Time-Dependent Changes in Risk of Progression During Use of Bevacizumab for Ovarian Cancer

[Importance] Although bevacizumab has been used in the treatment of ovarian cancer, its optimal use is unknown. [Objective] To investigate time-dependent changes in the outcomes of bevacizumab therapy. [Design, Setting, and Participants] This cohort study was conducted using published data from 7 previous randomized phase 3 clinical trials with bevacizumab (ICON7, GOG-0218, BOOST, GOG-0213, OCEANS, AURERIA, and MITO16B) from January 10 to January 31, 2023. From 2 ancillary analyses of the ICON7 trial with individual patient data and tumor gene expression profiles, an ICON7-A cohort was generated comprising 745 cases. From other studies, published Kaplan-Meier curves were graphically analyzed. [Exposures] Bevacizumab treatment vs placebo or no treatment. [Main Outcomes and Measures] Restricted mean survival time and relative risk of progression at a given time point between bevacizumab treatment and control groups. [Results] In the ICON7-A cohort (n = 745), restricted mean survival analysis showed that bevacizumab treatment (n = 384) had significantly better progression-free survival (PFS) than the control (n = 361) before bevacizumab discontinuation (restricted mean survival time ratio, 1.08; 95% CI, 1.05-1.11; P < .001), but had significantly worse PFS after bevacizumab discontinuation (0.79; 95% CI, 0.69-0.90; P < .001), showing rebound. In a post hoc analysis, the rebound was similarly observed both in homologous recombination deficiency (HRD) (before, 1.05; 95% CI, 1.02-1.09; P < .001; after, 0.79; 95% CI, 0.63-0.98; P = .04) and non-HRD tumors (before, 1.08; 95% CI, 1.03-1.15; P < .001; after, 0.71; 95% CI, 0.56-0.90; P < .001) of the serous subtype, but not in the nonserous subtype (before, 1.11; 95% CI, 1.05-1.18; P < .001; after, 0.94; 95% CI, 0.78-1.15; P = .57). In Kaplan-Meier curve image–based analysis, the trend of rebound effect was consistently observed in the overall ICON7 and GOG-0218 cohorts and their subgroups stratified by prognostic factors, homologous recombination–associated mutations, and chemotherapy sensitivity. In contrast, no such trend was observed in the studies GOG-0213, OCEANS, AURERIA, and MITO16B, in which patients who experienced relapse received bevacizumab until progression. [Conclusions and Relevance] In ovarian cancer, bevacizumab may reduce progression for approximately 1 year after initiation, but discontinuation may increase subsequent progression in the serous subtype regardless of HRD status. The results suggest that in the first-line treatment, bevacizumab may be more beneficial in patients with a shorter prognosis who are less likely to experience the rebound outcome.