Specifying the Associations between PM2.5 Constituents and Gastrointestinal Cancer Incidence: Findings from a Prospective Cohort Study in Beijing, China

This study aimed to test the association between PM2.5 and the incidence of gastrointestinal (GI) cancer, and further to detect the primary constituents on this association. A sum of 142,982 participants without GI cancer at baseline were derived from the National Urban Cancer Screening Program in Beijing (2013–2019). The 5 year averaged concentrations of PM2.5 mass and its five constituents, namely, black carbon (BC), ammonium (NH4+), nitrate (NO3–), organic matter (OM), and inorganic sulfate (SO42–), were estimated by using a hybrid machine learning model. The Cox proportional hazard model with fixed effects was used to explore the associations between PM2.5 mass and its constituents with the incidence of GI cancer. The double-exposure linear model, the mixture exposure model of quantile-based g-computation, and an explainable machine learning model were utilized to evaluate the importance of different PM2.5 constituents. Long-term exposure to PM2.5 mass and its constituents was linearly associated with GI cancer; the estimated hazard ratio and 95% confidence interval (95% CI) of per standard deviation increment were 1.367 (95% CI: 1.257 to 1.487) for PM2.5 mass, 1.434 (95% CI: 1.307 to 1.574) for BC, 1.255 (95% CI: 1.169 to 1.349) for NH4+, 1.217 (95% CI: 1.139 to 1.301) for NO3–, 1.410 (95% CI: 1.287 to 1.546) for OM, and 1.410 (95% CI: 1.288 to 1.542) for SO42–. By using multiple methods, results indicated that SO42– and BC were the most important constituents. Results indicated that long-term exposure to PM2.5 was associated with a high incidence of GI cancer, and BC and SO42– were robustly identified as the primary constituents.