Noninferiority testing for matched-pair ordinal data with misclassification.

Area of Science:

• Biostatistics
• Clinical Trials
• Medical Research Methodology

Background:

• Noninferiority testing is crucial for evaluating new treatments that offer advantages beyond superiority, such as improved safety or convenience.
• Categorical variables, particularly ordinal ones from patient-reported outcomes, are common in medical research.
• Existing noninferiority tests may not adequately address correlated ordinal data or potential misclassification issues.

Purpose of the Study:

• To develop a novel noninferiority testing procedure for correlated ordinal categorical variables.
• To extend the procedure to handle misclassification in ordinal data using partially validated data.
• To evaluate the performance of the proposed method through simulation studies.

Main Methods:

• Development of a noninferiority testing procedure based on a latent normal distribution approach for paired ordinal categorical data.
• Incorporation of a method to address misclassification using partially validated data.
• Conducting simulation studies to assess accuracy, type I error rates, and statistical power.

Main Results:

• The proposed noninferiority testing procedure demonstrates accurate estimation and controlled type I error rates.
• The method effectively handles correlated ordinal categorical variables and accounts for misclassification.
• Simulation results indicate good power for the developed procedure.

Conclusions:

• The developed noninferiority testing procedure provides a robust tool for analyzing correlated ordinal categorical data in clinical research.
• The approach is valuable for situations involving patient-reported outcomes where misclassification is a concern.
• This methodology enhances the evaluation of new treatments that are noninferior to existing ones but offer other patient benefits.