Assessment of Serum microRNA Biomarkers to Predict Reclassification of Prostate Cancer in Patients on Active Surveillance

Abstract

Conventional clinical variables cannot accurately differentiate indolent from aggressive prostate cancer in patients on active surveillance. We investigated promising circulating miRNA biomarkers to predict the reclassification of active surveillance cases. We collected serum samples from 2 independent active surveillance cohorts of 196 and 133 patients for the training and validation, respectively, of candidate miRNAs. All patients were treatment naïve and diagnosed with Gleason score 6 prostate cancer. Samples were collected prior to potential reclassification. We analyzed 9 circulating miRNAs previously shown to be associated with prostate cancer progression. Logistic regression and ROC analyses were performed to assess the predictive ability of miRNAs and clinical variables. A 3-miR (miRNA-223, miRNA-24 and miRNA-375) score was significant to predict patient reclassification (training OR 2.72, 95% CI 1.50-4.94 and validation OR 3.70, 95% CI 1.29-10.6). It was independent of clinical characteristics in multivariable models. The ROC AUC was maximized when combining the 3-miR score and prostate specific antigen, indicating additive predictive value. The 3-miR score plus the prostate specific antigen panel cutoff achieved 89% to 90% negative predictive value and 66% to 81% specificity. The 3-miR score combined with prostate specific antigen represents a noninvasive biomarker panel with high negative predictive value. It may be used to identify patients on active surveillance who have truly indolent prostate cancer.