Abstract Background Prostate cancer screening depends on a careful balance of benefits, in terms of reduced prostate cancer mortality, and harms, in terms of overdiagnosis and overtreatment. We aimed to estimate the effect on overdiagnosis of restricting prostate specific antigen (PSA) testing by age and baseline PSA. Methods Estimates of the effects of age on overdiagnosis were based on population based incidence data from the US Surveillance, Epidemiology and End Results database. To investigate the relationship between PSA and overdiagnosis, we used two separate cohorts subject to PSA testing in clinical trials (n = 1,577 and n = 1,197) and a population-based cohort of Swedish men not subject to PSA-screening followed for 25 years (n = 1,162). Results If PSA testing had been restricted to younger men, the number of excess cases associated with the introduction of PSA in the US would have been reduced by 85%, 68% and 42% for age cut-offs of 60, 65 and 70, respectively. The risk that a man with screen-detected cancer at age 60 would not subsequently lead to prostate cancer morbidity or mortality decreased exponentially as PSA approached conventional biopsy thresholds. For PSAs below 1 ng/ml, the risk of a positive biopsy is 65 (95% CI 18.2, 72.9) times greater than subsequent prostate cancer mortality. Conclusions Prostate cancer overdiagnosis has a strong relationship to age and PSA level. Restricting screening in men over 60 to those with PSA above median (>1 ng/ml) and screening men over 70 only in selected circumstances would importantly reduce overdiagnosis and change the ratio of benefits to harms of PSA-screening.