INTRODUCTION: We hypothesized that either very low (0-2.5 ng/mL) or very high (>20 ng/mL) PSA values may limit the accuracy of pathological stage predictions. To test this hypothesis, we examined 5193 consecutive patients subjected to radical prostatectomy (RP) for localized prostate cancer (PCa). MATERIAL AND METHODS: Patients were divided into three cohorts according to their pre-treatment PSA value: 20 ng/mL (n=317). Subsequently in each cohort, the ability of PSA, clinical stage and biopsy Gleason sum was tested in multivariable logistic regression models predicting three separate endpoints: extracapsular extension (ECE), seminal vesicle invasion (SVI) and lymph node invasion (LNI). Predictive accuracy represented the performance benchmark. All models were adjusted for year of surgery and subjected to 200 bootstrap resamples to reduce overfit bias. RESULTS: For PSA 20 ng/mL, predictive accuracy was 63.6%, 63.7% and 70.6%. CONCLUSIONS: The ability to predict pathological stage in patients with PSA values in excess of 20 ng/mL significantly decreased, compared to patients with lower PSA values. Therefore, accurate staging of these patients may require alternative markers or staging schemes.
INTRODUCTION: We hypothesized that either very low (0-2.5 ng/mL) or very high (>20 ng/mL) PSA values may limit the accuracy of pathological stage predictions. To test this hypothesis, we examined 5193 consecutive patients subjected to radical prostatectomy (RP) for localized prostate cancer (PCa). MATERIAL AND METHODS: Patients were divided into three cohorts according to their pre-treatment PSA value: 20 ng/mL (n=317). Subsequently in each cohort, the ability of PSA, clinical stage and biopsy Gleason sum was tested in multivariable logistic regression models predicting three separate endpoints: extracapsular extension (ECE), seminal vesicle invasion (SVI) and lymph node invasion (LNI). Predictive accuracy represented the performance benchmark. All models were adjusted for year of surgery and subjected to 200 bootstrap resamples to reduce overfit bias. RESULTS: For PSA 20 ng/mL, predictive accuracy was 63.6%, 63.7% and 70.6%. CONCLUSIONS: The ability to predict pathological stage in patients with PSA values in excess of 20 ng/mL significantly decreased, compared to patients with lower PSA values. Therefore, accurate staging of these patients may require alternative markers or staging schemes.