We examined cancer incidence among 14,407 men and women who were enrolled in the National Health and Nutrition Survey I in the early 1970s and then followed through 1987. We studied 657 male and 593 female cancer cases, using Cox regression. Analyses were conducted for all cancers, lung, colorectal, breast, and prostate cancer. Analyses focused on diabetes, cholesterol, pulse, and physical activity, four risk factors with limited or inconsistent prior evidence. All four risk factors were modestly associated with all cancers for men but not for women. For diabetic men, the rate ratio for all cancers was 1.38 [95% confidence interval (CI) = 1.00-1.91]; the elevated risk was particularly evident for colorectal and prostate cancer. Slight inverse trends of cancer risk with cholesterol were apparent for men but not for women and were diminished compared to prior analyses of these data with less follow-up. Males with the lowest quartile of cholesterol versus the highest had a rate ratio of 1.21 (CI = 0.98-1.51) for all cancers. A modest positive trend between pulse and all cancers was seen for males [rate ratio of 1.27 (CI = 1.04-1.57)] for the highest versus the lowest quartile). The rate ratio for men with the least amount of nonrecreational physical activity was 1.29 (CI = 0.99-1.69). There is some evidence in these data that findings for cholesterol and nonrecreational physical activity could be artifacts of the early effects of disease because they diminished when cases were restricted to those with longer follow-up.