The Friends of Cancer Research Real World Data (RWD) Pilot 2.0 Collaboration: Implementation of Standardized Methodological Approaches for Evaluation of Oncology RWD

ISPE Annual Meeting 2021 Presentation
Authors Donna R Rivera, Henry Henk, Elizabeth Garrett-Mayer, Mark S Walker, Joseph Wagner, Emily Valice, Connor Sweetnam, Elad Sharon, Lori Sakoda, Nicholas Robert, Ruth Pe Benito, Yanina Natanzon, Larry Kushi, Monika Izano, Rebecca Honnold, Eric Hansen, Janet Espirito, Lindsey Enewold, Jennifer Christian, Suanna S Bruinooge, Marley Boyd, Andrew Belli, Laura Lasiter, and Jeff Allen

Background: Real-world evidence (RWE) can reflect more broad and diverse patient populations than traditional clinical trials but challenges remain in using real-world data (RWD) to support clinical research, drug development, and regulatory decision-making. Operationalizing standardized variable definitions and implementing consistent statistical methods are essential.

Objectives: We determined the feasibility of developing and implementing a RWE master protocol across 8 oncology RWD partners, with data comprised of administrative claims, electronic health records, and/or cancer registry data, to evaluate treatment effects between platinum doublet chemotherapy (PC) and PD-(L)1 monotherapy (PDL1) in frontline treatment.

Methods: Patients meeting broad eligibility criteria, treated with a qualifying first-line therapy for advanced non-small cell lung cancer (aNSCLC), were examined to reflect the real-world populations. Three endpoints were assessed based on clinical relevance and operational practicality: real-world overall survival (rwOS), time to treatment discontinuation (rwTTD) of frontline regimen, and time to next treatment (rwTTNT). RWD partners conducted parallel protocol-based analyses accounting for key sources of variation in data availability due to population included, length of follow-up, and year of treatment initiation. Endpoints for each regimen were described by Kaplan-Meier estimation for each dataset. Hazard ratios (HRs) were estimated using Cox regression adjusting for prognostic factors available in all datasets. HRs are not included for one group which only had PC data.

Results: Median rwOS in the PC arm ranged from 10-17 months across groups vs 12-18 months in the PD-L1 arm. Adjusted HRs for rwOS (PDL1 vs PC) ranged from 0.88-1.22. The 95% CIs all included 1.0, but the direction of association varied: three groups had HRs >1.0 (1.06, 1.09, 1.22), two <1.0 (0.88, 0.88) and two at 1.0 (0.99, 0.99). Larger differences were observed in HRs between datasets for rwTTD (range 0.40-0.52) than rwTTNT (range 0.51-0.80) but all HRs were significantly <1 indicating that NSCLC patients had longer time on first-line treatment with PDL1 than PC.

Conclusions: This study demonstrated the feasibility of developing and implementing a common master protocol across RWD sources is feasible. Studies designed to evaluate uses of RWD require careful and informed cohort definition and variable selection. Understanding of RWE in the larger context of clinical practice is critical to this work. These efforts will be pivotal to help advance the use of RWE in regulatory decision making.