(253) An Emulated Target Trial Case Study of Real-World Overall Survival in Second-line Maintenance Niraparib vs Active Surveillance in Patients with Recurrent Ovarian Cancer

Publication Number: 913

Background: Niraparib, a poly (ADP-ribose) polymerase inhibitor, was assessed as a second-line maintenance (2Lm) therapy for patients (pts) with platinum-sensitive recurrent ovarian cancer (OC) in the NOVA trial; progression-free survival was the primary endpoint and overall survival (OS) was a secondary endpoint.

Objectives: This real-world (RW) study used an emulated target trial method to minimize biases inherent to retrospective observational studies when comparing OS in breast cancer gene wild-type (BRCAwt) pts with recurrent OC receiving 2Lm niraparib monotherapy vs under active surveillance (AS).

Methods: This study used the US nationwide Flatiron Health de-identified electronic health record-derived database. A target trial emulation cloned inverse probability of censoring weighting (IPCW) method was selected a priori to minimize confounding and immortal time bias. Eligible pts, diagnosed with epithelial OC (1 Jan 2011–31 May 2021), known BRCAwt status, and completed 2L therapy (1 Jan 2017–2 Mar 2022), were cloned at index date, assigned to niraparib 2Lm and AS cohorts, and censored when treatment deviated from clone assignment ≤120 days after end of 2L therapy (index date). Follow-up was measured from index date until the earliest of study end (31 May 2022), last activity/death. Pt characteristics were assessed as balanced when the standardized mean difference was < 15%. Median OS (mOS) and hazard ratios (HR) were estimated from stabilized IPCW Kaplan–Meier (KM) curves and Cox regression models; naïve unadjusted, cloned unweighted, and cloned weighted were also estimated.

Results: In total, 199 and 707 pts received niraparib 2Lm or AS, respectively. Key characteristics were balanced across cohorts after cloning and stabilized IPCW (ie, age, race, region, practice type, histology, ECOG status, disease stage, and time from end of first-line treatment to 2L initiation). Median follow-up was 15.6 and 9.3 months pre-cloning, and IPCW mOS was 24.1 (95% CI: 20.9, 29.5) and 18.4 (95% CI: 15.1, 22.8) months for niraparib 2Lm and AS cohorts, respectively (HR: 0.77 [95% CI: 0.66, 0.89]). mOS estimated by naïve, cloned unweighted, and cloned weighted KMs was longer in niraparib 2Lm than AS pts; the difference in mOS decreased with each model (11.5, 6.9, and 6.6 months, respectively) and was shortest (5.7 months) in the final model.

Conclusions: This RW study provides informative data on OS outcomes in BRCAwt pts receiving 2Lm niraparib or AS. This case study demonstrates a practical application of an emulated trial approach which appeared to minimize bias (eg, results closer to the null) to a greater extent than models which did not adjust for immortal time bias.

Funding: GSK 219306. Editorial support provided by Fishawack Health, funded by GSK.