We examined the association of serum alanine aminotransferase (ALT) with features of the metabolic syndrome and whether it predicted incident diabetes independently of routinely measured factors in 5,974 men in the West of Scotland Coronary Prevention Study. A total of 139 men developed new diabetes over 4.9 years of follow-up. ALT, but not aspartate aminotransferase, levels increased progressively with the increasing number of metabolic syndrome abnormalities from (means +/- SD) 20.9 +/- 7.6 units/l in those with none to 28.1 +/- 10.1 units/l in those with four or more (P < 0.001). In a univariate analysis, men with ALT in the top quartile (ALT >/=29 units/l) had an elevated risk for diabetes (hazard ratio 3.38 [95% CI 1.99-5.73]) versus those in the bottom quartile (<17 units/l). ALT remained a predictor with adjustment for age, BMI, triglycerides, HDL cholesterol, systolic blood pressure, glucose, and alcohol intake (2.04 [1.16-3.58] for the fourth versus first quartile). In stepwise regression, incorporating ALT and C-reactive protein (CRP) together with metabolic syndrome criteria, elevated ALT (>/=29 units/l), and CRP (>/=3 mg/l) predicted incident diabetes, but low HDL cholesterol and hypertension did not. Thus, elevated ALT levels within the "normal" range predict incident diabetes. The simplicity of ALT measurement and its availability in routine clinical practice suggest that this enzyme activity could be included in future diabetes prediction algorithms.