genRCT: a statistical analysis framework for generalizing RCT findings to real-world population

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Health Services Research

Background:

• Randomized clinical trials (RCTs) are crucial for evaluating treatment efficacy but often suffer from generalizability bias.
• Differences in participant characteristics and risk factors between RCTs and real-world populations limit the applicability of trial findings.
• Observational studies offer large, representative samples but may have confounding biases.

Purpose of the Study:

• To review statistical methods for enhancing the generalizability of randomized clinical trial (RCT) findings.
• To leverage information from large, real-world observational studies to correct for generalizability bias in RCTs.
• To introduce and compare methods for estimating treatment effects across different endpoint types using combined data.

Main Methods:

• Discusses the selection of appropriate data sources and variables to satisfy key theoretical assumptions for generalizability.
• Introduces calibration weighting methods (genRCT) to enforce covariate balance between RCT and observational data.
• Compares estimation techniques for continuous, binary, and survival endpoints, including a case study using the `genRCT` R package.

Main Results:

• Demonstrates the application of genRCT methods to improve the generalizability of treatment effect estimates.
• Provides a framework for integrating RCT and observational data to obtain more representative real-world evidence.
• The case study successfully estimates the average treatment effect of adjuvant chemotherapy for stage 1B non-small cell lung cancer.

Conclusions:

• Statistical methods combining RCTs with large observational studies can overcome generalizability limitations.
• The genRCT framework offers a robust approach to harnessing real-world data for more reliable treatment effect estimation.
• This methodology enhances the clinical utility of RCT findings by ensuring relevance to diverse patient populations.