Water Quality and Chronic Disease: Arsenic Exposure and EHR-Based Target Trial Emulations

North Carolina’s coastal plain has some of the highest naturally occurring arsenic concentrations in eastern U.S. groundwater, and a large share of rural residents rely on unregulated private wells with no required testing. This project uses large-scale UNC Health EHR data and target trial emulation to estimate the causal effects of that exposure on multiple chronic disease outcomes, in collaboration with NC TraCS and UNC Gillings.

Arsenic and Bladder Cancer. With Eaves (co-first author), Pfaff, Fry, and collaborators at UNC Gillings and NC TraCS, this arm emulates a target trial to estimate the effect of private well-water arsenic exposure on bladder cancer incidence. Bladder cancer is one of the best-documented downstream effects of chronic arsenic exposure, but causal estimates from real-world U.S. populations remain sparse.

Arsenic and Type 2 Diabetes. This arm applies the same target trial framework to incident type 2 diabetes, using the same NC well-water arsenic exposure data and UNC Health OMOP cohort. The study population is NC residents with greater than 90% estimated private well water use, and exposure is defined relative to the EPA maximum contaminant level (5 ppb) as well as percentile-based thresholds. Outcome ascertainment uses a multi-source confirmation algorithm combining inpatient diagnoses, outpatient diagnoses, HbA1c laboratory values, and antihyperglycemic medications. Estimation follows the same IPTW, IPCW, and g-computation pipeline as the bladder cancer study, with a 90-day lag period added to guard against differential diagnostic workup at cohort entry. Arsenic has been linked to impaired insulin signaling in laboratory models; this study provides population-level causal evidence from a high-exposure U.S. setting.