Identifying Fit-for-Emulation Data: Adaptation of a Structured Data Feasibility Assessment Process for Real-World Oncology Trial Emulations
Author(s)
Levy N1, Campbell U1, Sheridan P1, Lenis D2, Madsen A3, O'Doherty I1, Estrin A4, Iyer M1, McDonald S3, Becnel L5, Belli A6, Carrigan G7, Chan KA8, Chen J9, Chia VM10, Dhopeshwarkar N11, Eckert JC12, Fernandes L13, Goldstein M14, Greshock J15, Hendricks-Sturrup R16, Huang J17, Jiao X18, Khosla S19, Lunacsek O20, McRoy L21, Natanzon Y22, Ovbiosa O23, Pace ND23, Pinheiro S23, Quinn J24, Rees M25, Rider J26, Rimawi M27, Robinson T25, Rodriguez-Watson C28, Sangli C29, Sarsour K15, Schneeweiss S30, Shapiro M31, Stewart M32, Taylor A17, Wang C6, Wasserman A24, Zhang Y19
1Aetion, Inc, New York, NY, USA, 2Aetion, Inc., NY, NY, USA, 3Aetion, Inc, Brooklyn, NY, USA, 4Aetion Inc., New York, NY, USA, 5Pfizer, New York, NY, USA, 6COTA, Inc, Boston, MA, USA, 7Amgen Inc., Thousand Oaks, CA, USA, 8TriNetX, LLC, Cambridge, MA, USA, 9Tempus, Chicago, IL, USA, 10Amgen Inc, Thousand Oaks, CA, USA, 11TriNetX, Cambridge, MA, USA, 12Reagan-Udall Foundation for the Food and Drug Administration, Washington, DC, USA, 13ConcertAI, New York, NY, USA, 14XCures, oakland, CA, USA, 15Johnson and Johnson, New Brunswick, NJ, USA, 16Duke-Margolis Center for Health Policy, Washington, DC, USA, 17Gilead Sciences, Foster City, CA, USA, 18Duke Margolis Center, Washington, DC, USA, 19AstraZeneca, Gaithersburg, MD, USA, 20Bayer, Whippany, NJ, USA, 21Pfizer, Inc., New York, NY, USA, 22ConcertAI, Cambridge, MA, USA, 23AbbVie, North Chicago, IL, USA, 24xCures, Oakland, CA, USA, 25Loopback Analytics, Dallas, TX, USA, 26Aetion, Inc, Boston, MA, USA, 27Baylor College of Medicine, Houston, TX, USA, 28Reagan-Udall Foundation for the FDA, Washington DC, DC, USA, 29Tempus AI, Inc., Chicago, IL, USA, 30Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA, 31xCures, DURHAM, NC, USA, 32Friends of Cancer Research, Washington, DC, USA
Presentation Documents
OBJECTIVES: The Coalition to Advance Real-World Evidence through Randomized Controlled Trial Emulation (CARE) Initiative seeks to advance understanding of when real-world data (RWD) can generate valid treatment effectiveness estimates by emulating oncology randomized controlled trials (RCTs) using RWD. Successful emulation requires fit-for-emulation data. We describe learnings from a structured feasibility assessment process to evaluate potential datasets for CARE studies, which may inform other RCT emulations.
METHODS: Candidate RCTs were identified from active comparator trials for common tumor types leading to approvals during 2015-2020. Feasibility assessments included two phases. First, in each potential dataset, we confirmed availability of the RCT indication and outcomes and sample size ≥1.5-times the RCT population, estimated as counts of patients with the indication receiving an RCT treatment or comparator therapy. Second, we modified the Structured Process to Identify Fit-For-Purpose Data (SPIFD2) framework to conduct detailed data fitness assessments. Key RCT design elements (e.g., research question, treatment strategies, eligibility criteria, outcomes, covariates) and potential confounders were identified. The ability to operationalize each element was assessed and ranked (1-low, 5-high), based on reliability/validation of measures and missingness. Overall ratings were calculated by averaging required design element rankings.
RESULTS: Of 49 possible RCT-dataset combinations, 9 passed initial screening and proceeded to detailed feasibility assessment. Key drivers of overall ratings included: availability of performance status, non-cancer diagnoses/treatments, and progression measures; diagnosis date quality; and death data validity. Measurable disease and prognosis eligibility criteria were not captured in any dataset. Two datasets were determined fit-for-emulation of two RCTs (n=3 emulations).
CONCLUSIONS: Oncology RCT emulations require specific eligibility criteria and outcomes that make identifying fit-for-emulation RWD especially challenging. In particular, routinely-captured, non-cancer diagnosis/treatment variables are absent from datasets containing high-quality oncology information. Rigorous feasibility assessments are critical for identifying fit-for-emulation RWD, contextualizing results, and identifying gaps in existing datasets.
Conference/Value in Health Info
Value in Health, Volume 27, Issue 12, S2 (December 2024)
Code
RWD125
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, Oncology