This investigation examined the ability of four distinct prognostic risk-stratification systems to predict for BFFS.
Univariable Cox proportional hazards regression was performed for BFFS for each risk-stratification variable, separately by database.
The concordance index (C-index) was calculated and reported from Cox proportional hazards regression for BFFS using 10-fold cross-validation for all possible risk-stratification variables, as described previously (CAPRA score, CAPRA-5, CAPRA-3, ProCaRS, GUROC, and NCCN).
External validation of the previously published ProCaRS nomograms to predict five-year BFFS (as defined previously) for LDR-brachytherapy only and EBRT only was performed using the CHUM database.
Biochemical failure was observed in 134 patients (6%) and death in 102 patients (5%) with a corresponding five-year BFFS of 92.
Additionally, each of the examined risk-stratification systems was found to be an overall significant predictor of BFFS for both databases (all p<0.
The calibration plots of the ProCaRS nomograms predictive of five-year BFFS for LDR-brachytherapy and EBRT are shown in Fig.
We found that the CAPRA score and ProCaRS were superior to the GUROC and NCCN risk-assessment models for predicting BFFS in both cohorts.
Although our results demonstrate that clinical scoring systems perform reasonably well in predicting BFFS, further improvement in risk-stratification is possible.
This study also validated the ProCaRS five-year BFFS nomograms for LDR-brachytherapy and EBRT in an independent cohort.