We previously reported a strong positive association between vasectomy and the risk of prostatic cancer that arose in multiple comparisons made within data collected from 1976 to 1988 in an ongoing hospital-based surveillance study of many exposures and diseases. We have reassessed this association with data collected in the surveillance study during 1988-1992 from a new set of patients (355 cases of prostatic cancer and 2,048 controls with nonmalignant conditions). Because some studies have reported increased relative risks of lung cancer and testicular cancer in vasectomized men, we also used the surveillance database (4,126 men with various cancers, 7,027 men with nonmalignant conditions) to assess the relation of vasectomy to the risk of these and other cancers. In the newly collected data, the multivariate relative risk estimate for prostatic cancer in vasectomized men was 1.2 (95% confidence interval (CI) 0.6-2.7). For lung cancer and testicular cancer, the relative risk estimates were 1.3 (95% CI 0.8-2.1) and 0.8 (95% CI 0.4-1.9), respectively; for lung cancer occurring > or = 15 years after vasectomy, the relative risk estimate was 1.9 but it was not statistically significant (95% CI 0.7-5.0). For pancreatic cancer, the relative risk estimate was 1.8 (95% CI 1.0-3.1). For each of the other cancers considered--malignant melanoma, large bowel cancer, bladder cancer, kidney cancer, lymphoma, leukemia, and other cancers--the relative risk estimate was 1.3 or less and compatible with a value of 1.0. The present data provide little support for an association of vasectomy with the risk of prostatic cancer or other cancers. In addition, the data from two sets of cases of prostatic cancer and controls interviewed consecutively illustrate that increased relative risks detected in screening for statistically significant associations may tend to have an upward bias and to be lower in subsequent data.