Development and validation of a microRNA-based signature (MiROvaR) to predict early relapse or progression of epithelial ovarian cancer: a cohort study

Summary Background Risk of relapse or progression remains high in the treatment of most patients with epithelial ovarian cancer, and development of a molecular predictor could be a valuable tool for stratification of patients by risk. We aimed to develop a microRNA (miRNA)-based molecular classifier that can predict risk of progression or relapse in patients with epithelial ovarian cancer. Methods We analysed miRNA expression profiles in three cohorts of samples collected at diagnosis. We used 179 samples from a Multicenter Italian Trial in Ovarian cancer trial (cohort OC179) to develop the model and 263 samples from two cancer centres (cohort OC263) and 452 samples from The Cancer Genome Atlas epithelial ovarian cancer series (cohort OC452) to validate the model. The primary clinical endpoint was progression-free survival, and we adapted a semi-supervised prediction method to the miRNA expression profile of OC179 to identify miRNAs that predict risk of progression. We assessed the independent prognostic role of the model using multivariable analysis with a Cox regression model. Findings We identified 35 miRNAs that predicted risk of progression or relapse and used them to create a prognostic model, the 35-miRNA-based predictor of Risk of Ovarian Cancer Relapse or progression (MiROvaR). MiROvaR was able to classify patients in OC179 into a high-risk group (89 patients; median progression-free survival 18 months [95% CI 15–22]) and a low-risk group (90 patients; median progression-free survival 38 months [24–not estimable]; hazard ratio [HR] 1·85 [1·29–2·64], p=0·00082). MiROvaR was a significant predictor of progression in the two validation sets (OC263 HR 3·16, 95% CI 2·33–4·29, p