The morbidity and mortality rates of prostate cancer (PC) have been increasing over the years. Screening tests for a prostate-specific antigen (PSA) are essential for early diagnosis. However, differentiation between PC and benign prostatic hyperplasia by a screening test may be difficult. In patients with false-negative and false-positive PSA measurements, oversight, overdiagnosis and overtreatment of PC may occur as disadvantages. Most of the patients were admitted to the department of urology owing to lower urinary tract symptoms and underwent prostate biopsies to differentiate between PC and other diseases. We examined the merits of PSA measurements in those with suspected PC using ROC curve analysis and Bayesian theorem (BT). ROC curve analysis revealed that the cutoff level of PSA was 15.0 ng/mL and the AUC cutoff was 0.73. When the virtual pretest probability is 50%, the post-test probability is improved to 85% in the calculation based on BT. However, when the cutoff level was set at 4.0 ng/mL in those patients, there was no improvement in the post-test probability compared with the pre-test probability. We consider that it may be necessary to determine the cutoff level for clinical medicine apart from that for preventive medicine. If we choose an approach oriented toward preventive medicine for the young and toward practical medicine for the old, PSA measurement with a suitably chosen cutoff level may increase the diagnostic accuracy of PC in clinical and preventive medicines. The analytical use of the Bayesian theorem may enable us to realize the actual power of PSA and, in the future, we expect that the Bayesian theorem will become widely used as one of the powerful methods in many fields.