Abstract VAN is a bispecific IgG-like antibody targeting both VEGF-A and Ang-2, two key factors in tumor angiogenesis pathway that was investigated as monotherapy in a single arm study in platinum-resistant ovarian cancer (PROC) patients. A modeling approach working with tumor size data was applied to predict and compare overall survival (OS) of VAN + chemotherapy (CT) against BEV + CT. For VAN, individual tumor size data from 41 patients enrolled in study NCT01688206 were available. Primary endpoint of the study was objective response rate; survival data were not collected. For CT and BEV +CT, historical tumor size data were obtained from AURELIA, a randomized phase 3 study designed to compare progression free survival (PFS) in patients treated with CT alone or in combination with BEV. In AURELIA, patients were followed for OS after treatment discontinuation. In both studies, sum of longest diameters (SLD) were measured on CT/MRI scans collected every 6 to 8 weeks, as per RECIST 1.1. Non-linear tumor kinetics (TK) models accounting for the dynamics of tumor growth, drug effect and resistance to the drug effect were used to fit SLD following each treatment (VAN, CT and BEV+CT). Several tumor metrics summarizing the individual TK were derived: early tumor shrinkage at week 6 (ETS6), tumor size at baseline (TS0), maximum shrinkage (MaxSh) and time to tumor growth (TTG). As no VAN+CT data were available, individual tumor responses were simulated assuming an additive drug effect from CT on VAN. The median TTG was prolonged by 4 weeks and median MaxSh was increased by 7% for V+CT as compared to BEV+CT. To correlate tumor metrics to OS, a time-to-event model was fitted to the OS data from the AURELIA study. Covariates including ECOG, FIGO score at baseline, histological grade and subtype, presence of ascites, CA-125 at baseline, TS0, ETS6 and TTG were tested as prognostic factors in the survival model. In the survival model, two sets of covariates were found significant: those related to disease severity (ECOG, FIGO, presence of ascites) and those describing the key features of TK (ETS6, TTG, TS0). Treatment group was not retained in the final model making the model drug independent. Assuming the same survival model and a similar mechanism for VAN and BEV, the BEV+CT TK metrics were replaced by the VAN+CT TK metrics in the survival model. Based on this assumption, the median OS with VAN+CT was predicted to be 2 months longer compared to BEV+CT. This approach shows the benefit of using TK modeling, when drugs show similar mechanism of action, in early phase to predict potential outcome of drug combination, not yet tested in patients. Leveraging historical data for the development of survival model integrating TK metrics can be used to inform the expected OS outcome in phase 3 and can provide a quantitative tool to evaluate the chance of drug success. Citation Format: Alexandre Sostelly, Kevin Smart, Felix Jaminion, Christophe Boetsch, Francois Mercier. Leveraging tumor size and time to death from bevacizumab (BEV) historical data to predict overall survival in ovarian cancer patients treated with vanucizumab (VAN) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1643.