An E-value-Informed Sensitivity Analysis Framework for Hybrid Controlled Trials.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Real-World Evidence Integration

Background:

• Hybrid controlled trials (HCTs) combine internal randomized controlled trial (RCT) data with external real-world data (RWD).
• HCTs can enhance statistical power and improve patient recruitment through unequal randomization, but face threats to validity from unmeasured confounding in external controls.

Purpose of the Study:

• To develop and evaluate a sensitivity analysis framework for assessing the robustness of HCT results to unmeasured confounding.
• To introduce a practical, interpretable method for rigorous inference when integrating RWD into RCTs.

Main Methods:

• Developed a tipping point analysis, adapting the E-value framework to the HCT setting where confounding affects trial participation.
• Introduced a data-driven benchmark for interpreting the strength of unmeasured confounding.
• Proposed an operational decision rule and evaluated its performance via simulation studies and an asthma trial example.

Main Results:

• The proposed decision rule effectively controlled Type I error near the nominal 5% level, even with moderate unmeasured confounding.
• Power increased by 10-20% compared to RCTs alone when incorporating external data.
• The framework demonstrated robustness and preserved power gains from HCTs.

Conclusions:

• The developed sensitivity analysis framework provides a practical and interpretable method for assessing the robustness of HCTs.
• This approach supports rigorous inference when integrating external real-world data into clinical trials.
• The methodology enhances the reliability of hybrid controlled trials, balancing power gains with validity concerns.