Identifying exposome-driven cancer incidence patterns using multi-cancer principal components analysis.

Abstract

e22606 Background: Populations are exposed to complex mixtures of environmental, behavioral, and social factors that influence multiple cancers simultaneously. While exposomic approaches have increasingly been applied to individual cancers, multicancer analyses leveraging shared, geographically patterned exposures remain underutilized. As many exposures cluster spatially and affect multiple malignancies, cancer incidence may reflect shared regional risk structures. Methods: Age-adjusted average incidence rates for 12 cancers from 2017–2021 were obtained from the NCI State Cancer Profiles. Counties with complete, non-suppressed data were included (n = 1567). Principal component analysis (PCA) was applied to the county-by-cancer incidence matrix to identify orthogonal multicancer incidence patterns. County-level exposures from national surveillance datasets were correlated with each principal component to characterize exposome signatures. Results: PCA identified 12 orthogonal components of multicancer incidence, with the first three explaining the majority of shared variation across counties. PC1 represented a general cancer burden axis dominated by smoking and behavior-associated malignancies. PC2 reflected cancer screening and healthcare access, distinguishing screen-detected from screen-preventable cancers.PC3 captured a metabolic–industrial exposure axis and showed the strongest association with recent county-level changes in cancer incidence (r = 0.59). Conclusions: Population-level cancer burden appears structured along a small number of orthogonal axes spanning healthcare access and screening, health behaviors, and environmental exposures. This exposome-informed, multicancer framework uses PCA to capture geographically patterned influences on cancer incidence and provides a scalable strategy for prioritizing exposures and markers for downstream causal and mechanistic investigation. Multicancer principal components and associated exposome signatures. PC PC Variance (%) Dominant Cancers (r) Key Correlated Exposures (r) Conceptual Axis PC1 25.2 Lung (0.40)Colon (0.34)Oral (0.34)Bladder (0.33) Carbon tetrachloride (0.37)Smoking (0.37)Obesity (0.30) General cancer burden PC2 16.4 Melanoma (0.45)Breast (0.41)Colon (−0.38) Median household income (0.60)Colonoscopy screening (0.53)Social Vulnerability Index (−0.52) Screening / access PC3 10.1 Prostate (0.63)Breast (0.36)Pancreatic (0.33) Diabetes (0.28)PM2.5 (0.24)Ethylene oxide (0.18) Metabolic–industrial Datasets: EPA National Air Toxics Assessment (NATA, 2025); US Geological Survey EPest (2013–2017); BRFSS (2018–2020); USDA Economic Research Service (2020); U.S. Census (2023); NCI State Cancer Profiles (2017–2021). Cancers in study: Bladder, Breast, Colon, Kidney, Leukemia, Lung, Lymphoma, Melanoma, Oral, Pancreatic, Prostate and Uterine.