Although a number of obesity-related variables are recognized risk factors for NIDDM, few studies have addressed which one is the best predictor. A cohort of 721 Mexican Americans aged 25-64 years who were free of NIDDM at baseline were followed for an average of 7.2 years; 105 new cases of NIDDM were diagnosed. Body weight, body mass index (BMI), waist and hip circumferences, waist/hip ratio (WHR), triceps and subscapular skinfolds were all positively predictive of NIDDM independent of age and sex. There were modest to strong correlations between these anthropometric variables, however, waist circumference was the strongest predictor of NIDDM. The predictive power of a single measurement of waist circumference was at least equal to that of WHR and BMI combined. The risk of NIDDM for those in the highest quartile of waist circumference was 11 times greater than for those in the lowest quartile (95% confidence interval: 4.2-28.8). The waist-NIDDM relation was stronger in subjects with BMI < or = 27 kg/m2 (OR: 6.0 for a 1 SD difference) than in subjects with BMI > 27 kg/m2 (OR: 1.7 for a 1 SD difference). In multivariate analysis, waist circumference was the only significant predictor of NIDDM in models that included other anthropometric variables either separately or simultaneously. WHR and BMI were independent predictors of NIDDM after adjustment for each other, however, their predictive abilities disappeared after adjustment for waist circumference. The data indicate that waist is the best obesity-related predictor of NIDDM. This finding suggests that the distribution of body fat, especially abdominal localization, is a more important determinant than the total amount of body fat of the development of NIDDM in Mexican Americans.