We evaluated the diagnostic and prognostic value of PSA density (PSAD) in a prospective study of 100 men consulting for various urinary tract disorders. PSAD was calculated by the ratio of serum PSA over the ultrasonographic volume of the prostate gland, and was expressed in ng/ml/ml. The mean PSAD of 31 patients with prostate cancer was significantly higher than that of the 69 patients without prostate cancer (p = 0.0003). The patients were classified into two groups according to the serum PSA value. Forty nine patients had a serum PSA < or = 10 ng/ml, 31 of whom had a PSAD < or = 0.15 ng/ml/ml and 29 of these 31 patients (94%) did not have prostate cancer; 18 patients had a PSAD > 0.15 ng/ml/ml and six of them (33%) had prostate cancer (p = 0.04). Fifty one patients had a serum PSA > 10 ng/ml and 23 of them (45%) has prostate cancer; 14 of these 51 patients had a PSAD < or = 0.30 ng/ml/ml and two of them (14% à had prostate cancer; 37 patients had a PSAD > 0.30 ng/ml/ml, 21 of whom (57%) had prostate cancer (p = 0.016). Twenty one of these 31 patients with prostate cancer were candidates for radical prostatectomy. Four patients were N+ and 17 underwent radical prostatectomy. Capsular invasion was significantly correlated with PSAD. None of the four patients with PSAD < or = 0.30 ng/ml/ml showed capsular invasion, while 11 of the 13 patients (85%) with PSAD > 0.30 ng/ml/ml had capsular invasion. In conclusion, in patients with PSA < or = 10 ng/ml, calculation of the PSAD avoids approximately two-thirds of prostatic biopsies, with a very low risk of missing a clinically significant cancer. On the other hand, PSAD is of more limited value in patients with PSA > 10 ng/ml. In cancers considered to be operable clinically, the cut-off value of 0.30 ng/ml/ml for PSAD is predictive of capsular or lymph node invasion with a diagnostic accuracy of 94%.