Probability of prostate cancer detection based on results of a multicenterstudy using the AxSYM free PSA and total PSA assays

Objectives. The determination of the percentage of free prostate-specific a ntigen (%fPSA) enhances the specificity of prostate cancer (CaP) detection. This study was undertaken to assess the performance of %fPSA in differenti ating benign prostate disease from CaP and to determine the CaP probability estimates using the AxSYM Free PSA and AxSYM Total PSA assays. Methods. In this prospective study, 297 men, 50 years old or older, with a total PSA level between 4 and 10 ng/mL and a nonsuspicious digital rectal e xamination were enrolled at 10 clinical sites. All subjects underwent at le ast sextant prostate biopsies to establish the diagnosis. fPSA and total PS A (tPSA) levels were determined using the AxSYM Free PSA and AxSVM Total PS A assays. Percent fPSA values were compared with tPSA values to determine t he appropriate cutoffs for prostate biopsy and to calculate the CaP probabi lity estimates. Results. The strongest predictor of CaP in a logistic regression model was %fPSA (odds ratio 2.29), which contributed significantly more than age or t PSA to the predictive model. In this study population, a %fPSA cutoff of 26 .4% would have detected 96% of subjects with CaP (sensitivity) and would ha ve eliminated 27.4% of unnecessary biopsies (specificity). CaP probability estimates ranged from 9% to 69% and increased as the %fPSA value decreased. Men with a %fPSA level of 10% or lower had a 69% probability of CaP, and m en with a %fPSA level of greater than 26% had a 9% probability of CaP. Conclusions. Percent fPSA values can help differentiate CaP from benign pro state disease and reduce unnecessary biopsies in 27% of men 50 years old or older whose digital rectal examination was normal and whose tPSA level was between 4 and 10 ng/mL. A %fPSA result can assist the physician and patien t in determining the probability of CaP and assessing the need for prostate biopsy. UROLOGY 55: 909-914, 2000. (C) 2000, Elsevier Science Inc.