To evaluate the usefulness of serum prostate specific antigen in the preoperative staging of prostate cancer we examined tumor volume and differentiation, as well as benign prostatic hyperplasia volume to determine their influence on serum antigen levels. Serum prostate specific antigen was measured in 350 men with clinically localized prostate cancer and preoperatively in 72 men with documented benign prostatic hyperplasia. Although the mean antigen levels increased with advancing pathological stage, the usefulness of prostate specific antigen to predict pathological stage for an individual patient was limited: 1 of 102 men (0.9%) with prostate specific antigen levels of less than 2.8 ng./ml. had positive lymph nodes and 5 of 5 men with levels of greater than 100 ng./ml. had either seminal vesicle or lymph node involvement. However, for the majority of men (greater than 70%) with prostate specific antigen values between these 2 extremes the antigen levels did not accurately predict pathological stage. Because serum prostate specific antigen levels correlated with morphometrically determined tumor volume (r equals 0.535, p less than 0.01) they should, in fact, be predictive of pathological stage. However, most men with prostate cancer also have varying degrees of benign prostatic hyperplasia tissue in the gland producing prostate specific antigen. We have found that serum prostate specific antigen does not correlate with the volume of benign hyperplasia within the gland (r equals 0.21, p greater than 0.05). In addition, immunohistochemical studies have suggested that the lack of correlation between pathological stage and serum prostate specific antigen might be explained by a decrease in the production of antigen with increasing histological grade. Our findings of a negative correlation (r equals -0.37, p less than 0.01) between serum prostate specific antigen levels and Gleason score adjusted for tumor volume confirmed this suggestion. Consequently, serum prostate specific antigen levels do not reflect tumor burden and pathological stage accurately in individual patients for 2 reasons: 1) the unpredictable contribution from the benign prostatic hyperplasia component of the gland and 2) the decreasing production of prostate specific antigen by higher grade lesions as tumor volume increases.