Exposomic and Metabolomic Approaches to Understanding Cancer Risk in Females: A Secondary PLCO Study
Principal Investigator
Name
Jordan Kuiper
Degrees
Ph.D., M.S.
Institution
The George Washington University
Position Title
Assistant Professor of Environmental and Occupational Health, tenure-track
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-2027
Initial CDAS Request Approval
Mar 23, 2026
Title
Exposomic and Metabolomic Approaches to Understanding Cancer Risk in Females: A Secondary PLCO Study
Summary
Per- and polyfluoroalkyl substances (PFAS) are persistent endocrine-disrupting chemicals widely detected in human serum and increasingly linked to breast cancer risk, yet most epidemiologic studies have evaluated individual PFAS compounds rather than real-world co-exposures. Humans are concurrently exposed to PFAS and other xenobiotic chemicals (e.g., pesticides, industrial pollutants, persistent organic contaminants), raising concern for joint effects on metabolic processes relevant to carcinogenesis. Mechanistic and metabolomic studies demonstrate that PFAS and xenobiotic compounds perturb lipid metabolism, mitochondrial function, and redox balance, pathways that are central to cellular energy homeostasis and are rewired during breast tumor initiation and progression. Despite compelling biologic plausibility, few human studies have jointly evaluated PFAS with co-occurring xenobiotics or integrated biochemical intermediates such as metabolomic pathways that reflect carcinogenic vulnerability. Recent analyses in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Screening Trial demonstrate associations between PFAS exposure and breast cancer within a large U.S. cohort with stored biospecimens, underscoring the feasibility and relevance of studying endocrine-disrupting mixtures in this population. Together, these findings support a conceptual framework in which PFAS and xenobiotic mixtures influence oxidative, lipid, and mitochondrial pathways that shape breast cancer risk, while highlighting critical gaps in mixture-focused, mechanistically informed epidemiologic research.
Aims
Aim 1: Develop metabolomic pathway scores capturing oxidative stress, lipid remodeling, and mitochondrial energy metabolism among cancer-free women in the Prostate, Lung, Colon, and Ovary (PLCO) Screening Trial.
Approach 1: Using LC-MS metabolomics data from cancer-free female participants in PLCO, we will develop pathway scores for oxidative stress, lipid remodeling, and mitochondrial function. We will combine biologically curated metabolite sets informed by mechanistic literature and deep learning approaches to derive robust pathway scores via a PLCO-specific metabolomic aging clock. Internal validation will include split-sample and cross-validation procedures to assess stability, variance explained, and associations with known correlates. We will implement this as a reproducible analytic “tool” that can be applied to other PLCO sub-studies.
Aim 2: Evaluate whether PFAS and xenobiotic mixture burden and metabolomic pathway scores derived from pre-diagnostic serum are associated with future breast cancer case status and whether pathway scores improve risk discrimination beyond established risk factors in PLCO.
Sub-aim 2a: Estimate mixture-breast cancer associations.
Approach 2a: Use quantile-based g-computation (qgcomp) to estimate the overall OR per quantile increase in the joint mixture and Bayesian kernel machine regression (BKMR) to evaluate non-linearity and interactions among chemicals. Adjust for demographic, behavioral, and biospecimen-related confounders and account for the biospecimen sampling design using design-consistent methods (e.g., inverse probability weighting or appropriate adjustment for matching/sampling factors). Lag-stratified analyses and interaction tests with time-to-diagnosis will be conducted as sensitivity analyses to assess robustness and potential influence of preclinical disease processes.
Sub-aim 2b: Evaluate added predictive value of pathway scores.
Approach 2b: Fit baseline models using established risk factors, then evaluate incremental performance when adding pathway scores (e.g., cross-validation AUC/ΔAUC).
Aim 3: Evaluate how PFAS and xenobiotic mixtures influence breast cancer risk by identifying metabolomic pathway-based latent states and estimating marginal causal effects in PLCO.
Sub-aim 3a: Identify latent metabolic states linking environmental mixtures to cancer risk.
Approach 3a: Using PFAS, xenobiotic chemical, and LC-MS metabolomics data from PLCO nested case-control studies of breast cancers, we will apply latent unknown clustering integrating multi-omics data (LUCID) models to jointly characterize relationships among chemical mixtures, metabolomic signatures, and cancer outcomes. LUCID will estimate how mixture burden influences latent state membership and how latent states relate to cancer risk, adjusting for demographic, lifestyle, and biospecimen-related covariates.
Sub-aim 3b: Estimate marginal causal effects of mixtures on cancer risk.
Approach 3b: We will estimate causal effects of PFAS and xenobiotic mixtures on cancer risk by utilizing targeted maximum likelihood estimation (TMLE), which accommodates flexible machine learning for confounding adjustment and supports two-phase sampling designs via inverse-probability weighting. Exploratory analyses will incorporate latent metabolic state membership as an intermediate biomarker to explore pathway-specific causal relevance.
Collaborators
Jordan Kuiper Milken Institute School of Public Health at The George Washington University
Cindy Weinman Milken Institute School of Public Health at The George Washington University
Alexander P. Keil National Cancer Institute