BACKGROUND With serum prostate specific antigen (PSA) levels of 4.1 to 10 ng/ml a significant number of patients are unnecessary subjected to biopsy. We try to determine if the calculation of prostate specific antigen density (PSAD) and prostate specific antigen density of the transition zone (PSADTZ) improve the capacity to discriminate between non-malignant disease and prostate cancer. METHODS A prospective study including 314 males with PSA levels between 4.1 and 10 ng/ml is reported. Transrectal ultrasonography and prostatic biopsy were performed in all of them and total prostate and transition-zone volumes were calculated. PSA density and PSA density of the transition zone were calculated for each patient. Receiver operating characteristics (ROC) curves for PSA, PSAD and PSADTZ were constructed for all the patients and for those patients with digital rectal examination unsuspicous of malignancy, determining the sensitivity and specificity for several cutoff values. RESULTS The area under the curve for both, PSAD and PSADTZ, were greater than for PSA (p < 0.05), without any significant differences between PSADTZ and PSAD. The cutoff value of greatest diagnostic efficiency for PSAD was 0.17 ng/ml/cc (71.4% sensitivity and 55.7% specificity), while it was 0.41 ng/ml/cc for DPSATZ (70% sensitivity and 61.5% specificity). For those cases of normal digital rectal examination, no differences were observed between PSA and PSAD but they were between PSA and PSADTZ. In any event, the area under ROC curves was always less than 0.7, and, in order to avoid a large number of biopsies (high specificity), a large number of cancers are left without diagnosis (low sensitivity). CONCLUSIONS We conclude that PSAD and PSADTZ are not excessively useful for adequately discriminating between patients with prostate cancer and those with non-malignant disease, particularly when digital rectal examination is normal.