Introduction

Currently utilized nomograms for prostate cancer were developed using populations primarily composed of men with low- and intermediate-risk disease. However, there is a need to elucidate accurate, patient-specific point estimates for oncologic outcomes in men with high-risk (HR) and very high-risk (VHR) disease to enable patients and providers to make superior clinical decisions. We therefore sought to construct a novel nomogram that predicts biochemical recurrence (BCR) from a contemporary cohort of men with HR and VHR prostate cancer.

Methods

A total of 1,241 men with HR or VHR prostate cancer who underwent radical prostatectomy from 2005 to 2015 were identified from Johns Hopkins (n = 620, training cohort) and the Cleveland Clinic (n = 621, validation cohort). The primary endpoint was BCR after radical prostatectomy. Cox multivariable regression was performed to model characteristics and outcomes in the training cohort. The AUC of the model in the training cohort was adjusted for optimism by subjecting the model to bootstrapping with 100 resamples. Model accuracy was assessed using the time-dependent area under the receiver operator characteristic curve (AUC) in the validation cohort.

Results

A total of 494 men developed BCR, with 245 arising from the training cohort and 249 from the validation cohort. The overall BCR-free probability was 49.0% (95% CI: 45.4%-52.9%) at 5 years. The nomogram for postoperative BCR probability was developed using age, race, PSA, Gleason grade group, clinical stage, and number of cores with Gleason score 8-10 disease [Figure 1]. Model AUC was 0.730 after optimism-adjustment, as compared to 0.700 and 0.654 in the existing Stephenson and Cancer of the Prostate Risk Assessment (CAPRA) nomograms, respectively [Figure 2]. The nomogram demonstrated similar accuracy in the external validation cohort (AUC = 0.734).

Conclusions

Accurate and individualized risk assessment for the outcome of BCR is imperative for optimizing clinical decisions and designing clinical trials. We have herein described a novel predictive tool created exclusively from men with HR and VHR prostate cancer demonstrating better discriminative ability than existing nomograms for the prediction of postoperative BCR in this important patient population.

Funding

None