There is a high unmet medical need for disease control in platinum resistant epithelial ovarian cancer patients (PROC). PemBOv trial (NCT03596281) showed promising results with the combination of pembrolizumab plus short-term flat dose bevacizumab with or without pegylated liposomal doxorubicin (PLD) based chemotherapy. However, overall, 20 % of pts showed trough levels for bevacizumab below the targeted threshold of plasma levels associated with efficacy. Based on the PemBOv data, the current analysis aimed to develop a pharmacometric model to characterize the tumor growth early kinetics and quantify the effect of treatment in the study cohorts. A focus was made on the association between bevacizumab exposure and tumor response. Methods: Tumor growth kinetic was studied to predict the clinical outcome. Population pharmacodynamic non-linear mixed-effects modeling approach was used to characterize tumor size evolution in patients. The Kkill parameter expressed as week^-1 was defined to best describe the shrinkage in tumor size depending on the treatments. In addition, Exposure-Response relationships was performed to evaluate whether bevacizumab plasma levels were associated with the previously determined Kkill parameter. The exposure to bevacizumab was defined by the trough level at the second treatment cycle (PreC2, µg/ml). Results: A strong difference in Kkill values was found between cohort A (pembrolizumab plus PLD, n=6, Kkill = 0.00091 week^-1) and either cohorts B (pembrolizumab + bevacizumab, n=19, Kkill = 0.002) and C (pembrolizumab + bevacizumab + PLD, n=19, Kkill = 0.02 week^-1) respectively, whereas little difference was observed between cohorts B and C. This difference indicates higher tumor shrinkage in cohorts B and C as compared to cohort A. No clear Exposure-Response relationship was evidenced between bevacizumab plasma levels and tumor shrinkage. Still, in patients treated with bevacizumab plus pembrolizumab (cohorts B and C), mean bevacizumab exposure was higher in patients with an ORR and Stable disease of at least 4 months as compared with patients with no clinical benefit (i.e., 69.4 VS. 48 ug/ml). Association between bevacizumab levels and clinical outcome deserves further investigations. Because Kkill can be estimated in an early fashion through in silico modeling, this could be evaluated as a surrogate marker to forecast clinical outcome. Conclusions: Bevacizumab exposure correlated with ORR and clinical benefit in PROC patients treated with bevacizumab plus pembrolizumab plus minus PLD, since higher plasma trough levels were associated with better clinical outcome. The early Kkill in silico modeling could pave the way for the development of prediction models for personalized treatments of PROC patients. Clinical trial information: NCT03596281 .