Serum gamma-glutamyl transferase (GGT), a marker of oxidative stress, has been shown to be associated with diabetes mellitus in some population-based studies, but not all. Also, it is not clear if there is a continuous dose-response relationship in this association, or if this association is evident only beyond a particular threshold level of GGT. We examined the association between serum GGT and diabetes mellitus in a representative sample of US adults aged ≥20 years, in a cross-sectional study involving 7,976 National Health and Nutrition Examination Survey 1999–2002 participants. Diabetes mellitus was defined as a fasting glucose ≥ 126mg/dl, nonfasting glucose ≥200 mg/dl, or use of oral hypoglycemic medication or insulin (n = 805). Higher serum GGT levels were positively associated with diabetes mellitus, independent of, alcohol consumption, body mass index, hypertension and other confounders. Multivariable odds ratio (95% confidence interval) comparing quartile 4 of GGT (>33 U/L) to quartile 1 (<15U/L) was 2.33 (1.59–3.41), P-trend < 0.0001. This association persisted in separate analysis among men and women. In nonparametric models, the positive association between serum GGT and diabetes appeared to be present across the full range of GGT, without any threshold effect. Higher serum GGT levels are positively associated with diabetes mellitus.