Latent class analysis to model multiple chemical exposures among children

Estimating effects of longitudinal and cumulative exposure to PFAS mixtures on early adolescent body composition

Few methods have been used to characterize repeatedly measured biomarkers of chemical mixtures. We applied latent profile analysis (LPA) to serum concentrations of 4 perfluoroalkyl and polyfluoroalkyl substances (PFAS) measured at 4 time points from gestation to age 12 years. We evaluated the relationships between profiles and z scores of height, body mass index, fat mass index, and lean body mass index at age 12 years (n = 218). We compared LPA findings with an alternative approach for cumulative PFAS mixtures using g-computation to estimate the effect of simultaneously increasing the area under the receiver operating characteristic curve (AUC) for all PFAS. We identified 2 profiles: a higher PFAS profile (35% of sample) and a lower PFAS profile (relative to each other), based on their average PFAS concentrations at all time points. The higher PFAS profile had generally lower z scores for all outcomes, with somewhat larger effects for males, though all 95% CIs crossed the null. For example, the higher PFAS profile was associated with a 0.50-unit lower (β = -0.50; 95% CI, -1.07 to 0.08) BMI z score among males but not among females (β = 0.04; 95% CI, -0.45 to 0.54). We observed similar patterns with AUCs. We found that a higher childhood PFAS profile and higher cumulative PFAS mixtures may be associated with altered growth in early adolescence. This article is part of a Special Collection on Environmental Epidemiology.

Evaluating the association between longitudinal exposure to a PFAS mixture and adolescent cardiometabolic risk in the HOME Study

Background

Exposure to per- and polyfluoroalkyl substances (PFAS) throughout gestation and childhood may impact cardiometabolic risk.

Methods

In 179 HOME Study participants (Cincinnati, OH; recruited 2003-2006), we used latent profile analysis to identify two distinct patterns of PFAS exposure from serum concentrations of four PFAS measured at birth and ages 3, 8, and 12 years. We assessed the homeostatic model of insulin resistance, triglycerides-to-high-density lipoprotein cholesterol ratio, leptin-to-adiponectin ratio, systolic blood pressure, visceral fat, and hemoglobin A1c levels at age 12 years. We used multivariable linear regression to assess the association of membership in the longitudinal PFAS mixture exposure group with a summary measure of overall cardiometabolic risk and individual components.

Results

One PFAS exposure profile (n = 66, 39%) had higher geometric means of all PFAS across all visits than the other. Although adjusted associations were null in the full sample, child sex modified the association of longitudinal PFAS mixture exposure group with overall cardiometabolic risk, leptin-to-adiponectin ratio, systolic blood pressure, and visceral fat (interaction term P values: 0.02-0.08). Females in the higher exposure group had higher cardiometabolic risk scores (ß = 0.43; 95% CI = -0.08, 0.94), systolic blood pressures (ß = 0.6; 95% CI = 0.1, 1.1), and visceral fat (ß = 0.44; 95% CI = -0.13, 1.01); males had lower cardiometabolic risk scores (ß = -0.52; 95% CI = -1.06, -0.06), leptin-to-adiponectin ratios (ß = -0.7; 95% CI = -1.29, -0.1), systolic blood pressures (ß = -0.14; 95% CI = -0.7, 0.41), and visceral fat (ß = -0.52; 95% CI = -0.84, -0.19).

Conclusions

Exposure to this PFAS mixture throughout childhood may have sex-specific effects on adolescent cardiometabolic risk.

The mediating role of telomere length in multi-pollutant exposure associated with metabolic syndrome in adults

Metabolic syndrome is a chronic and complex disease characterized by environmental and genetic factors. However, the underlying mechanisms remain unclear. This study assessed the relationship between exposure to a mixture of environmental chemicals and metabolic syndrome (MetS) and further examined whether telomere length (TL) moderated these relationships. A total of 1265 adults aged > 20 years participated in the study. Data on multiple pollutants (polycyclic aromatic hydrocarbons, phthalates, and metals), MetS, leukocyte telomere length (LTL), and confounders were provided in the 2001-2002 National Health and Nutrition Examination Survey. The correlations between multi-pollutant exposure, TL, and MetS in the males and females were separately assessed using principal component analysis (PCA), logistic and extended linear regression models, Bayesian kernel machine regression (BKMR), and mediation analysis. Four factors were generated in PCA that accounted for 76.2% and 77.5% of the total environmental pollutants in males and females, respectively. The highest quantiles of PC2 and PC4 were associated with the risk of TL shortening (P < 0.05). We observed that the relationship between PC2, PC4, and MetS risk was significant in the participants with median TL levels (P for trend = 0.04 for PC2, and P for trend = 0.01 for PC4). Furthermore, mediation analysis revealed that TL could explain 26.1% and 17.1% of the effects of PC2 and PC4 associated with MetS in males, respectively. The results of BKMR model revealed that these associations were mainly driven by 1-PYE (cPIP = 0.65) and Cd (cPIP = 0.29) in PC2. Meanwhile, TL could explain 17.7% of the mediation effects of PC2 associated with MetS in the females. However, the relationships between pollutants and MetS were sparse and inconsistent in the females. Our findings suggest that the effects of the risk of MetS associated with mixed exposure to multiple pollutants are mediated by TL, and this mediating effect in the males is more pronounced than that in the females.

Reducing the complexity of high-dimensional environmental data: An analytical framework using LASSO with considerations of confounding for statistical inference

PURPOSE

Frameworks for selecting exposures in high-dimensional environmental datasets, while considering confounding, are lacking. We present a two-step approach for exposure selection with subsequent confounder adjustment for statistical inference.

METHODS

We measured cognitive ability in 338 children using the Woodcock-Muñoz General Intellectual Ability (GIA) score, and potential associated features across several environmental domains. Initially, 111 variables theoretically associated with GIA score were introduced into a Least Absolute Shrinkage and Selection Operator (LASSO) in a 50% feature selection subsample. Effect estimates for selected features were subsequently modeled in linear regressions in a 50% inference (hold out) subsample, first adjusting for sex and age and later for covariates selected via directed acyclic graphs (DAGs). All models were adjusted for clustering by school.

RESULTS

Of the 15 LASSO selected variables, eleven were not associated with GIA score following our inference modeling approach. Four variables were associated with GIA scores, including: serum ferritin adjusted for inflammation (inversely), mother's IQ (positively), father's education (positively), and hours per day the child works on homework (positively). Serum ferritin was not in the expected direction.

CONCLUSIONS

Our two-step approach moves high-dimensional feature selection a step further by incorporating DAG-based confounder adjustment for statistical inference.

Associations between prepregnancy body mass index, gestational weight gain and weight catch-up in small-for-gestational-age children

Inadequate gestational weight gain (GWG) was related with a higher incidence of small‐for‐gestational‐age (SGA) births than appropriate GWG; however, the long‐term association of maternal GWG with weight catch‐up growth in SGA children remains unknown. The objective of this study is to evaluate the associations between prepregnancy body mass index (pBMI), GWG and weight catch‐up patterns in SGA children. Data were from the Collaborative Perinatal Project, an American multicentre prospective cohort study. A total of 56,990 gravidas were recruited at the first prenatal visit, and children were followed up until school age. Maternal pBMI, GWG and physical growth of the offspring at birth, 4 months, 1 year, 4 years and 7 years old were recorded. The latent class analysis was employed to form weight catch‐up growth patterns (appropriate, excessive, slow, regression and no catch‐up patterns) in SGA children. SGA children who developed the ‘appropriate catch‐up growth’ pattern and whose mothers had appropriate pBMI and GWG were chosen as the reference. Associations between GWG for different pBMI and weight catch‐up patterns were analysed by multivariate logistic regression models. A total of 1619 infants (9.45%) were born term SGA. After adjusting for relevant confounders, compared with SGA children whose mothers had appropriate pBMI and GWG, SGA children with maternal prepregnancy underweight (for inadequate GWG, GWG below recommendations, adjusted OR: 2.88, 95% CI: 1.13–7.31; for appropriate/excessive GWG, adjusted OR: 3.07, 95% CI: 1.74–5.42) or with prepregnancy normal weight but inadequate GWG (adjusted OR: 2.14, 95% CI: 1.36–3.38) were at a higher risk of having the ‘no catch‐up growth’ pattern. We suggest that SGA children with maternal prepregnancy underweight or inadequate GWG tend to have a poor weight catch‐up growth at least until school age.

Latent classes for chemical mixtures analyses in epidemiology: an example using phthalate and phenol exposure biomarkers in pregnant women

Latent class analysis (LCA), although minimally applied to the statistical analysis of mixtures, may serve as a useful tool for identifying individuals with shared real-life profiles of chemical exposures. Knowledge of these groupings and their risk of adverse outcomes has the potential to inform targeted public health prevention strategies. This example applies LCA to identify clusters of pregnant women from a case-control study within the LIFECODES birth cohort with shared exposure patterns across a panel of urinary phthalate metabolites and parabens, and to evaluate the association between cluster membership and urinary oxidative stress biomarkers. LCA identified individuals with: "low exposure," "low phthalates, high parabens," "high phthalates, low parabens," and "high exposure." Class membership was associated with several demographic characteristics. Compared with "low exposure," women classified as having "high exposure" had elevated urinary concentrations of the oxidative stress biomarkers 8-hydroxydeoxyguanosine (19% higher, 95% confidence interval [CI] = 7, 32%) and 8-isoprostane (31% higher, 95% CI = -5, 64%). However, contrast examinations indicated that associations between oxidative stress biomarkers and "high exposure" were not statistically different from those with "high phthalates, low parabens" suggesting a minimal effect of higher paraben exposure in the presence of high phthalates. The presented example offers verification of latent class assignments through application to an additional data set as well as a comparison to another unsupervised clustering approach, k-means clustering. LCA may be more easily implemented, more consistent, and more able to provide interpretable output.

Latent class analysis of the association between polycyclic aromatic hydrocarbon exposures and body mass index

BACKGROUND

People experience multiple co-occurring exposures to environmental pollutants, but analyses of multiple exposures have rarely been reported.

OBJECTIVES

We used latent class analysis to estimate co-exposures to multiple polycyclic aromatic hydrocarbons (PAH), and tested the associations of latent classes to body mass index.

METHODS

We analyzed National Health and Nutrition Examination Survey (NHANES) 2013-2014 data. The sample included 2354 people aged 6-80 years. Measures included seven urinary PAH metabolites, BMI, and demographic and behavioral covariates. People were classified into mutually exclusive latent classes characterized by unique profiles of multiple PAH exposures. These classes were used as categorical independent variables in weighted multiple regression models with BMI as the dependent measure. Models were analyzed overall and by age groups (6-19, 20-59, and 60 and over.) We compared results using latent classes to results using a summed PAH exposure measure.

RESULTS

Five latent classes were identified. Two of these classes were significantly associated with higher BMI overall (p < .0001) and for the two youngest age groups. One of these classes was characterized by high multiple exposures across all PAHs, and one by moderate exposures but relatively high naphthalene and phenanthrene. The summed PAH score was associated with higher BMI only for the youngest age group.

CONCLUSIONS

Persons experience multiple co-exposures to PAHs that are related to BMI and obesity across age groups. Latent class analysis provides information on higher order interactions among multiple chemicals that a summed score does not. Future work may apply this approach to other outcomes or types of co-exposures.