Long-term exposure to PM2.5 constituents and incident cancer among Medicare beneficiaries in the USA: a national cohort study
Menée à partir de données portant sur 15 138 652 bénéficiaires du programme Medicare (durée médiane de suivi : 9 ans), cette étude analyse l'association entre 15 constituants des PM2.5, leurs sources et l’incidence de 5 cancers (poumon, prostate, sein, endomètre, côlon-rectum)
Background: The health burden from cancer has markedly increased over the past 20 years. Total PM2·5 has been identified as an environmental risk factor for cancer. However, how each constituent of PM2·5 contributes to the development of cancer is largely unknown. We aimed to investigate the association between 15 PM2·5 constituents, sources of PM2·5, and the incidence of five common cancers among adults aged 65 years and older in the USA.
Methods: For this national cohort study, we included beneficiaries of Medicare in the USA, aged 65 or older, who were followed for various time periods between Jan 1, 2000, and Dec 31, 2018. The concentrations of 15 PM2·5 constituents in the contiguous USA were estimated using ensemble machine learning models, and the sources of PM2·5 were identified from the PM2·5 constituent data using non-negative matrix factorisation. The outcomes were incident lung, colorectal, prostate, breast, and endometrial cancers, for which information was extracted from the Medicare Chronic Conditions Data Warehouse database. Associations between PM2·5 constituents and the outcomes were investigated using a generalised weighted-quantile sum regression model with the Cox-approximate Poisson method, and associations between PM2·5 sources and outcomes were investigated using Cox-approximate Poisson regression models.
Findings: 15 138 652 Medicare beneficiaries were included in the study, with a median follow-up period of 9 years (IQR 5–15). The incidences per 1000 person-years were 15·9 for lung cancer, 18·9 for colorectal cancer, 73·5 for prostate cancer, 50·7 for breast cancer, and 11·4 for endometrial cancer. Each one-decile increase in exposure to the PM2·5 mixture was associated with incident rate increases of 6·4% (95% CI 5·9 to 6·9) for lung cancer, 4·3% (3·8 to 4·8) for colorectal cancer, 3·6% (3·1 to 3·9) for prostate cancer, and 2·0% (1·7 to 2·4) for breast cancer. No association was observed for endometrial cancer (0·4% [−0·4 to 1·1]). Of the PM2·5 constituents, vanadium contributed the largest relative weight in the observed associations (ranging from 23·7–36·1%). Increased incidence of all five cancers was associated with PM2·5 sourced from fuel oil combustion (14·1% [95% CI 6·5–22·3] for lung cancer, 15·6% [7·5–24·2] for colorectal cancer, 18·4% [5·2–33·4] for prostate cancer, 9·2% [7·9–10·5] for breast cancer, and 5·6% [2·9–8·3] for endometrial cancer for each 1
μg/m3 increase in concentration) and with PM2
·5 sourced from coal combustion (6·9% [3·3–10·7] for lung cancer, 9·2% [6·6–11·9] for colorectal cancer, 7·3% [3·6–11·1] for prostate cancer, 6·2% [3·2–9·3] for breast cancer, and 3·3% [2·2–4·3] for endometrial cancer).
Interpretation: PM2·5 exposure was associated with an increased risk of cancer, with vanadium—a marker of fuel oil combustion—contributing the largest weight in the observed association. Future PM2·5 regulations should consider targeting the constituents and sources that have the largest health effects.
The Lancet Planetary Health , article en libre accès, 2025