AIM: Cross-sectional studies have demonstrated a relationship between obesity and insulin sensitivity (S(I)); however, there is a lack of evidence from longitudinal studies. METHODS: The Insulin Resistance Atherosclerosis Study (IRAS) estimated S(I) (x10(-4)/min.microU/ml) directly using a frequently sampled intravenous glucose tolerance test with minimal model analysis in 504 normoglycaemic subjects. Partial correlation coefficients (r) were calculated to compare the relationship of change in S(I) from baseline to 5 years later (DeltaS(I)) with baseline waist circumference (waist) as a measure of abdominal obesity and body mass index (BMI) as a measure of overall obesity. Mean DeltaS(I) was -1.06 (SD = 1.85). RESULTS: Higher baseline waist (r = -0.16; p = 0.0005), but not BMI (r = -0.005; p = 0.91), was associated with (-) DeltaS(I) in models including sex, ethnicity, clinical centre and baseline S(I), BMI, waist, age and physical activity. The waist-DeltaS(I) relationship differed across the levels of baseline BMI, being significant only in normal weight (r = -0.21) and overweight subjects (r = -0.16), but not in obese subjects. DeltaS(I) was correlated with a 5-year change in either obesity measure (Deltawaist: r = -0.22 and DeltaBMI: r = -0.20; p = 0.0001). CONCLUSIONS: Among non-diabetics, waist circumference was a strong predictor of declining S(I) among lean subjects, a modest predictor among overweight subjects, but was not predictive among obese individuals. Waist circumference should be considered, in addition to BMI, when identifying individuals at high risk of diabetes or the insulin resistance syndrome.