Leveraging External Data for Testing Experimental Therapies with Biomarker Interactions in Randomized Clinical Trials

Area of Science:

• Oncology
• Biostatistics
• Clinical Trial Design

Background:

• Therapeutic efficacy in oncology frequently varies among patient subgroups, posing challenges for early drug development.
• Randomized clinical trials may yield false negative results when treatment effects are concentrated in subpopulations.
• External data from clinical studies and electronic health records can enhance decision-making in drug development.

Purpose of the Study:

• To introduce a novel permutation procedure for testing treatment efficacy in subpopulations using external data.
• To increase the statistical power of randomized clinical trials when heterogeneous treatment effects are present.
• To control the false positive rate while accommodating discrepancies between trial and external data.

Main Methods:

• A permutation procedure is proposed to test the null hypothesis of no treatment effect in any subpopulation.
• External data is leveraged to enhance the power of the statistical test.
• The procedure is designed to maintain the desired false positive rate ($\alpha$) without strict assumptions on external data.

Main Results:

• The permutation test increases power for detecting heterogeneous treatment effects.
• The method controls the false positive rate even with unmeasured confounders or differing patient profiles.
• The test is shown to be optimal and is illustrated with simulations and glioblastoma trial data.

Conclusions:

• External data can be effectively used with a permutation test to evaluate experimental therapeutics with potential heterogeneous effects.
• This approach improves the detection of subgroup-specific treatment benefits in oncology.
• The proposed method offers a robust tool for clinical trial analysis, enhancing drug development decision-making.