EXTERNAL CONTROL ARM FEASIBILITY ACROSS EXTERNAL DATA SOURCES IN ONCOLOGY: METHODOLOGICAL AND REGULATORY SCIENCE CONSIDERATIONS
Author(s)
Bernat Navarro, Ph.D.1, Kawther Abdilleh, Ph.D.2, Amy Alabaster, MPH3, Peter Ansell, Ph.D.4, Li Chen, Ph.D.5, Gregory S. Calip, MPH, PharmD, Ph.D.4, Ruthanna Davi, Ph.D.6, Janet Espirito, PharmD7, Laura L. Fernandes, Ph.D.8, Sebastian Zavala Hoffmann, MS9, Patricia Luhn, Ph.D.10, Xinran Ma, MS11, Patricia Prince, MPH12, Mark Riffon, MPH13, Xiang Yin, Ph.D.6, Mark Stewart, Ph.D.1, Hillary Andrews, Ph.D.1, Jeff Allen, PhD1;
1Friends of Cancer Research, Washington, DC, USA, 2Pancreatic Cancer Action Network, El Segundo, CA, USA, 3ConcertAI, Cambridge, MA, USA, 4AbbVie, Chicago, IL, USA, 5Amgen, Thousand Oaks, CA, USA, 6Medidata, Belmont, MA, USA, 7Ontada, Boston, MA, USA, 8COTA Healthcare, New York, NY, USA, 9iOMEDICO, Freiburg im Breisgau, Germany, 10Genentech, San Francisco, CA, USA, 11Flatiron Health, Sammamish, WA, USA, 12Aetion/Datavant, New York, NY, USA, 13American Society of Clinical Oncology (ASCO), Alexandria, VA, USA
1Friends of Cancer Research, Washington, DC, USA, 2Pancreatic Cancer Action Network, El Segundo, CA, USA, 3ConcertAI, Cambridge, MA, USA, 4AbbVie, Chicago, IL, USA, 5Amgen, Thousand Oaks, CA, USA, 6Medidata, Belmont, MA, USA, 7Ontada, Boston, MA, USA, 8COTA Healthcare, New York, NY, USA, 9iOMEDICO, Freiburg im Breisgau, Germany, 10Genentech, San Francisco, CA, USA, 11Flatiron Health, Sammamish, WA, USA, 12Aetion/Datavant, New York, NY, USA, 13American Society of Clinical Oncology (ASCO), Alexandria, VA, USA
OBJECTIVES: External control arms (ECAs) can generate scientifically rigorous comparison data when randomized controls are infeasible, but methodologies for developing ECAs using real-world data (RWD) and historical clinical trials (HCTs) are not universally accepted. As a proof-of-concept, Friends’ pilot evaluated ECAs from multiple external data sources mirroring the control arm of the metastatic pancreatic RESOLVE HCT. Objectives included: (1) operationalizing target-trial eligibility criteria across heterogeneous RWD sources; (2) characterizing data missingness and quality; and (3) comparing outcomes against the HCT control arm.
METHODS: We developed a statistical analysis plan (SAP) to translate target-trial eligibility criteria, propensity score modeling, and variable definitions into standardized operational rules. Nine groups applied the SAP to independent RWD or HCT datasets, documenting operational choices and attrition. We examined missingness to understand variable capture and identify gaps affecting feasibility and covariate balance. Propensity score matching and baseline comparisons were conducted to evaluate balance between external and target control arms.
RESULTS: Three eligibility criteria—adequate hepatic/renal function, adequate hematologic function, and no prior systemic therapy—accounted for the greatest attrition across cohorts, with median losses of ~35%, ~30%, and ~22%, respectively (ranges: 0-86%, 0-61%, and 8-29%), driven primarily by clinical ineligibility. In contrast, most structural trial criteria resulted in minimal attrition (<10%). Missingness patterns differed across sources, underscoring where incomplete capture may affect ECA feasibility. Iterative clarification, documentation of operational decisions, and cross-partner feedback supported alignment and baseline balance assessment. Next steps include outcome comparisons with the target trial control arm.
CONCLUSIONS: Feasible ECA construction depends on the strategic application of trial-defined criteria to ensure comparability with the target control arm, together with careful management of data limitations. Eligibility-related attrition and covariate missingness highlight the need for large, high-quality datasets and harmonized processes so ECAs can approximate target-trial populations and support regulatory decision-making.
METHODS: We developed a statistical analysis plan (SAP) to translate target-trial eligibility criteria, propensity score modeling, and variable definitions into standardized operational rules. Nine groups applied the SAP to independent RWD or HCT datasets, documenting operational choices and attrition. We examined missingness to understand variable capture and identify gaps affecting feasibility and covariate balance. Propensity score matching and baseline comparisons were conducted to evaluate balance between external and target control arms.
RESULTS: Three eligibility criteria—adequate hepatic/renal function, adequate hematologic function, and no prior systemic therapy—accounted for the greatest attrition across cohorts, with median losses of ~35%, ~30%, and ~22%, respectively (ranges: 0-86%, 0-61%, and 8-29%), driven primarily by clinical ineligibility. In contrast, most structural trial criteria resulted in minimal attrition (<10%). Missingness patterns differed across sources, underscoring where incomplete capture may affect ECA feasibility. Iterative clarification, documentation of operational decisions, and cross-partner feedback supported alignment and baseline balance assessment. Next steps include outcome comparisons with the target trial control arm.
CONCLUSIONS: Feasible ECA construction depends on the strategic application of trial-defined criteria to ensure comparability with the target control arm, together with careful management of data limitations. Eligibility-related attrition and covariate missingness highlight the need for large, high-quality datasets and harmonized processes so ECAs can approximate target-trial populations and support regulatory decision-making.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
RWD154
Topic
Real World Data & Information Systems
Topic Subcategory
Reproducibility & Replicability
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, SDC: Oncology