Data from several epidemiologic studies suggest that exposure to unusually high amounts of dioxin (2,3,7,8-tetrachlorodibenzo-p-dioxin) increases the risk of diabetes mellitus, and experimental data suggest that the mechanism for this is decreased cellular glucose uptake. To investigate the dose-response relation more closely, we examined the association of serum dioxin level with prevalence of diabetes mellitus and with levels of serum insulin and glucose among 1,197 veterans in the Air Force Health Study who never had contact with dioxin-contaminated herbicides and whose serum dioxin level was within the range of background exposure typically seen in the United States (< or =10 ng/kg lipid). Compared with those whose serum dioxin level was in the first quartile (<2.8 ng/kg lipid), the multivariate-adjusted odds of diabetes among those in the highest quartile (> or =5.2 ng/kg lipid) was 1.71 (95% confidence interval = 1.00-2.91). The association was slightly attenuated after adjustment for serum triglycerides. Whether adjustment for serum triglycerides was appropriate, however, cannot be determined with available data. The association of background-level dioxin exposure with the prevalence of diabetes in these data may well be due to reasons other than causality, although a causal contribution cannot be wholly dismissed.