Correct and logical causal inference for binary and time-to-event outcomes in randomized controlled trials

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Pharmacogenomics

Background:

• Targeted therapies often exhibit differential efficacy across biomarker-defined subgroups.
• Current statistical methods in randomized controlled trials (RCTs) for analyzing these subgroups contain critical oversights.
• Existing approaches can lead to misidentification of predictive biomarkers and misleading efficacy interpretations.

Purpose of the Study:

• To identify and correct three prevailing oversights in stratified analyses of RCTs with binary and time-to-event outcomes.
• To propose a novel statistical principle, subgroup mixable estimation (SME), for accurate causal inference.
• To ensure reliable treatment effect estimation in biomarker-defined subgroups.

Main Methods:

• Analytical derivations demonstrating how common efficacy measures (odds ratio, hazard ratio) fail causal estimand requirements (ICH E9R1).
• Analysis of real-world immunotherapy patient data to validate theoretical findings.
• Development and application of the subgroup mixable estimation (SME) principle, which averages probabilities within treatment arms before computing subgroup efficacy.

Main Results:

• Odds ratios and hazard ratios can erroneously suggest prognostic biomarkers are predictive due to confounding factors.
• Mixing subgroup efficacies on a logarithmic scale creates artificial population estimands.
• The SME principle provides accurate causal inference by averaging probabilities and computing simultaneous confidence intervals for subgroup and mixed efficacies.

Conclusions:

• Prevailing statistical practices in RCT subgroup analysis are flawed and require correction.
• The SME principle offers a rigorous mathematical framework for reliable causal inference in targeted therapy trials.
• Implementing SME ensures that biomarker-defined subgroups receive appropriate and effective treatments.