A meta-analytic framework to adjust for bias in external control studies.

Area of Science:

• Biostatistics
• Clinical Trials
• Real-World Data Analysis

Background:

• Randomized controlled trials (RCTs) are standard for treatment effect estimation.
• Real-world data (RWD) is increasingly used in drug development, especially for external control arms in single-arm trials.
• Non-randomized external control arms can introduce bias due to population differences.

Approach:

• Developed a meta-analytic framework to adjust log hazard ratio estimates for bias and variability.
• Uses historical reference studies to compare trial controls against constructed external control arms.
• External control studies can be performed independently using causal inference techniques.

Key Points:

• The framework corrects for bias and variability in time-to-event outcome analyses.
• Historical studies are used to meta-analyze trial controls versus external control arms.
• Empirical analysis demonstrated bias correction in advanced non-small cell lung cancer studies.

Conclusions:

• The proposed meta-analytic framework effectively adjusts for bias in external control arms.
• This methodology enhances the reliability of treatment effect estimation using RWD.
• An R package 'ecmeta' is available for implementing the approach.