When to order genomic tests: development and external validation of a model to predict high-risk prostate cancer at the genotypic level.

Abstract

The aim of this study was to develop a model to predict high-genomic-risk prostate cancer (PCa) according to Decipher score, a validated 22 gene prognostic panel. By doing so, one might select the individuals who are likely to benefit from genomic testing and improve pre-op counseling about the need for adjuvant treatments. We retrospectively reviewed IRB-approved databases at two institutions. All patients had preoperative magnetic resonance imaging (MRI) and Decipher prostate radical prostatectomy (RP), a validated 22 gene prognostic panel. We used binary logistic regression to estimate high-risk Decipher (Decipher score > 0.60) probability on RP specimen. Area under the curve (AUC) and calibration were used to assess the accuracy of the model in the development and validation cohort. Decision curve analysis (DCA) was performed to assess the clinical benefit of the model. The development and validation cohort included 622 and 185 patients with 283 (35%) and 80 (43%) of those with high-risk Decipher. The multivariable model included PSA density, biopsy Gleason Grade Group, percentage of positive cores and MRI extracapsular extension. AUC was 0.73 after leave-one-out cross-validation. DCA showed a clinical benefit in a range of probabilities between 15 and 60%. In the external validation cohort, AUC was 0.70 and calibration showed that the model underestimates the actual probability of the outcome. The proposed model to predict high-risk Decipher score at RP is helpful to improve risk stratification of patients with PCa and to assess the need for additional testing and treatments.