All of the analyses were adjusted for sex, age, race, and study center. A second set of models was additionally adjusted for baseline WHR, BMI, or waist circumference in the analyses of change in WHR, BMI, or waist circumference, respectively. A third set of models was also adjusted for baseline BMI or WHR, smoking status, years of cigarette smoking, change in smoking status during follow-up, serum glucose level, systolic blood pressure, serum total cholesterol level, the presence of coronary heart disease or diabetes, physical activity level, fibrinogen level, and white blood cell count, all measured at baseline. If one of the variables was used for stratification, it was no longer included in the model. Subgroup analyses were performed after stratifying by sex, age group, weight group, diabetes status, smoking status, and race. Interaction terms for each of the stratification variables with change in WHR were also tested in the logistic regression models. All of the analyses were performed using Stata/SE version 8.0 statistical software (Stata Corp, College Station, Texas).