An unfavourable body fat distribution has been associated with an increased prevalence and incidence of non-insulin dependent diabetes mellitus (NIDDM). The potential utility of assessing body fat distribution in diabetes screening, however, has not been assessed. We compared the impact of upper body fat distribution (assessed by the waist-to-hip ratio (WHR)) and body mass index (BMI) and NIDDM using the population attributable risk approach of Levin in 1965 Mexican Americans from the San Antonio Heart Study, a population-based study of diabetes and cardiovascular disease. The population attributable risk percentage (PAR%) was 52.0% for WHR compared to 43.4% for body mass index. After stratification by BMI, women with a high WHR had a PAR% of approximately 50% and men had a PAR% of 28-58%. For any given cutpoint (e.g. the 10th percentile, 20th percentile, etc.) of WHR used to screen for NIDDM, WHR had both a higher sensitivity and a lower false positive rate than the corresponding cutpoint of BMI. To evaluate the relative contribution of WHR in identifying prevalent cases of NIDDM, multiple logistic regression analyses were performed, and the number of subjects identified as being in the top 20% of the risk score distribution was compared using a model that included WHR and a model that included BMI. In men, BMI did not increase the sensitivity in detecting NIDDM subjects once age was accounted for; WHR increased the sensitivity only slightly. In women, sensitivity was enhanced modestly using both measures, although WHR again was the more sensitive method. These data suggest that WHR is a better single screening measure for NIDDM than BMI.