The ability of serum prostate-specific antigen (PSA) and PSA density (PSAD) to distinguish patients with prostate cancer from those with benign diseases of the prostate was assessed in 495 men. All men were evaluated with PSA determination, digital rectal examination (DRE), transrectal ultrasonography (TRUS) and ultrasound-guided prostatic biopsies. PSA was analysed by the polyclonal (Yang) assay. Prostate volume was estimated from TRUS. PSAD was determined by dividing the serum PSA by the volume of the prostate. Prostatic biopsies identified cancer in 246 of the 495 patients (49.7%). The entire group was divided into 6 subgroups according to PSA level at presentation. Cancer and noncancer patients were compared in each subgroup with respect to the values of PSA, prostate volume and PSAD. For the entire group of patients, there was no statistically significant advantage, for PSAD over serum PSA alone, in distinguishing between benign and malignant prostatic conditions. However, when patients were stratified according to PSA level, PSAD was statistically significantly superior to serum PSA alone in the detection of prostate cancer for PSA values in the intermediate range (2.6-30 ng/ml). This analysis with respect to the DRE and TRUS results showed PSAD to be superior to PSA when both examinations are normal. Our results demonstrate that the influence of PSAD level on cancer detection proportionally increases as the PSAD value increases. Curves constructed from the incidence of prostate cancer according to PSAD values may be useful to select patients with intermediate levels of serum PSA, and normal DRE and TRUS for prostatic biopsies.