MIRROR KEYNOTE: Generating Real-World External Control Arms in Non-Small Cell Lung Cancer Using a Curated Database
Author(s)
Sreeram Ramagopalan, PhD1, Paul Arora, PhD2, Emma Mackay, MA, MSc2, Carrie Manthei, MSc3.
1Kings College London, London, United Kingdom, 2Inka Health, Toronto, ON, Canada, 3OneMedNet, Eden Prairie, MN, USA.
1Kings College London, London, United Kingdom, 2Inka Health, Toronto, ON, Canada, 3OneMedNet, Eden Prairie, MN, USA.
OBJECTIVES: External control arms (ECA) using real-world data (RWD) are increasingly recognized as a valuable approach to accelerate clinical development for rare disease. However, regulatory and HTA bodies remain cautious due to concerns about RWD quality. This study outlines an approach following target trial emulation principles to construct an ECA using a curated clinical RWD database together with the application of matching-adjusted indirect comparison (MAIC) when individual patient data (IPD) is lacking for the comparator arm.
METHODS: We outline a protocol for a planned ECA analysis with the aim of replicating the results of the KEYNOTE-189 trial. From OneMedNet's database, an ECA will be made by identifying patients with metastatic non-squamous NSCLC without EGFR/ALK mutations who received first-line platinum-based chemotherapy (cisplatin or carboplatin plus pemetrexed) with ECOG performance status 0-1. Key baseline characteristics will include age, sex, smoking status, ECOG, brain metastases status, and PD-L1 expression levels. Pseudo-IPD for both KEYNOTE-189 arms will be reconstructed using the Guyot algorithm from published survival curves. MAICs will be used to compare ECA with KEYNOTE arms, with the primary endpoint being overall survival (OS).
RESULTS: Baseline characteristics will be compared between the ECA and KEYNOTE-189 arms. Kaplan-Meier curves for OS will be plotted for the ECA and KEYNOTE-189 arms, along with hazard ratios and 95% confidence intervals (CI). Differences between the ECA and KEYNOTE-189 control arm will be assessed using restricted mean survival time (RMST) differences with accompanying 95% CI.
CONCLUSIONS: This study will demonstrate the implementation of an ECA constructed from RWD via MAIC. The goal will be to inform best practices for developing and validating ECA analyses for future clinical development programs, and to illustrate the applicability of methods like MAIC for the construction of ECAs with limited IPD.
METHODS: We outline a protocol for a planned ECA analysis with the aim of replicating the results of the KEYNOTE-189 trial. From OneMedNet's database, an ECA will be made by identifying patients with metastatic non-squamous NSCLC without EGFR/ALK mutations who received first-line platinum-based chemotherapy (cisplatin or carboplatin plus pemetrexed) with ECOG performance status 0-1. Key baseline characteristics will include age, sex, smoking status, ECOG, brain metastases status, and PD-L1 expression levels. Pseudo-IPD for both KEYNOTE-189 arms will be reconstructed using the Guyot algorithm from published survival curves. MAICs will be used to compare ECA with KEYNOTE arms, with the primary endpoint being overall survival (OS).
RESULTS: Baseline characteristics will be compared between the ECA and KEYNOTE-189 arms. Kaplan-Meier curves for OS will be plotted for the ECA and KEYNOTE-189 arms, along with hazard ratios and 95% confidence intervals (CI). Differences between the ECA and KEYNOTE-189 control arm will be assessed using restricted mean survival time (RMST) differences with accompanying 95% CI.
CONCLUSIONS: This study will demonstrate the implementation of an ECA constructed from RWD via MAIC. The goal will be to inform best practices for developing and validating ECA analyses for future clinical development programs, and to illustrate the applicability of methods like MAIC for the construction of ECAs with limited IPD.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
RWD126
Topic
Health Technology Assessment, Methodological & Statistical Research, Real World Data & Information Systems
Topic Subcategory
Health & Insurance Records Systems
Disease
Oncology