Elevated fasting insulin is an independent risk factor for hyperlipidemia, hypertension, and cardiovascular disease, but determinants of insulin other than age and body mass remain poorly described. Potentially modifiable factors associated with insulin were identified by correlating anthropometric, dietary and physical activity data in the CARDIA cohort of 2643 black and 2472 white men and women aged 18-30 years. Insulin was positively correlated with serum glucose, body mass index (BMI), skinfold thickness, waist/hip ratio and sucrose intake, and negatively correlated with heavy physical activity score, treadmill exercise duration, and magnesium intake (each p less than 0.01). After adjustment for other covariates, the positive association of insulin with waist/hip ratio, skinfold thickness, and sucrose intake remained in the group as a whole, as did the negative associations with magnesium and treadmill duration. These relationships provide insight into potentially modifiable factors affecting insulin levels, and should be considered in interpreting associations between insulin levels and cardiovascular disease.