Development of a novel risk model to predict CRPC progression following IMRT: Implications for tailoring treatment intensity

Abstract

Objectives

To develop a novel risk score (RS) model to predict the probability of progression to castration-resistant prostate cancer (PCa) (CRPC) after intensity-modulated radiation therapy (IMRT) for patients with high- and very high-risk PCa according to the National Comprehensive Cancer Network (NCCN) risk classification, since accurate prediction of the clinical outcome of definitive radiation therapy for patients with high- and very high-risk PCa remains challenging due to its heterogeneity.

Materials and Methods

We conducted a retrospective review of 600 patients with high- and very high-risk PCa treated with IMRT at our institution. They were randomly divided into discovery (n = 300) and validation (n = 300) cohorts. A predictive RS model was created using a dataset from the discovery cohort based on the following parameters: T-stage, Gleason score, prostate-specific antigen and age at initiation of IMRT. The model was internally validated using a dataset from the validation cohort. RS was calculated using multivariable Cox regression analysis, and patients were categorized into low-risk, intermediate-risk or high-risk based on the value.

Results

The median follow-up period of the 600 patients was 9.1 (IQR: 6.1–11.6) years. The 10-year CRPC-free rates for low-, intermediate- and high-risk categories were 100.0, 90.4 and 61.4% in the discovery cohort, respectively (p < 0.001). Such differences were reproduced in the validation cohort. Specifically, those rates for low-, intermediate- and high-risk categories were 96.4, 90.7 and 74.8% in the validation cohort, respectively (p < 0.001). Harrell's C-index for this model was 0.692, being higher than that of the NCCN risk classification (0.617).

Conclusion

This RS model provided useful information to enable tailoring of the treatment intensity for this heterogeneous population.