The introduction of PARP inhibitors (PARPi) has greatly changed Ovarian Cancer (OC) patients' journey. Nonetheless, subgroups of patients still benefit of Bevacizumab (Bev) particularly those at clinical high risk or proficient for BRCA and/or Homologous Recombination. We developed, and validated, a robust and independent miRNA-based molecular predictor, MiROvaR, successfully identifying patients at high risk of early relapse. MiROvaR may contribute in refining poor-prognosis patients predicting those who could greatly benefit from Bev treatment. Samples used for the analyses were from two clinical trials: MITO16A-MANGO-Ov2 (MITO16A), single arm including Bev treatment/maintenance in front-line and MITO16B-MANGO-Ov2-ENGOT-Ov17 (MITO16B), randomized phase III including or not Bev treatment/maintenance in platinum sensitive patients relapsing after receiving Bev in front-line. RNA of adequate quality from patient enrolled in MITO16A (n=197) and MITO16B (n=102 standard arm; n=108 experimental arm) was profiled for miRNA expression (Agilent 8x60K miRBase21 version). MiROvaR-Index was derived to classify patients for being at high/low risk of relapse and assess association with clinical/pathological parameters and prognostic/predictive impact. The biomarker-evaluable populations comprising 49.5% and 51.7% of the intent-to-treat populations of MITO16A and MITO16B respectively had representative baseline characteristics and outcomes. In MITO16A, MiROvaR confirmed its performance in progression-free survival (PFS) and maintained an independent prognostic power in multivariable analysis with residual disease and FIGO stage (HR 1.74, 95% CI 1.131–2.67; P=0.011). In MITO16B, high MiROvaR-Index was predictive of a therapeutic advantage with Bev for PFS (Pinteraction = 0.00754) but not for Overall Survival. Patients with high MiROvaR-Index treated with Bev had longer PFS (13 vs. 8 months; log-rank P<0.0001) compared to those in the control arm. High MiROvaR-Index confirmed its prognostic power of early relapse independently of the treatment schedule and suggested a predictive potential of Bev response.