Applying Bias Correction Methods to Build Hybrid Controls Using Real-World Patients for a Phase IIb Randomized...

Area of Science:

• Rheumatology
• Clinical Epidemiology
• Data Science

Background:

• Randomized controlled trials (RCTs) face challenges when incorporating real-world (RW) patients due to potential selection bias and unmeasured confounding.
• Hybrid control arms, combining RCT controls with eligible RW patients, offer a potential solution to enhance generalizability.
• Evaluating methods to construct these hybrid arms is crucial for robust clinical trial evidence.

Purpose of the Study:

• To assess two methods, Match, Test-then-Pool (MTP) and Matching and Bias Adjustment (MBA), for constructing a hybrid control arm.
• To replicate the treatment effect of baricitinib on rheumatoid arthritis (RA) outcomes using this hybrid approach.
• To compare the performance of MTP and MBA in a moderately to severely active RA population.

Main Methods:

• Utilized concurrent RCT control patients and trial-eligible RW patients from the CorEvitas Rheumatoid Arthritis Registry.
• Applied optimal propensity score (PS) matching to create comparable control groups.
• MTP involved replacing RCT controls with RW controls and testing for outcome differences before pooling; MBA adjusted for residual differences between RCT and RW controls.

Main Results:

• Both MTP and MBA methods successfully replicated the original RCT's findings for baricitinib's treatment effect on ACR20 response and CDAI change.
• Optimal matching resulted in a RW cohort with good covariate balance compared to RCT participants.
• The odds ratio for ACR20 was 4.71 (MTP) and 4.76 (MBA), closely mirroring the RCT's 4.74.
• Mean CDAI change was -8.96 (MTP) and -9.85 (MBA), consistent with the RCT's -8.88.

Conclusions:

• Hybrid control arms constructed using MTP or MBA, combining well-matched RW data with RCT controls, can effectively replicate Phase IIb trial results.
• These methods provide a viable strategy to enhance the generalizability of RCT findings by incorporating real-world evidence.
• The study demonstrates the successful application of advanced statistical techniques in real-world data analysis for rheumatoid arthritis.