The impact of genomic biomarkers on a clinical risk prediction model for upgrading/upstaging among men with favorable-risk prostate cancer

Abstract

Background: The challenge of distinguishing indolent from aggressive prostate cancer (PCa) complicates decision-making for men considering active surveillance (AS). Genomic classifiers (GCs) may improve risk stratification by predicting end points such as upgrading or upstaging (UG/US). The aim of this study was to assess the impact of GCs on UG/US risk prediction in a clinicopathologic model.

Methods: Participants had favorable-risk PCa (cT1-2, prostate-specific antigen [PSA] ≤15 ng/mL, and Gleason grade group 1 [GG1]/low-volume GG2). A prediction model was developed for 864 men at the University of California, San Francisco, with standard clinical variables (cohort 1), and the model was validated for 2267 participants from the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) registry (cohort 2). Logistic regression was used to compute the area under the receiver operating characteristic curve (AUC) to develop a prediction model for UG/US at prostatectomy. A GC (Oncotype Dx Genomic Prostate Score [GPS] or Prolaris) was then assessed to improve risk prediction.

Results: The prediction model included biopsy GG1 versus GG2 (odds ratio [OR], 5.83; 95% confidence interval [CI], 3.73-9.10); PSA (OR, 1.10; 95% CI, 1.01-1.20; per 1 ng/mL), percent positive cores (OR, 1.01; 95% CI, 1.01-1.02; per 1%), prostate volume (OR, 0.98; 95% CI, 0.97-0.99; per mL), and age (OR, 1.05; 95% CI, 1.02-1.07; per year), with AUC 0.70 (cohort 1) and AUC 0.69 (cohort 2). GPS was associated with UG/US (OR, 1.03; 95% CI, 1.01-1.06; p < .01) and AUC 0.72, which indicates a comparable performance to the prediction model.

Conclusions: GCs did not substantially improve a clinical prediction model for UG/US, a short-term and imperfect surrogate for clinically relevant disease outcomes.

Keywords: active surveillance; genomic classifiers; prediction; prostate cancer.