Improving efficiency of inference in clinical trials with external control data

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Epidemiology

Background:

• Clinical trials often face limited statistical power due to small sample sizes.
• External control data from historical studies are frequently available but underutilized.
• Integrating external data can potentially overcome sample size limitations in clinical trials.

Purpose of the Study:

• To improve the efficiency of statistical inference in clinical trials by incorporating external control data.
• To develop a robust and efficient estimator for the average treatment effect (ATE) using both clinical trial and external data.
• To explore the extrapolation of causal effects to different populations and inform trial design.

Main Methods:

• Developed a semiparametric approach assuming exchangeability of potential outcome means.
• Derived a doubly robust and locally efficient estimator for the average treatment effect.
• Investigated scenarios with relaxed overlap assumptions, including trials with only a treated group.

Main Results:

• Incorporating external control data significantly reduces the semiparametric efficiency bound for ATE estimation.
• Efficiency gains are most pronounced with large, low-variability external control datasets.
• The proposed methods demonstrated effective extrapolation of causal effects and showed superior performance in simulations.

Conclusions:

• Borrowing strength from external control data is a viable strategy to enhance clinical trial efficiency and inference.
• The developed doubly robust and locally efficient estimators provide a powerful tool for analyzing integrated data.
• Findings have significant implications for optimizing clinical trial design and data collection strategies, as seen in the Helicobacter pylori infection example.