Partition testing for real-world evidence studies.

Area of Science:

• Pharmacovigilance
• Biostatistics
• Clinical Trial Methodology

Background:

• Traditional hypothesis testing in post-approval studies is limited, often restricting inferences to superiority or inferiority.
• There is a need for methods that can simultaneously assess multiple hypotheses, including equivalence and non-inferiority.

Purpose of the Study:

• To introduce and illustrate the application of a partition testing framework for comparative medical product studies.
• To demonstrate how partition testing allows for a broader range of inferences beyond simple superiority or inferiority.

Main Methods:

• The partition testing framework requires selecting an equivalence margin and calculating a two-sided Wald confidence interval.
• This methodology was applied to a published study on Ondansetron safety during pregnancy.

Main Results:

• Applying partition testing to the Ondansetron study indicated equivalence in the risk of cardiac malformations and oral clefts between exposed and unexposed women.
• This contrasts with traditional testing, which focused solely on inferiority/superiority.

Conclusions:

• Partition testing enables a more comprehensive and nuanced assessment of medical product safety and effectiveness.
• Pre-specifying an equivalence margin enhances transparency and impartiality in evaluating clinical importance, mitigating publication bias.