BACKGROUND To evaluate whether serum total prostate-specific antigen (PSA), PSA density (serum total PSA level divided by prostate volume), gamma-seminoprotein and gamma-seminoprotein/total PSA ratio could predict prostate cancer (PCa) prior to biopsy. METHODS A total of 316 consecutive patients who had undergone transrectal prostate biopsy and/or transurethral resection were examined. The prostate volume was determined by transrectal ultrasonography (TRUS) and the ability of the above-mentioned four variables to distinguish PCa from benign prostatic hyperplasia (BPH) was evaluated. RESULTS PCa was detected in 61 cases. Receiver-operating characteristic (ROC) analysis revealed that both the PSA density and serum total PSA were the most useful predictors of PCa among the four variables. For the patients with a serum total PSA level of 4.1-10.0 ng/ml, PSA density was significantly more accurate than total PSA (p < 0.005). An optimum PSA density value of 0.18 was chosen as a cutoff because it showed the highest sum of sensitivity and specificity, 92 and 54%, respectively. Using this PSA density cutoff, the number of biopsies could have been reduced to 57 from 63% when compared with a PSA density of 0.15. CONCLUSIONS PSA density was significantly more accurate than other variables in predicting PCa. To avoid unnecessary biopsies, the PSA density cutoff value of 0.18 would be recommendable for determining a prostate biopsy for Japanese males with a serum total PSA level of 4.1-10.0 ng/ml.