Jessica Madrigal

PhD, MS

Jessica Madrigal

Research Fellow

PLCO (Learn more about this study)

PLCO-1259

Jun 22, 2023

Characterizing the independent and joint contributions of the chemical, non-chemical, and social-structural environment on cancer incidence and mortality in the PLCO Study

In 2018, 21,557 industrial facilities across the U.S. reported annual releases of 600 million pounds of toxic chemicals into the air (United States Environmental Protection Agency 2020), including known and probable carcinogens such as benzene, dioxins and dioxin-like compounds, and metals (International Agency for Research on Cancer 2020). In addition to potential exposure to carcinogens, point sources and their emissions may impact other factors that are intermediates in the etiologic pathway for cancer such as inflammation (Cakmak et al. 2020; Olvera Alvarez et al. 2018; Shacter and Weitzman 2002; Thompson et al. 2015) and/or epigenetic changes (Chappell et al. 2016; Herceg and Vaissière 2011). These exposures could be important contributors to carcinogenic exposures for individuals living in proximity to industrial facilities. However, epidemiologic evaluation of cancer risks related to many of these exposures at emissions levels in the ambient environment remains limited. Moreover, few studies have considered how environmental pollution and social-structural barriers may act together to contribute to disparities in cancer outcomes, despite it being known that these factors are often interrelated. The environment can be considered as consisting of chemical pollutants (e.g., toxic industrial emissions of asbestos, benzene, cadmium; regulated criteria air pollutants, including fine particulate matter), the non-chemical and built environment (e.g., noise, availability of primary and specialized health care, food stores, outdoor space for physical activity), and social-structural environment (e.g., residential segregation, isolation, unemployment).

In this study, we propose to summarize probable exposure to multiple chemical, non-chemical, and social-structural factors experienced by participants. We will evaluate the association of point source carcinogenic air emissions with cancer risk and all-cause and cause-specific mortality and consider the non-chemical and social-structural factors as potential confounders, modifiers, or exposure determinants in these analyses. We will use geographic information systems technology combined with data from the U.S. Census, U.S. Environmental Protection Agency, other federal agencies, and proprietary business lists to develop a novel set of factors that can be quantified to characterize conditions that may impact cancer etiology and mortality. Most of these data have already been linked to participant addresses in PLCO (e.g., PM2.5 exposures, EPA’s Toxics Release Inventory (TRI) facility locations). We anticipate working with IMS to generate specific distance-based exposure metrics from TRI based on participant addresses.

Aim 1: To describe the industrial environmental exposures, specifically to known and probable carcinogenic emissions, and social-structural factors experienced by adults across racial/ethnic and socioeconomic groups in PLCO.

Aim 2: To evaluate individual and joint associations of environmental and structural factors with incidence of bladder, breast, colorectal, gastric, hematopoietic, liver, lung, pancreas, prostate, and kidney cancer, and cancer mortality in PLCO.

Statistical Analysis

In Aim 1 (descriptive analyses), we will use the inverse distance weighted sum of each chemical (pounds/km2) as the exposure metric. To describe the relationship between exposure to chemical emissions and participants characteristics of race/ethnicity and SES (e.g., educational attainment), we will generate basic descriptive statistics to evaluate distributions of each chemical by race/ethnicity and indicators of SES. Chi-square or Fisher’s exact tests will be used to compare the proportions of minority/lower SES populations exposed to different levels of each exposure factor. We will evaluate the correlation structure for all the exposures among the various race/ethnicity and SES groups. We will construct separate exposure-specific regression models to evaluate the cross-sectional associations between race/ethnicity and SES and each pollutant exposure.

For Aim 2 (risk analyses), we will use Cox regression to estimate hazard ratios (HRs) and 95% confidence intervals separately for each chemical and outcome (e.g., any incident breast cancer). Where power allows, we will evaluate relationships by histologic subtype. Models will be adjusted for a set of standard (e.g., age at study entry, educational attainment, state of residence) and outcome-specific covariates. For example, analyses of breast cancer, we will evaluate parity, age at first live birth, age at menarche, menopausal status, use of oral contraceptives or hormone therapy, family history of breast or ovarian cancer (first degree relative), body mass index, smoking status, alcohol intake, census tract level annual averages of criteria air pollutants, and neighborhood socioeconomic status as potential confounders. Where power allows, we will evaluate associations for breast cancer by hormone receptor (ER, PR) status and extent of disease (DCIS vs. invasive). We will include analyses stratified by 1) race and ethnicity and 2) individual and neighborhood SES and formally test for interaction with Wald or likelihood ratio tests.

Jared A. Fisher, Occupational and Environmental Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute
Rena R. Jones, Occupational and Environmental Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute

1