Studies of the potential effect of prostate specific antigen (PSA) screening on a less than yearly basis have been limited to computer simulations using relatively small sample sets. Primary clinical data on this relationship have not been generally available. We examined the relationship of less frequent testing and the risk of nonlocalized incurable cancer. The effect of testing frequency on the risk of prostate biopsy in men ultimately diagnosed with cancer was also assessed. The study included a population based sample of 36,422 men 65 years old or older residing in 9 geographic areas with newly diagnosed prostate cancer during 1989 to 1993. The primary end point was the risk of nonlocalized cancer, as determined by logistic regression. Patient age, geographic region, year of diagnosis and race were included as covariates. In men who would be diagnosed with prostate cancer the risk of nonlocalized cancer did not differ in those tested 2 or 3 years compared with the risk in those tested 1 year before diagnosis (relative risk 1.00, 95% confidence interval 0.84 to 1.20 and 1.02, 95% confidence interval 0.74 to 1.41, respectively). However, the risk of prostate biopsy in these men was directly related to the number of PSA tests performed (test for trend p = 0.0061). Patients who choose to undergo PSA testing may be tested on a biennial instead of annual basis without an increased risks of nonlocalized cancer. Decreasing the frequency of PSA testing may lead to fewer prostate biopsies.