Purposes
(a) To externally validate the Crippa and colleagues’ nomograms combining PSA, percentage of positive biopsy cores (PPBC) and biopsy Gleason score to predict organ-confined disease (OCD) in a contemporary sample of patients treated at a tertiary teaching institution. (b) To adjust such variables, resulting in predictive nomograms for OCD and seminal vesicle invasion (SVI): the USP nomograms.
Materials and Methods
The accuracy of Crippa and colleagues’ nomograms for OCD prediction was examined in 1002 men submitted to radical prostatectomy between 2005 and 2010 at the University of São Paulo (USP). ROC-derived area under the curve (AUC) and Brier scores were used to assess the discriminant properties of nomograms for OCD. Nomograms performance was explored graphically with LOESS smoothing plots. Furthermore, univariate analysis and logistic regression models targeted OCD and SVI. Variables consisted of PSA, PPBC, biopsy Gleason score and clinical stage. The resulted predictive nomograms for OCD and SVI were internally validated with bootstrapping and the same abovementioned procedures.
Results
Crippa and colleagues’ nomograms for OCD showed ROC AUC = 0.68 (CI: 0.65-0.70), Brier score = 0.17 and overestimation in LOESS plots. USP nomograms for OCD and SVI showed ROC AUC of 0.73 (CI: 0.70-0.76) and 0.77 (CI: 0.73-0.79), respectively, and Brier scores of 0.16 and 0.08, respectively. The LOESS plots showed excellent calibration for OCD and underestimation for SVI.
Conclusions
Crippa and colleagues’ nomograms showed moderate discrimination and considerable OCD overestimation. USP nomograms showed good discrimination for OCD and SVI, as well as excellent calibration for OCD and SVI underestimation.
Prostate; Prostatic Neoplasms; Risk Factors; Nomograms; Models; Statistical
