Increased abdominal obesity has been related to lower insulin sensitivity (SI), independent of overall obesity, but it has been suggested that this relationship may be weaker in non-whites. In the Insulin Resistance and Atherosclerosis Study (IRAS), SI was estimated using a minimal model analysis of the frequently sampled intravenous glucose tolerance test in 1,625 men and women aged 40-69 years. Subjects included African-Americans, Hispanics, and non-Hispanic whites from Oakland and Los Angeles, CA, San Antonio, TX, and the San Luis Valley, CO. Minimum waist circumference was significantly (P = 0.0001) associated with SI after adjusting for age, sex, height, BMI, glucose tolerance status, ethnicity, and clinic. This relationship was significantly (P = 0.0001) stronger in subjects with normal glucose tolerance (NGT) (beta = -0.030, P = 0.0001) than in those with impaired glucose tolerance (IGT) (beta = -0.010, P = 0.02; NIDDM: beta = -0.013, P = 0.0001). There were no significant ethnic differences in effect size across the spectrum of glucose tolerance. Waist circumference was also positively related to fasting insulin, an indirect measure of insulin sensitivity, in NGT (P = 0.0001), IGT (P = 0.0003), and NIDDM (P = 0.0002). The waist-fasting insulin relationship was significantly weaker in African-Americans, relative to non-Hispanic whites, in NGT and IGT (tests of statistical interaction: P = 0.04 and P = 0.02, respectively). In general, these patterns were similar in models specifying waist-to-hip ratio (WHR), rather than waist circumference, as the independent variable. While some ethnic variability exists, a negative relationship between abdominal obesity and insulin sensitivity was confirmed for all three ethnic groups across the spectrum of glucose tolerance.