Low-level exposure to arsenic in drinking water and risk of type 2 diabetes: a systematic review and meta-analysis

The association between diabetes and low-level arsenic exposure in drinking water has not been adequately investigated. We conducted a two-stage meta-analysis of dose-response relationships from studies reporting incidence or prevalence of diabetes with average exposure up to 150 µg/L arsenic in drinking water. We conducted a literature search and included cross-sectional, case control and cohort studies and we extracted results for exposure to arsenic up to average concentration of 150 µg/L water. Prevalence and incidence were generically referred to as risk. We conducted study-specific linear regression to estimate the relative risk (RR) of diabetes for a 10 µg/L increase of arsenic, and combined the study-specific results in a random-effect meta-analysis. We addressed publication bias by excluding studies with the largest standard error; and identified a subset of high-quality studies. We also performed a one-stage meta-regression. We selected 11 studies including 12 populations from seven countries. The meta-analysis of their results yielded a relative risk (RR) of type 2 diabetes for a 10 µg/L arsenic increase within the range 0–150 µg/L equal to 1.07 (95% confidence interval [CI] 1.02–1.21, p heterogeneity = 0.006). There was evidence of publication bias (p = 0.004); after accounting for publication bias, the summary RR was 1.02 (95% CI 0.99–1.06, eight studies). The RR based on the high-quality studies was 1.03 (95% CI 0.97–1.09, five studies). The corresponding RR of the meta-regression was 1.007 (95% CI 0.998–1.017, 32 risk estimates). Accounting for publication bias and study quality, and considering heterogeneity of risk estimates, our two-stage meta-analysis supports a null association between low levels of arsenic in drinking water and type 2 diabetes.