Stratified exact tests for the weak causal null hypothesis in randomized trials with a binary outcome.

Area of Science:

• Biostatistics
• Causal Inference
• Clinical Trials

Background:

• Fisher's exact test commonly analyzes binary outcomes in randomized trials, focusing on the sharp causal null hypothesis.
• It does not adequately test the weak causal null hypothesis, meaning rejection doesn't confirm a zero causal risk difference.

Purpose of the Study:

• To extend Chiba's exact test for the weak causal null hypothesis to stratified analyses in randomized trials.
• To provide a general method for estimating treatment effects adjusted for stratification factors.
• To enable straightforward extension to noninferiority trials and confidence interval construction.

Main Methods:

• Development of a novel stratified exact test for the weak causal null hypothesis.
• The proposed test is assumption-free and not reliant on large sample theory.
• General applicability for treatment effect estimation with adjustment for stratification factors.

Main Results:

• The proposed stratified exact test accurately evaluates the weak causal null hypothesis for binary outcomes.
• It offers a more general approach compared to Jung's stratified Fisher's exact test, which addresses the sharp null hypothesis.
• The method is adaptable for noninferiority trial analysis and confidence interval generation.

Conclusions:

• The new stratified exact test provides a robust method for causal inference with binary outcomes in stratified randomized trials.
• This approach improves upon existing methods by directly testing the weak causal null hypothesis.
• The test's flexibility supports its use in advanced trial designs and effect estimation.