T he role of the liver in the pathogenesis of type 2 diabetes is attracting increasing interest. In a recent study (1), directly determined liver fat content was shown to correlate with several features of insulin resistance in normal weight and moderately overweight subjects independent of BMI and intraabdominal or overall obesity. However, direct measurements of liver fat require ultrasound, computed tomography scan, or proton spectroscopy, and such techniques are unlikely to be recommended for this purpose in routine clinical practice. Fortunately, circulating concentrations of a number of variables appear to give insight into the extent of liver fat accumulation. Among these are ␥-glutamyltransferase, alanine aminotransferase (ALT), and aspartate aminotransferase (AST). Of these three, ALT is the most specific marker of liver pathology and appears to be the best marker for liver fat accumulation (2). In addition, circulating concentrations of plasminogen activator inhibitor-1 may give insight into the extent of liver fat content (3) but, unlike ALT, its measurement is perhaps not as simple, standardized, or routinely available in laboratories.In light of the above observations, it is of interest that ALT has been shown to predict incident type 2 diabetes in two prospective studies (4,5). Ohlson et al. (4) determined risk factors for diabetes in 766 men, 47 of whom developed diabetes over 13.5 years of follow-up. They reported that elevated ALT predicted diabetes independently of classical predictors inclusive of BMI, blood pressure, triglycerides, and family history. However, diabetes ascertainment was not uniform in their study, and the potential utility of ALT to predict diabetes was not examined in any detail. Vozarovoa et al. (5) examined the ALT levels in a cohort of 370 Pima Indians with normal glucose tolerance, 63 of whom developed diabetes over an average follow-up of 6.9 years. In their analyses, individuals in the top decile for ALT (Ն70 units/l) had a relative risk of 2.5 (95% CI 1.7-3.7) for diabetes compared with those in the bottom decile (ALT Յ12 units/l), with adjustment for age, sex, percentage body fat, clamp-derived insulin resistance, and acute insulin response. The only minor weaknesses in that study were the absence of data on alcohol consumption and routine clinical measures known to predict diabetes, namely fasting lipids and blood pressure. Moreover, they