGS

The health effects of environmental chemicals or air pollutants have been widely described. Often, there exists a complex mixture of different substances, potentially highly correlated with each other and with other (environmental) stressors. Single-exposure approaches do not allow disentangling effects of individual factors and fail to detect potential interactions between exposures. In the last years, sophisticated methods have been developed to investigate the joint or independent health effects of multi-pollutant mixtures or multiple environmental exposures. A classification of the most recent methods is proposed. A non-technical description of each method is provided, together with epidemiological applications and operational details for implementation with standard software.

Acute respiratory symptoms associated with short term fluctuations in ambient pollutants among schoolchildren in Durban, South Africa

Ambient air pollution has been associated with adverse respiratory outcomes, especially among children with asthma. This study reports on associations between daily ambient air pollutant concentrations and the respiratory symptoms of schoolchildren living in Durban, South Africa. This city is Africa's busiest port and a key hub for imported crude oil and exported refined petroleum and petrochemical products, and it experiences a mixture of air pollutants that reflects emissions from industry, traffic and biomass burning. Children in four communities in the highly industrialized southern portion of the city were compared to children of similar socio-economic profiles living in the north of the city. One school was selected in each community. A total of 423 children were recruited. Symptom logs were completed every 1.5-2 h over 3-week period in each of four seasons. Ambient concentrations of NO₂, NO, SO₂, CO, O_3, PM_2.5 and PM₁₀ were measured throughout the study. Generalized estimating equation (GEE) models were used to estimate odds ratios (ORs) and assess lag effects (1-5 days) using single pollutant (single lags or distributed lags) models. Concentrations of SO₂ and NO_x were markedly higher in the south, while PM₁₀ did not vary. Significant increase in the odds ratios of cough were identified for the various lags analyzed. The OR of symptoms was further increased among those living in the south compared to the north. In conclusion, in this analysis of over 70,000 observations, we provide further evidence that exposure to PM₁₀, SO₂, NO₂ and NO is associated with significantly increased occurrence of respiratory symptoms among children. This was evident for cough, shortness of breath, and chest tightness, across the four pollutants and for different lags of exposure. This is the first study describing these changes in sub-Saharan Africa.

Environmental exposure to metals and the risk of hypertension: A cross-sectional study in China

Metal pollution is a severe environmental issue in China, which has been recently linked with the risk of hypertension. However, relevant epidemiological studies are limited. The present exploratory study was conducted to assess the associations of environmental exposure to metals with the odds of hypertension as well as blood pressure (BP) levels using urine samples in a Chinese general population. From May 2016 to April 2017, a total of 823 eligible participants were consecutively enrolled in our study in Wuhan, China. Hypertension was defined as systolic BP (SBP) of ≥140 mmHg or diastolic BP (DBP) of ≥90 mmHg, a self-reported physician diagnosis, or current use of antihypertensive medication. We used urine samples as biomarkers to reflect the levels of environmental exposure to 20 metals. Multivariable regression models were applied to assess the potential association. Multi-metal models were conducted to investigate the impacts of co-exposure to various metals. Based on the results from various models, positive trends for increased odds of hypertension with increasing quartiles of vanadium (V), iron (Fe), zinc (Zn) and selenium (Se) were suggested. Compared with those in the lowest quartiles, participants in the highest quartiles of V, Fe, Zn and Se had a 4.4-fold, 4.9-fold, 4.2-fold and 2.5-fold increased odds of having hypertension, respectively. High urinary Hg level was found to increase the levels of DBP. Individuals in the highest group of Hg were found to have a 4.3 mmHg higher level of DBP. Our findings suggest that environmental exposure to V, Fe, Zn, Se and Hg might increase the risk of hypertension or elevate the levels of BP. These findings warrant further prospective studies in a larger population.

Using machine learning to identify air pollution exposure profiles associated with early cognitive skills among U.S. children

Data-driven machine learning methods present an opportunity to simultaneously assess the impact of multiple air pollutants on health outcomes. The goal of this study was to apply a two-stage, data-driven approach to identify associations between air pollutant exposure profiles and children's cognitive skills. Data from 6900 children enrolled in the Early Childhood Longitudinal Study, Birth Cohort, a national study of children born in 2001 and followed through kindergarten, were linked to estimated concentrations of 104 ambient air toxics in the 2002 National Air Toxics Assessment using ZIP code of residence at age 9 months. In the first-stage, 100 regression trees were learned to identify ambient air pollutant exposure profiles most closely associated with scores on a standardized mathematics test administered to children in kindergarten. In the second-stage, the exposure profiles frequently predicting lower math scores were included within linear regression models and adjusted for confounders in order to estimate the magnitude of their effect on math scores. This approach was applied to the full population, and then to the populations living in urban and highly-populated urban areas. Our first-stage results in the full population suggested children with low trichloroethylene exposure had significantly lower math scores. This association was not observed for children living in urban communities, suggesting that confounding related to urbanicity needs to be considered within the first-stage. When restricting our analysis to populations living in urban and highly-populated urban areas, high isophorone levels were found to predict lower math scores. Within adjusted regression models of children in highly-populated urban areas, the estimated effect of higher isophorone exposure on math scores was -1.19 points (95% CI -1.94, -0.44). Similar results were observed for the overall population of urban children. This data-driven, two-stage approach can be applied to other populations, exposures and outcomes to generate hypotheses within high-dimensional exposure data.

Gestational exposure to endocrine disrupting chemicals in relation to infant birth weight: a Bayesian analysis of the HOME Study

BACKGROUND

Pregnant women are exposed to a mixture of endocrine disrupting chemicals (EDCs). Gestational EDC exposures may be associated with changes in fetal growth that elevates the risk for poor health later in life, but few studies have examined the health effects of simultaneous exposure to multiple chemicals. This study aimed to examine the association of gestational exposure to five chemical classes of potential EDCs: phthalates and bisphenol A, perfluoroalkyl substances (PFAS), polychlorinated biphenyls (PCBs), polybrominated diphenyl ethers (PBDEs), and organochlorine pesticides (OCPs) with infant birth weight.

METHODS

Using data from the Health Outcomes and Measures of Environment (HOME) Study, we examined 272 pregnant women enrolled between 2003-2006. EDC concentrations were quantified in blood and urine samples collected at 16 and 26 weeks gestation. We used Bayesian Hierarchical Linear Models (BHLM) to examine the associations between newborn birth weight and 53 EDCs, 2 organochlorine pesticides (OPPs) and 2 heavy metals.

RESULTS

For a 10-fold increase in chemical concentration, the mean differences in birth weights (95% credible intervals (CI)) were 1 g (-20, 23) for phthalates, -11 g (-52, 34) for PFAS, 0.2 g (-9, 10) for PCBs, -4 g (-30, 22) for PBDEs, and 7 g (-25, 40) for OCPs.

CONCLUSION

Gestational exposure to phthalates, PFAS, PCBs, PBDEs, OCPs or OPPs had null or small associations with birth weight. Gestational OPP, Pb, and PFAS exposures were most strongly associated with lower birth weight.

A birth cohort study in the Middle East: the Qatari birth cohort study (QBiC) phase I

Background

The latest scientific reports raise concerns about the rapidly increasing burden of chronic diseases in the state of Qatar. Pregnant Qatari women often confront complications during pregnancy including gestational diabetes, hypertension, abortion and stillbirth. The investigation of early life environmental, genetic, nutritional and social factors that may affect lifelong health is of great importance. Birth cohort studies offer a great opportunity to address early life hazards and their possible long lasting effects on health.

Methods/design

The Qatari Birth Cohort study is the first mother-child cohort study in the Middle East Area that aims to assess the synergetic role of environmental exposure and genetic factors in the development of chronic disease and monitor woman and child health and/or obstetric characteristics with high prevalence. The present manuscript describes the recruitment phase of the study (duration: 2 years; expected number: 3000 families), where the pregnant Qatari women and their husbands are being contacted before the 15th week of gestation at the Primary Health Care Centers. The consented participants are interviewed to obtain information on several factors (sociodemographic characteristics, dietary habits, occupational/environmental exposure) and maternal characteristics are assessed based on anthropometric measurements, spirometry, and blood pressure. Pregnant women are invited to provide biological samples (blood and urine) in each trimester of their pregnancy, as well as cord blood at delivery. Fathers are also asked to provide biological samples.

Discussion

The present study provides invaluable insights into a wide range of early life factors affecting human health. With a geographical focus on the Middle East, it will be a resource for information to the wider scientific community and will allow the formulation of effective policies with a primary focus on public health interventions for maternal and child health.

Construction of environmental risk score beyond standard linear models using machine learning methods: application to metal mixtures, oxidative stress and cardiovascular disease in NHANES

BACKGROUND

There is growing concern of health effects of exposure to pollutant mixtures. We initially proposed an Environmental Risk Score (ERS) as a summary measure to examine the risk of exposure to multi-pollutants in epidemiologic research considering only pollutant main effects. We expand the ERS by consideration of pollutant-pollutant interactions using modern machine learning methods. We illustrate the multi-pollutant approaches to predicting a marker of oxidative stress (gamma-glutamyl transferase (GGT)), a common disease pathway linking environmental exposure and numerous health endpoints.

METHODS

We examined 20 metal biomarkers measured in urine or whole blood from 6 cycles of the National Health and Nutrition Examination Survey (NHANES 2003-2004 to 2013-2014, n = 9664). We randomly split the data evenly into training and testing sets and constructed ERS's of metal mixtures for GGT using adaptive elastic-net with main effects and pairwise interactions (AENET-I), Bayesian additive regression tree (BART), Bayesian kernel machine regression (BKMR), and Super Learner in the training set and evaluated their performances in the testing set. We also evaluated the associations between GGT-ERS and cardiovascular endpoints.

RESULTS

ERS based on AENET-I performed better than other approaches in terms of prediction errors in the testing set. Important metals identified in relation to GGT include cadmium (urine), dimethylarsonic acid, monomethylarsonic acid, cobalt, and barium. All ERS's showed significant associations with systolic and diastolic blood pressure and hypertension. For hypertension, one SD increase in each ERS from AENET-I, BART and SuperLearner were associated with odds ratios of 1.26 (95% CI, 1.15, 1.38), 1.17 (1.09, 1.25), and 1.30 (1.20, 1.40), respectively. ERS's showed non-significant positive associations with mortality outcomes.

CONCLUSIONS

ERS is a useful tool for characterizing cumulative risk from pollutant mixtures, with accounting for statistical challenges such as high degrees of correlations and pollutant-pollutant interactions. ERS constructed for an intermediate marker like GGT is predictive of related disease endpoints.

Performance of variable selection methods for assessing the health effects of correlated exposures in case-control studies

OBJECTIVES

There is growing recognition that simultaneously assessing multiple exposures may reduce false positive discoveries and improve epidemiological effect estimates. We evaluated the performance of statistical methods for identifying exposure-outcome associations across various data structures typical of environmental and occupational epidemiology analyses.

METHODS

We simulated a case-control study, generating 100 data sets for each of 270 different simulation scenarios; varying the number of exposure variables, the correlation between exposures, sample size, the number of effective exposures and the magnitude of effect estimates. We compared conventional analytical approaches, that is, univariable (with and without multiplicity adjustment), multivariable and stepwise logistic regression, with variable selection methods: sparse partial least squares discriminant analysis, boosting, and frequentist and Bayesian penalised regression approaches.

RESULTS

The variable selection methods consistently yielded more precise effect estimates and generally improved selection accuracy compared with conventional logistic regression methods, especially for scenarios with higher correlation levels. Penalised lasso and elastic net regression both seemed to perform particularly well, specifically when statistical inference based on a balanced weighting of high sensitivity and a low proportion of false discoveries is sought.

CONCLUSIONS

In this extensive simulation study with multicollinear data, we found that most variable selection methods consistently outperformed conventional approaches, and demonstrated how performance is influenced by the structure of the data and underlying model.

The association of air pollution with birthweight and gestational age: evidence from Hong Kong's 'Children of 1997' birth cohort

Background

Previous studies from Western settings have found inconsistent associations of air pollutants with birth outcomes, which are open to residual confounding by socioeconomic position (SEP). We assessed this association in the economically developed non-Western setting of Hong Kong, with high levels of air pollution but little social patterning of these outcomes.

Methods

We obtained PM10, SO2, NO and NO2 from monitoring stations, and assessed their associations with birthweight and gestational age in a large population-representative birth cohort 'Children of 1997', using partial least-square regression to account for the colinearity between pollutants.

Results

PM10 (per 5.7 µg/m3 higher) and NO2 (per 10.9 µg/m3 higher) were associated with birthweight lower by 47.0 g (95% confidence interval (CI) 36.2-56.3) and 16.9 g (95% CI 10.8-22.6), respectively; and were associated with gestational age shorter by 2.1 days (95% CI 1.7-2.4) and 0.7 days (95% CI 0.5-0.8), respectively.

Conclusions

Given minimal confounding by SEP in our setting, these findings provide unequivocal evidence of adverse effects of PM10 and NO2 exposure during pregnancy on birthweight and gestational age. Physiological mechanisms need to be better understood to support effective public health action globally.

Neonatal exposure to environmental pollutants and placental mitochondrial DNA content: A multi-pollutant approach

BACKGROUND

Placental mitochondrial DNA (mtDNA) content can be indicative of oxidative damage to the placenta during fetal development and is responsive to external stressors. In utero exposure to environmental pollutants that may influence placental mtDNA needs further exploration.

OBJECTIVES

We evaluated if placental mtDNA content is altered by environmental pollution in newborns and identified pollutants independently associated to alterations in placental mtDNA content.

METHODS

mtDNA content was measured in placental tissue of 233 newborns. Four perfluoroalkyl compounds and nine organochlorine compounds were quantified in cord blood plasma samples and six toxic metals in whole cord blood. We first applied a LASSO (least absolute shrinkage and selection operator) penalized regression model to identify independent associations between environmental pollutants and placental mtDNA content, without penalization of several covariates. Then adjusted estimates were obtained using an ordinary least squares (OLS) regression model evaluating the pollutants' association with placental mtDNA content, adjusted for several covariates.

RESULTS

Based on LASSO penalized regression, oxychlordane, p,p'-dichlorodiphenyldichloroethylene, β-hexachlorocyclohexane, perfluorononanoic acid, arsenic, cadmium and thallium were identified to be independently associated with placental mtDNA content. The OLS model showed a higher placental mtDNA content of 2.71% (95% CI: 0.3 to 5.2%; p=0.03) and 1.41% (0.1 to 2.8%, p=0.04) for a 25% concentration increase of respectively cord blood β-hexachlorocyclohexane and arsenic. For a 25% concentration increase of cord blood thallium, a 4.88% lower placental mtDNA content (95% CI: -9.1 to -0.5%, p=0.03) was observed.

CONCLUSION

In a multi-pollutant approach, low fetal exposure levels of environmental organic and inorganic pollutants might compromise placental mitochondrial function as exemplified in this study by alterations in mtDNA content.

A systematic comparison of statistical methods to detect interactions in exposome-health associations

BACKGROUND

There is growing interest in examining the simultaneous effects of multiple exposures and, more generally, the effects of mixtures of exposures, as part of the exposome concept (being defined as the totality of human environmental exposures from conception onwards). Uncovering such combined effects is challenging owing to the large number of exposures, several of them being highly correlated. We performed a simulation study in an exposome context to compare the performance of several statistical methods that have been proposed to detect statistical interactions.

METHODS

Simulations were based on an exposome including 237 exposures with a realistic correlation structure. We considered several statistical regression-based methods, including two-step Environment-Wide Association Study (EWAS₂), the Deletion/Substitution/Addition (DSA) algorithm, the Least Absolute Shrinkage and Selection Operator (LASSO), Group-Lasso INTERaction-NET (GLINTERNET), a three-step method based on regression trees and finally Boosted Regression Trees (BRT). We assessed the performance of each method in terms of model size, predictive ability, sensitivity and false discovery rate.

RESULTS

GLINTERNET and DSA had better overall performance than the other methods, with GLINTERNET having better properties in terms of selecting the true predictors (sensitivity) and of predictive ability, while DSA had a lower number of false positives. In terms of ability to capture interaction terms, GLINTERNET and DSA had again the best performances, with the same trade-off between sensitivity and false discovery proportion. When GLINTERNET and DSA failed to select an exposure truly associated with the outcome, they tended to select a highly correlated one. When interactions were not present in the data, using variable selection methods that allowed for interactions had only slight costs in performance compared to methods that only searched for main effects.

CONCLUSIONS

GLINTERNET and DSA provided better performance in detecting two-way interactions, compared to other existing methods.

Pesticide Exposure and Risk of Rheumatoid Arthritis among Licensed Male Pesticide Applicators in the Agricultural Health Study

BACKGROUND

The occupation of farming has been associated with rheumatoid arthritis (RA); pesticides may account for this association, but there are few studies.

OBJECTIVES

We investigated associations between RA and use of pesticides in the Agricultural Health Study.

METHODS

The study sample was drawn from male pesticide applicators enrolled in 1993–1997 who provided questionnaire data at baseline and at least once during follow-up (over a median 18 y; interquartile range 16–19). Incident RA cases (n=220), confirmed by physicians or by self-reported use of disease-modifying antirheumatic drugs, were compared with noncases (n=26,134) who did not report RA. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using logistic regression, adjusting for enrollment age, state, smoking pack-years, and education. We evaluated the association of RA with the use of 46 pesticides and across 4 levels (never use and tertiles) of lifetime days of use for 16 pesticides with OR≥1.2 for ever use.

RESULTS

Incident RA was associated with ever use of fonofos (OR = 1.70; 95% CI: 1.22, 2.37), carbaryl (OR = 1.51; 95% CI: 1.03, 2.23), and chlorimuron ethyl (OR = 1.45; 95% CI: 1.01, 2.07) compared with never use. Statistically significant exposure–response trends in association with RA were observed for lifetime days of use of atrazine [OR_tertile3= 1.62 (95% CI: 1.09, 2.40); p_trend=0.01] and toxaphene [OR_tertile3= 2.42 (95% CI: 1.03, 5.68); p_trend=0.02]. Exposure–response was nonlinear for fonofos [OR_tertile1= 2.27 (95% CI: 1.44, 3.57); OR_tertile2= 0.98 (95% CI: 0.54, 1.80); OR_tertile3= 2.10 (95% CI: 1.32, 3.36); p_trend=0.005] and suggestive for carbaryl (p_trend=0.053).

CONCLUSIONS

Our results provide novel evidence of associations between exposure to some pesticides and RA in male farmers. https://doi.org/10.1289/EHP1013.

Perfluoroalkyl Substances during Pregnancy and Offspring Weight and Adiposity at Birth: Examining Mediation by Maternal Fasting Glucose in the Healthy Start Study

BACKGROUND

Certain perfluoroalkyl and polyfluoroalkyl substances (PFAS) are widespread, persistent environmental contaminants. Prenatal PFAS exposure has been associated with lower birth weight; however, impacts on body composition and factors responsible for this association are unknown.

OBJECTIVES

We aimed to estimate associations between maternal PFAS concentrations and offspring weight and adiposity at birth, and secondarily to estimate associations between PFAS concentrations and maternal glucose and lipids, and to evaluate the potential for these nutrients to mediate associations between PFAS and neonatal outcomes.

METHODS

Within the Healthy Start prospective cohort, concentrations of 11 PFAS, fasting glucose, and lipids were measured in maternal mid-pregnancy serum (n=628). Infant body composition was measured using air displacement plethysmography. Associations between PFAS and birth weight and adiposity, and between PFAS and maternal glucose and lipids, were estimated via linear regression. Associations were decomposed into direct and indirect effects.

RESULTS

Five PFAS were detectable in >50% of participants. Maternal perfluorooctanoate (PFOA) and perfluorononanoate (PFNA) concentrations were inversely associated with birth weight. Adiposity at birth was approximately 10% lower in the highest categories of PFOA, PFNA, and perfluorohexane sulfonate (PFHxS) compared to the lowest categories. PFOA, PFNA, perfluorodecanoate (PFDeA), and PFHxS were inversely associated with maternal glucose. Up to 11.6% of the effect of PFAS on neonatal adiposity was mediated by maternal glucose concentrations. Perfluorooctane sulfonate (PFOS) was not significantly associated with any outcomes studied.

CONCLUSIONS

Follow-up of offspring will determine the potential long-term consequences of lower weight and adiposity at birth associated with prenatal PFAS exposure. https://doi.org/10.1289/EHP641.

Exposure to multiple chemicals in a cohort of reproductive-aged Danish women

BACKGROUND

Current exposure assessment research does not sufficiently address multi-pollutant exposure and their correlations in human media. Understanding the extent of chemical exposure in reproductive-aged women is of particular concern due to the potential for in utero exposure and fetal susceptibility.

OBJECTIVES

The objectives of this study were to characterize concentrations of chemical biomarkers during preconception and examine correlations between and within chemical classes.

METHODS

We examined concentrations of 135 biomarkers from 16 chemical classes in blood and urine from 73 women aged 18-40 enrolled in Snart Foraeldre/Milieu, a prospective cohort study of pregnancy planners in Denmark (2011-2014). We compared biomarker concentrations with United States similarly-aged, non-pregnant women who participated in the National Health and Nutrition Environmental Survey (NHANES) and with other international biomonitoring studies. We performed principal component analysis to examine biomarker correlations.

RESULTS

The mean number of biomarkers detected in the population was 92 (range: 60-108). The most commonly detected chemical classes were phthalates, metals, phytoestrogens and polycyclic aromatic hydrocarbons. Except blood mercury, urinary barium and enterolactone, geometric means were higher in women from NHANES. Chemical classes measured in urine generally did not load on a single component, suggesting high between-class correlation among urinary biomarkers, while there is high within-class correlation for biomarkers measured in serum and blood.

CONCLUSIONS

We identified ubiquitous exposure to multiple chemical classes in reproductive-aged Danish women, supporting the need for more research on chemical mixtures during preconception and early pregnancy. Inter- and intra-class correlation between measured biomarkers may reflect common exposure sources, specific lifestyle factors or shared metabolism pathways.

Informatics and Data Analytics to Support Exposome-Based Discovery for Public Health

The complexity of the human exposome-the totality of environmental exposures encountered from birth to death-motivates systematic, high-throughput approaches to discover new environmental determinants of disease. In this review, we describe the state of science in analyzing the human exposome and provide recommendations for the public health community to consider in dealing with analytic challenges of exposome-based biomedical research. We describe extant and novel analytic methods needed to associate the exposome with critical health outcomes and contextualize the data-centered challenges by drawing parallels to other research endeavors such as human genomics research. We discuss efforts for training scientists who can bridge public health, genomics, and biomedicine in informatics and statistics. If an exposome data ecosystem is brought to fruition, it will likely play a role as central as genomic science has had in molding the current and new generations of biomedical researchers, computational scientists, and public health research programs.

Identification of chemical mixtures to which Canadian pregnant women are exposed: The MIREC Study

Depending on the chemical and the outcome, prenatal exposures to environmental chemicals can lead to adverse effects on the pregnancy and child development, especially if exposure occurs during early gestation. Instead of focusing on prenatal exposure to individual chemicals, more studies have taken into account that humans are exposed to multiple environmental chemicals on a daily basis. The objectives of this analysis were to identify the pattern of chemical mixtures to which women are exposed and to characterize women with elevated exposures to various mixtures. Statistical techniques were applied to 28 chemicals measured simultaneously in the first trimester and socio-demographic factors of 1744 participants from the Maternal-Infant Research on Environment Chemicals (MIREC) Study. Cluster analysis was implemented to categorize participants based on their socio-demographic characteristics, while principal component analysis (PCA) was used to extract the chemicals with similar patterns and to reduce the dimension of the dataset. Next, hypothesis testing determined if the mean converted concentrations of chemical substances differed significantly among women with different socio-demographic backgrounds as well as among clusters. Cluster analysis identified six main socio-demographic clusters. Eleven components, which explained approximately 70% of the variance in the data, were retained in the PCA. Persistent organic pollutants (PCB118, PCB138, PCB153, PCB180, OXYCHLOR and TRANSNONA) and phthalates (MEOHP, MEHHP and MEHP) dominated the first and second components, respectively, and the first two components explained 25.8% of the source variation. Prenatal exposure to persistent organic pollutants (first component) were positively associated with women who have lower education or higher income, were born in Canada, have BMI ≥25, or were expecting their first child in our study population. MEOHP, MEHHP and MEHP, dominating the second component, were detected in at least 98% of 1744 participants in our cohort study; however, no particular group of pregnant women was identified to be highly exposed to phthalates. While widely recognized as important to studying potential health effects, identifying the mixture of chemicals to which various segments of the population are exposed has been problematic. We present an approach using factor analysis through principal component method and cluster analysis as an attempt to determine the pregnancy exposome. Future studies should focus on how to include these matrices in examining the health effects of prenatal exposure to chemical mixtures in pregnant women and their children.

Current approaches used in epidemiologic studies to examine short-term multipollutant air pollution exposures

PURPOSE

Air pollution epidemiology traditionally focuses on the relationship between individual air pollutants and health outcomes (e.g., mortality). To account for potential copollutant confounding, individual pollutant associations are often estimated by adjusting or controlling for other pollutants in the mixture. Recently, the need to characterize the relationship between health outcomes and the larger multipollutant mixture has been emphasized in an attempt to better protect public health and inform more sustainable air quality management decisions.

METHODS

New and innovative statistical methods to examine multipollutant exposures were identified through a broad literature search, with a specific focus on those statistical approaches currently used in epidemiologic studies of short-term exposures to criteria air pollutants (i.e., particulate matter, carbon monoxide, sulfur dioxide, nitrogen dioxide, and ozone).

RESULTS

Five broad classes of statistical approaches were identified for examining associations between short-term multipollutant exposures and health outcomes, specifically additive main effects, effect measure modification, unsupervised dimension reduction, supervised dimension reduction, and nonparametric methods. These approaches are characterized including advantages and limitations in different epidemiologic scenarios.

DISCUSSION

By highlighting the characteristics of various studies in which multipollutant statistical methods have been used, this review provides epidemiologists and biostatisticians with a resource to aid in the selection of the most optimal statistical method to use when examining multipollutant exposures.

A Systematic Comparison of Linear Regression-Based Statistical Methods to Assess Exposome-Health Associations

BACKGROUND

The exposome constitutes a promising framework to improve understanding of the effects of environmental exposures on health by explicitly considering multiple testing and avoiding selective reporting. However, exposome studies are challenged by the simultaneous consideration of many correlated exposures.

OBJECTIVES

We compared the performances of linear regression-based statistical methods in assessing exposome-health associations.

METHODS

In a simulation study, we generated 237 exposure covariates with a realistic correlation structure and with a health outcome linearly related to 0 to 25 of these covariates. Statistical methods were compared primarily in terms of false discovery proportion (FDP) and sensitivity.

RESULTS

On average over all simulation settings, the elastic net and sparse partial least-squares regression showed a sensitivity of 76% and an FDP of 44%; Graphical Unit Evolutionary Stochastic Search (GUESS) and the deletion/substitution/addition (DSA) algorithm revealed a sensitivity of 81% and an FDP of 34%. The environment-wide association study (EWAS) underperformed these methods in terms of FDP (average FDP, 86%) despite a higher sensitivity. Performances decreased considerably when assuming an exposome exposure matrix with high levels of correlation between covariates.

CONCLUSIONS

Correlation between exposures is a challenge for exposome research, and the statistical methods investigated in this study were limited in their ability to efficiently differentiate true predictors from correlated covariates in a realistic exposome context. Although GUESS and DSA provided a marginally better balance between sensitivity and FDP, they did not outperform the other multivariate methods across all scenarios and properties examined, and computational complexity and flexibility should also be considered when choosing between these methods. Citation: Agier L, Portengen L, Chadeau-Hyam M, Basagaña X, Giorgis-Allemand L, Siroux V, Robinson O, Vlaanderen J, González JR, Nieuwenhuijsen MJ, Vineis P, Vrijheid M, Slama R, Vermeulen R. 2016. A systematic comparison of linear regression-based statistical methods to assess exposome-health associations. Environ Health Perspect 124:1848-1856; http://dx.doi.org/10.1289/EHP172.

Novel application of statistical methods for analysis of multiple toxicants identifies DDT as a risk factor for early child behavioral problems

BACKGROUND

The aim of this study was to assess the association between postnatal exposure to multiple persistent organic pollutants (POPs) measured in breast milk samples and early behavioral problems using statistical methods to deal with correlated exposure data.

METHODS

We used data from the Norwegian HUMIS study. We measured concentrations of 24 different POPs in human milk from 612 mothers (median collection time: 32 days after delivery), including 13 polychlorinated biphenyls (PCB) congeners, 6 polybrominated diphenyl ethers (PBDE) congeners and five organochlorine compounds. We assessed child behavioral problems at 12 and 24 months using the infant toddler symptom checklist (ITSC). Higher score in ITSC corresponds to more behavioral problems. First we performed principal component analysis (PCA). Then two variable selection methods, elastic net (ENET) and Bayesian model averaging (BMA), were applied to select any toxicants associated with behavioral problems. Finally, the effect size of the selected toxicants was estimated using multivariate linear regression analyses.

RESULTS

p,p'-DDT was associated with behavioral problems at 12 months in all the applied models. Specifically, the principal component composed of organochlorine pesticides was significantly associated with behavioral problems and both ENET and BMA identified p,p'-DDT as associated with behavioral problems. Using a multiple linear regression model an interquartile increase in p,p'-DDT was associated with a 0.62 unit increase in ITSC score (95% CI 0.45, 0.79) at 12 months, corresponding to more behavioral problems. The association was modified by maternal education: the effect of p,p'-DDT was strongest in women with lower education (β=0.59; 95%CI: 0.38, 0.81) compared to the mother with higher education (β=0.14; 95%CI: -0.05, 0.34) (p-value for interaction=0.089). At 24 months, neither selection method consistently identified any toxicant associated with behavioral problems.

CONCLUSION

Within a mixture of 24 toxicants measured in breast milk, p,p'-DDT was the single toxicant associated with behavioral problems at 12 months using different methods for handling numerous correlated exposures.

Association between phthalate exposure and lower handgrip strength in an elderly population: a repeated-measures study

BACKGROUND

Decreased muscle strength can lead to adverse health outcomes in the elderly. A potential association between phthalate exposure and muscle strength was suggested previously, but has not been investigated directly. We hypothesized that phthalate exposure is associated with lower handgrip strength and that the association is modified by the dietary omega-6 to omega-3 ratio.

METHODS

We analyzed 1,228 participants (≥60 years of age) recruited in Seoul and Asan, Republic of Korea. The study participants were surveyed up to three times between 2012 and 2015. At every survey, we collected urine samples and measured handgrip strength twice for each hand. The associations between urine phthalate metabolite concentrations and handgrip strength were evaluated using linear mixed models. Based on dietary information from 391 individuals who participated in the first survey in Seoul, we evaluated the heterogeneity of the association for those with high and low omega-6 to omega-3 ratios, using 8.81 (the 75th quantile) as a cutoff value.

RESULTS

Log-transformed creatinine-adjusted concentrations of mono-(2-ethyl-5-oxohexyl phthalate (MEOHP), mono-(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), and mono-n-butyl phthalate (MnBP) were inversely associated with all measured handgrip strengths (β = -0.69 to -0.42, all p-values < 0.05). Associations between phthalate biomarkers and handgrip strength did not differ by sex. When the dietary subgroup was stratified by the omega-6 to omega-3 ratio, the associations were stronger among participants with high ratios.

CONCLUSIONS

We found inverse associations between phthalate biomarkers and handgrip strength in the elderly; this association was modified by the dietary omega-6 to omega-3 ratio.

Adolescent epigenetic profiles and environmental exposures from early life through peri-adolescence

Epigenetic perturbations induced by environmental exposures at susceptible lifestages contribute to disease development. Even so, the influence of early life and ongoing exposures on the adolescent epigenome is rarely examined. We examined the association of exposure biomarkers for lead (Pb), bisphenol A (BPA), and nine phthalates metabolites with blood leukocyte DNA methylation at LINE-1 repetitive elements and environmentally responsive genes ( IGF2 , H19 , and HSD11B2 ) in peri-adolescents. Participants ( n  = 247) from the Early Life Exposures in Mexico to Environmental Toxicants (ELEMENT) birth cohorts were followed-up once between the ages of 8 and 14 years, and concurrent exposures were measured in biospecimen collected at that time (blood Pb, urinary BPA, and phthalate metabolites). Prenatal and childhood exposures to Pb were previously approximated using maternal and child samples. BPA and phthalate metabolites were measured in third trimester maternal urine samples. Significant associations ( P  < 0.05) were observed between DNA methylation and exposure biomarkers that were gene and biomarker specific. For example, Pb was only associated with LINE-1 hypomethylation during pregnancy ( P  = 0.04), while early childhood Pb was instead associated with H19 hypermethylation ( P  = 0.04). Concurrent urinary mono (2-ethylhexyl) phthalate (MEHP) was associated with HSD11B2 hypermethylation ( P  = 0.005). Sex-specific associations, particularly among males, were also observed. In addition to single exposure models, principal component analysis was employed to examine exposure mixtures. This method largely corroborated the findings of the single exposure models. This study along with others in the field suggests that environment-epigenetic relationships vary by chemical, exposure timing, and sex.

Combined Effects of Prenatal Exposures to Environmental Chemicals on Birth Weight

Prenatal chemical exposure has been frequently associated with reduced fetal growth by single pollutant regression models although inconsistent results have been obtained. Our study estimated the effects of exposure to single pollutants and mixtures on birth weight in 248 mother-child pairs. Arsenic, copper, lead, manganese and thallium were measured in cord blood, cadmium in maternal blood, methylmercury in maternal hair, and five organochlorines, two perfluorinated compounds and diethylhexyl phthalate metabolites in cord plasma. Daily exposure to particulate matter was modeled and averaged over the duration of gestation. In single pollutant models, arsenic was significantly associated with reduced birth weight. The effect estimate increased when including cadmium, and mono-(2-ethyl-5-carboxypentyl) phthalate (MECPP) co-exposure. Combining exposures by principal component analysis generated an exposure factor loaded by cadmium and arsenic that was associated with reduced birth weight. MECPP induced gender specific effects. In girls, the effect estimate was doubled with co-exposure of thallium, PFOS, lead, cadmium, manganese, and mercury, while in boys, the mixture of MECPP with cadmium showed the strongest association with birth weight. In conclusion, birth weight was consistently inversely associated with exposure to pollutant mixtures. Chemicals not showing significant associations at single pollutant level contributed to stronger effects when analyzed as mixtures.

What Can Epidemiological Studies Tell Us about the Impact of Chemical Mixtures on Human Health?

Humans are exposed to a large number of environmental chemicals: Some of these may be toxic, and many others have unknown or poorly characterized health effects. There is intense interest in determining the impact of exposure to environmental chemical mixtures on human health. As the study of mixtures continues to evolve in the field of environmental epidemiology, it is imperative that we understand the methodologic challenges of this research and the types of questions we can address using epidemiological data. In this article, we summarize some of the unique challenges in exposure assessment, statistical methods, and methodology that epidemiologists face in addressing chemical mixtures. We propose three broad questions that epidemiological studies can address: a) What are the potential health impacts of individual chemical agents? b) What is the interaction among agents? And c) what are the health effects of cumulative exposure to multiple agents? As the field of mixtures research grows, we can use these three questions as a basis for defining our research questions and for developing methods that will help us better understand the effect of chemical exposures on human disease and well-being.

A Characterisation and Profiling of District Health Indicators in Zimbabwe: An Application of Principal Component Analysis in a Data Limited Setting

Background: The Ministry of Health and Child Care, Zimbabwe does not have a method for prioritization and equitable allocation of its share of the national health budget and other resources in the sector. Regional allocations at the provincial level are made regardless of the provinces' disease burden, population size, or needs. Currently there is no method available to show how the provinces eventually allocate these resources to the lower levels of care. In a data limited country such as Zimbabwe, Principal Component Analysis method can be used to identify a set of indicators that account for cross variation between different regions. This set of indicators could then be used by planners as reference indicators for equitable allocation of resources and prioritization of health care interventions. Objective: The aim of the study was to construct a set of simple, feasible, reliable and valid composite health indicators for use in characterising and profiling of the different districts in Zimbabwe. Method: This was a retrospective analysis of secondary data to derive composite indices for the 57 administrative health districts in Zimbabwe using routinely collected secondary data. The data was extracted from the 2012 Zimbabwe Health information database, the 2012 National Census and the 2011 Prices, Income and Expenditure Survey. Results: The analysis of the data resulted in the construction of 10 mutually exclusive principal composite indices, which included demographic, child related, disease related and health systems related indices. The 10 composite indices (population, immunisation, child mortality, antenatal care, HIV/TB, malaria, non-communicable diseases, socioeconomic, health seeking behaviour and infrastructure) were tested for construct and content validity and were found to be statistically robust, reliable and consistent with observed behaviour. Conclusion: The composite indices exhibited internal consistency and construct validity to be regarded as true representations of the cross variation of the 57 districts in Zimbabwe; hence these indices could be used to characterise the behaviour and assess the performance of these districts. There is also potential use for these indices in the areas of resource allocation and prioritisation of health interventions.

Exposure to Endocrine-Disrupting Chemicals during Pregnancy and Weight at 7 Years of Age: A Multi-pollutant Approach

BACKGROUND

Prenatal exposure to endocrine-disrupting chemicals (EDCs) may induce weight gain and obesity in children, but the obesogenic effects of mixtures have not been studied.

OBJECTIVE

We evaluated the associations between pre- and perinatal biomarker concentrations of 27 EDCs and child weight status at 7 years of age.

METHODS

In pregnant women enrolled in a Spanish birth cohort study between 2004 and 2006, we measured the concentrations of 10 phthalate metabolites, bisphenol A, cadmium, arsenic, and lead in two maternal pregnancy urine samples; 6 organochlorine compounds in maternal pregnancy serum; mercury in cord blood; and 6 polybrominated diphenyl ether congeners in colostrum. Among 470 children at 7 years, body mass index (BMI) z-scores were calculated, and overweight was defined as BMI > 85th percentile. We estimated associations with EDCs in single-pollutant models and applied principal-component analysis (PCA) on the 27 pollutant concentrations.

RESULTS

In single-pollutant models, HCB (hexachlorobenzene), βHCH (β-hexachlorocyclohexane), and polychlorinated biphenyl (PCB) congeners 138 and 180 were associated with increased child BMI z-scores; and HCB, βHCH, PCB-138, and DDE (dichlorodiphenyldichloroethylene) with overweight risk. PCA generated four factors that accounted for 43.4% of the total variance. The organochlorine factor was positively associated with BMI z-scores and with overweight (adjusted RR, tertile 3 vs. 1: 2.59; 95% CI: 1.19, 5.63), and these associations were robust to adjustment for other EDCs. Exposure in the second tertile of the phthalate factor was inversely associated with overweight.

CONCLUSIONS

Prenatal exposure to organochlorines was positively associated with overweight at age 7 years in our study population. Other EDCs exposures did not confound this association.

Associations between ambient air pollutant mixtures and pediatric asthma emergency department visits in three cities: a classification and regression tree approach

BACKGROUND

Characterizing multipollutant health effects is challenging. We use classification and regression trees to identify multipollutant joint effects associated with pediatric asthma exacerbations and compare these results with those from a multipollutant regression model with continuous joint effects.

METHODS

We investigate the joint effects of ozone, NO2 and PM2.5 on emergency department visits for pediatric asthma in Atlanta (1999-2009), Dallas (2006-2009) and St. Louis (2001-2007). Daily concentrations of each pollutant were categorized into four levels, resulting in 64 different combinations or "Day-Types" that can occur. Days when all pollutants were in the lowest level were withheld as the reference group. Separate regression trees were grown for each city, with partitioning based on Day-Type in a model with control for confounding. Day-Types that appeared together in the same terminal node in all three trees were considered to be mixtures of potential interest and were included as indicator variables in a three-city Poisson generalized linear model with confounding control and rate ratios calculated relative to the reference group. For comparison, we estimated analogous joint effects from a multipollutant Poisson model that included terms for each pollutant, with concentrations modeled continuously.

RESULTS AND DISCUSSION

No single mixture emerged as the most harmful. Instead, the rate ratios for the mixtures suggest that all three pollutants drive the health association, and that the rate plateaus in the mixtures with the highest concentrations. In contrast, the results from the comparison model are dominated by an association with ozone and suggest that the rate increases with concentration.

CONCLUSION

The use of classification and regression trees to identify joint effects may lead to different conclusions than multipollutant models with continuous joint effects and may serve as a complementary approach for understanding health effects of multipollutant mixtures.

Exploring associations between multipollutant day types and asthma morbidity: epidemiologic applications of self-organizing map ambient air quality classifications

BACKGROUND

Recent interest in the health effects of air pollution focuses on identifying combinations of multiple pollutants that may be associated with adverse health risks.

OBJECTIVE

Present a methodology allowing health investigators to explore associations between categories of ambient air quality days (i.e., multipollutant day types) and adverse health.

METHODS

First, we applied a self-organizing map (SOM) to daily air quality data for 10 pollutants collected between January 1999 and December 2008 at a central monitoring location in Atlanta, Georgia to define a collection of multipollutant day types. Next, we conducted an epidemiologic analysis using our categories as a multipollutant metric of ambient air quality and daily counts of emergency department (ED) visits for asthma or wheeze among children aged 5 to 17 as the health endpoint. We estimated rate ratios (RR) for the association of multipollutant day types and pediatric asthma ED visits using a Poisson generalized linear model controlling for long-term, seasonal, and weekday trends and weather.

RESULTS

Using a low pollution day type as the reference level, we found significant associations of increased asthma morbidity in three of nine categories suggesting adverse effects when combinations of primary (CO, NO2, NOX, EC, and OC) and/or secondary (O3, NH4, SO4) pollutants exhibited elevated concentrations (typically, occurring on dry days with low wind speed). On days with only NO3 elevated (which tended to be relatively cool) and on days when only SO2 was elevated (which likely reflected plume touchdowns from coal combustion point sources), estimated associations were modestly positive but confidence intervals included the null.

CONCLUSIONS

We found that ED visits for pediatric asthma in Atlanta were more strongly associated with certain day types defined by multipollutant characteristics than days with low pollution levels; however, findings did not suggest that any specific combinations were more harmful than others. Relative to other health endpoints, asthma exacerbation may be driven more by total ambient pollutant exposure than by composition.

Particulate matter components, sources, and health: Systematic approaches to testing effects

Exposure to particulate matter (PM) is associated with adverse health outcomes. There has long been a question as to whether some components of the PM mixture are of greater public health concern than others so that the sources that emit the more toxic components could be controlled. In this paper, we describe the National Particle Component Toxicity (NPACT) initiative, a comprehensive research program that combined epidemiologic and toxicologic approaches to evaluate this critical question, partly relying on information from a national network of air quality monitors that provided data on speciated PM2.5 (PM with an aerodynamic diameter<2.5 μm) starting in 2000. We also consider the results of the NPACT program in the context of selected research on PM components and health in order to assess the current state of the field. Overall, the ambitious NPACT research program found associations of secondary sulfate and, to a somewhat lesser extent, traffic sources with health effects. Although this and other research has linked a variety of health effects to multiple groups of PM components and sources of PM, the collective evidence has not yet isolated factors or sources that would be closely and unequivocally more strongly related to specific health outcomes. If greater success is to be achieved in isolating the effects of pollutants from mobile and other major sources, either as individual components or as a mixture, more advanced approaches and additional measurements will be needed so that exposure at the individual or population level can be assessed more accurately. Enhanced understanding of exposure and health effects is needed before it can be concluded that regulations targeting specific sources or components of PM2.5 will protect public health more effectively than continuing to follow the current practices of targeting PM2.5 mass as a whole.

IMPLICATIONS

This paper describes a comprehensive epidemiologic and toxicologic research program to evaluate whether some components and sources of PM may be more toxic than others. This question is important for regulatory agencies in setting air quality standards to protect people's health. The results show that PM from coal and oil combustion and from traffic sources was associated with adverse health outcomes, but other components and sources could not definitively be ruled out. Thus, given current knowledge, the current practice of setting air quality standards for PM mass as a whole likely remains an effective approach to protecting public health.

Urinary metals and heart rate variability: a cross-sectional study of urban adults in Wuhan, China

BACKGROUND

Epidemiological studies have suggested an association between external estimates of exposure to metals in air particles and altered heart rate variability (HRV). However, studies on the association between internal assessments of metals exposure and HRV are limited.

OBJECTIVES

The purpose of this study was to examine the potential association between urinary metals and HRV among residents of an urban community in Wuhan, China.

METHODS

We performed a cross-sectional analysis of 23 urinary metals and 5-min HRV indices (SDNN, standard deviation of normal-to-normal intervals; r-MSSD, root mean square of successive differences in adjacent normal-to-normal intervals; LF, low frequency; HF, high frequency; TP, total power) using baseline data on 2,004 adult residents of Wuhan.

RESULTS

After adjusting for other metals, creatinine, and other covariates, natural log-transformed urine titanium concentration was positively associated with all HRV indices (all p < 0.05). Moreover, we estimated negative associations between cadmium and r-MSSD, LF, HF, and TP; between lead and r-MSSD, HF, and TP; and between iron, copper, and arsenic and HF, SDNN, and LF, respectively, based on models adjusted for other metals, creatinine, and covariates (all p < 0.10). Several associations differed according to cardiovascular disease risk factors. For example, negative associations between cadmium and r-MSSD were stronger among participants ≤ 52 years of age (vs. > 52), current smokers (vs. nonsmokers), body mass index < 25 kg/m2 (vs. ≥ 25), and among those who were not hypertensive.

CONCLUSIONS

Urine concentrations of several metals were associated with HRV parameters in our cross-sectional study population. These findings need replication in other studies with adequate sample sizes.