OBJECTIVES: The contemporary ability of prostate-specific antigen (PSA) to predict pathologic stage in men with localized prostate cancer was recently questioned. METHODS: We quantified the added value related to the addition of pretreatment PSA to established pathologic stage predictors (namely clinical stage and biopsy Gleason sum) in 5921 consecutive radical prostatectomy (RP) patients. Univariable and multivariable logistic regression analyses predicting pathologic stage (extracapsular extension [ECE], seminal vesicle invasion [SVI], lymph node invasion [LNI], and organ-confined disease [OC]) were stratified according to four time quartiles. The gain in predictive accuracy (PA) related to the inclusion of PSA to multivariable models was quantified by using the area under the curve method. RESULTS: Temporal analyses showed a decrease in PSA levels over the study years (p<0.001). Conversely, the rate of nonpalpable disease and the rate of biopsy Gleason sum < or =6 increased over time (all p<0.001). At RP, the rate of OC increased over time, while the rate of ECE and SVI decreased over time (all p<0.001). The rate of LNI remained stable (p=0.1). In multivariable models, PSA represented an independent predictor of all pathologic stages over time (all p<0.03), except for ECE in the first and last time quartiles. The addition of PSA significantly increased the multivariable PA of all models predicting pathologic stages over time (all p<0.03), except for ECE predictions in the first quartile (p=0.1). CONCLUSIONS: In the new millennium, PSA has not lost its ability to accurately predict the pathologic stage in contemporary patients.
OBJECTIVES: The contemporary ability of prostate-specific antigen (PSA) to predict pathologic stage in men with localized prostate cancer was recently questioned. METHODS: We quantified the added value related to the addition of pretreatment PSA to established pathologic stage predictors (namely clinical stage and biopsy Gleason sum) in 5921 consecutive radical prostatectomy (RP) patients. Univariable and multivariable logistic regression analyses predicting pathologic stage (extracapsular extension [ECE], seminal vesicle invasion [SVI], lymph node invasion [LNI], and organ-confined disease [OC]) were stratified according to four time quartiles. The gain in predictive accuracy (PA) related to the inclusion of PSA to multivariable models was quantified by using the area under the curve method. RESULTS: Temporal analyses showed a decrease in PSA levels over the study years (p<0.001). Conversely, the rate of nonpalpable disease and the rate of biopsy Gleason sum < or =6 increased over time (all p<0.001). At RP, the rate of OC increased over time, while the rate of ECE and SVI decreased over time (all p<0.001). The rate of LNI remained stable (p=0.1). In multivariable models, PSA represented an independent predictor of all pathologic stages over time (all p<0.03), except for ECE in the first and last time quartiles. The addition of PSA significantly increased the multivariable PA of all models predicting pathologic stages over time (all p<0.03), except for ECE predictions in the first quartile (p=0.1). CONCLUSIONS: In the new millennium, PSA has not lost its ability to accurately predict the pathologic stage in contemporary patients.