A comparison of analysis procedures for correlated binary data in dedicated multi-rater imaging trials

Area of Science:

• Biostatistics
• Clinical Trial Analysis
• Statistical Methods

Background:

• Correlated binary data arise in clinical trials with multiple raters.
• Defining success criteria for regulatory imaging trials with multiple raters is crucial.
• No inherent order exists for measurements from different raters.

Purpose of the Study:

• To describe and investigate three analysis procedures for repeated correlated binary data.
• To compare the performance of 'Majority rater,' 'Average rater,' and 'Majority significant' methods.
• To provide an analytical result for the expectation of the 'Majority rater' method.

Main Methods:

• Analytical derivation for the expectation of the 'Majority rater'.
• Simulation study comparing the three analysis procedures.
• Investigation of statistical power and properties of each method.

Main Results:

• The 'Average rater' method frequently offers a gain in statistical power.
• 'Majority significant' demonstrates favorable properties in certain settings.
• 'Majority rater' is often challenging to interpret.

Conclusions:

• The choice of analysis method impacts statistical power and interpretability.
• 'Average rater' is generally advantageous for power in correlated binary data analysis.
• Further research may refine the application of 'Majority significant' in specific trial designs.