OBJECTIVE: To test whether a higher fasting insulin concentration is associated with a lower rate of weight gain over six to seven years. DESIGN: Two longitudinal epidemiologic cohorts including blacks and whites. SUBJECTS: The Coronary Artery Risk Development in Young Adults (CARDIA) Study examined subjects aged 18-30 y in 1985-86 and 1992-93 (n = 3536), and the Atherosclerosis Risk in Communities (ARIC) Study examined subjects aged 45-64 y in 1987-89 and 1993-95 (n = 11,179). MEASUREMENTS: In each study, fasting insulin at baseline and weight change during follow-up were measured in participants without diabetes. RESULTS: In whites and black men in CARDIA, there was a positive age-adjusted association between baseline insulin and weight change, although weight change was not entirely monotonic across the insulin quartiles. In these race-gender-groups, the linear regression coefficients indicated that each 50 pmol/L increment of baseline insulin was associated (P < 0.05) with approximately 0.10 kg/y greater rate of weight gain (95% confidence intervals (CI) for this estimate in kg/y were 0.023-0.187 for white women, 0.015-0.150 for black men, and 0.011-0.158 for white men). The association was eliminated entirely with adjustment for baseline weight. In contrast, among whites and black women in ARIC, the association was negative, with the linear regression coefficients suggesting that each 50 pmol/L higher fasting insulin concentration was associated (P < 0.05) with a 0.03-0.10 kg/y lower rate of weight gain (95% CI for this estimate in kg/y were -0.133 to -0.061 for black women, -0.106 to -0.054 for white women, and -0.055 to -0.009 for white men). This finding was generally strengthened by adjustment for baseline body mass index (BMI). CONCLUSIONS: A higher fasting insulin concentration is associated modestly with a lower rate of weight gain in ARIC, but not in CARDIA.