A Multipollutant Longitudinal Study of the Association between Urinary Tungsten and Incident Diabetes in a Rural Population

Background/Aim: Tungsten (W) is a transitional metal found in soil, water, food, and air. Cross-sectional analyses indicate that W is associated with fasting glucose (FG) levels. We conducted the first longitudinal analysis assessing the association between W and incident diabetes, accounting for arsenic (As), lead (Pb), and cadmium (Cd) coexposure. Methods: We analyzed data from 1684 individuals residing in rural Colorado (USA) and participating in the San Luis Valley Diabetes Study. Participants’ diabetes status, FG (mg/dL), and urinary concentrations of W, As, Pb, Cd, and creatinine were assessed between one and seven times over 14 years. Metal concentrations (μg/L) and FG were natural log-transformed. We assessed longitudinal associations between lnW exposure and lnFG using linear mixed effect models with a random intercept for each participant. We assessed the association between baseline lnW exposure and incidence rate of diabetes using Fine and Gray competing risk regression (competing event=all-cause mortality). An evidenced-based directed acyclic graph (DAG) was used to determine covariates (age, ethnicity (Hispanic/not), sex, education (<12/12/>12 years), smoker status (current/former/never), alcohol intake (g/week), carbohydrates (g/day), body mass index, urine creatine (μg/g), lnAs, lnPb, and lnCd). Results: At the first examination, 48% of participants were Hispanic, 53% were female, the mean age was 54 years (standard deviation=12), the median W concentration was 0.22 μg/L (25th =0.20; 75th=0.59), and the median FG level was 99 mg/DL (25th=92; 75th =115). Of 1387 participants without diabetes at baseline (mean time at-risk=9.2 years), 398 individuals developed diabetes (incidence=3 cases/100 person-years) and 301 died before developing diabetes. LnW exposure was not associated with lnFG (0.1% decrease in lnFG for a doubling in lnW; 95%CI= 0.2% decrease-0.0% increase) or the incidence rate of diabetes (subdistribution hazard ratio=1.02, 95%CI=0.90,1.15). Conclusion: Accounting for metal coexposures, competing risks, and DAG-determined covariates, we observed no association between W and diabetes.