Abstract
BACKGROUND: A prostate cancer diagnosis is based on biopsy sampling that is an invasive, expensive procedure, and doesn't accurately represent multifocal disease. METHODS: To establish a model using plasma miRs to distinguish Prostate cancer patients from non-cancer controls, we enrolled 600 patients histologically diagnosed as having or not prostate cancer at biopsy. Two hundred ninety patients were eligible for the analysis. Samples were randomly divided into discovery and validation cohorts. RESULTS: NGS-miR-expression profiling revealed a miRs signature able to distinguish prostate cancer from non-cancer plasma samples. Of 51 miRs selected in the discovery cohort, we successfully validated 5 miRs (4732-3p, 98-5p, let-7a-5p, 26b-5p, and 21-5p) deregulated in prostate cancer samples compared to controls (p ≤ 0.05). Multivariate and ROC analyses show miR-26b-5p as a strong predictor of PCa, with an AUC of 0.89 (CI = 0.83-0.95;p < 0.001). Combining miRs 26b-5p and 98-5p, we developed a model that has the best predictive power in discriminating prostate cancer from non-cancer (AUC = 0.94; CI: 0,835-0,954). To distinguish between low and high-grade prostate cancer, we found that miR-4732-3p levels were significantly higher; instead, miR-26b-5p and miR-98-5p levels were lower in low-grade compared to the high-grade group (p ≤ 0.05). Combining miR-26b-5p and miR-4732-3p we have the highest diagnostic accuracy for high-grade prostate cancer patients, (AUC = 0.80; CI 0,69-0,873). CONCLUSIONS: Noninvasive diagnostic tests may reduce the number of unnecessary prostate biopsies. The 2-miRs-diagnostic model (miR-26b-5p and miR-98-5p) and the 2-miRs-grade model (miR-26b-5p and miR-4732-3p) are promising minimally invasive tools in prostate cancer clinical management.