Estimating treatment effect in randomized trial after control to treatment crossover using external controls

Area of Science:

• Clinical Trial Design and Methodology
• Biostatistics
• Causal Inference

Background:

• Clinical trials often allow crossover to experimental treatments post-primary endpoint, particularly in open-label extension phases.
• Assessing long-term safety and efficacy in these phases is crucial but challenging without placebo controls.
• External controls offer a potential solution for robust estimation when placebo groups are unavailable.

Purpose of the Study:

• To propose novel causal inference methods for estimating long-term treatment effects using combined randomized controlled trials (RCTs) and external controls.
• To evaluate the performance of proposed difference-in-differences (DID) and synthetic control methods (SCM) in realistic scenarios.

Main Methods:

• Development of several difference-in-differences (DID) type estimators.
• Implementation of a synthetic control method (SCM).
• Combination of randomized controlled trial data with external control data within a causal inference framework.

Main Results:

• Simulation studies demonstrated the desirable performance of the proposed estimators across various practical scenarios.
• Difference-in-differences (DID) methods consistently outperformed the synthetic control method (SCM).
• The methods were successfully applied to a phase III clinical trial in a rare disease setting.

Conclusions:

• Combining RCTs with external controls is feasible for estimating long-term treatment effects in open-label extension phases.
• Proposed DID methods are effective and recommended for this purpose.
• The approach is particularly valuable for rare disease clinical trials where traditional controls may be limited.