Sample-size calculations for studies with correlated ordinal outcomes

Area of Science:

• Biostatistics
• Public Health Research
• Clinical Trial Design

Background:

• Correlated ordinal response data are common in public health studies.
• Accurate sample-size calculations are vital for study design to ensure sufficient power.
• Existing methods for sample-size estimation often do not adequately address ordinal data.

Purpose of the Study:

• To extend existing sample-size calculation methods for repeated binary responses to the correlated ordinal data setting.
• To provide a robust method for sample-size estimation in public health studies involving ordinal outcomes.
• To ensure adequate statistical power for detecting significant effects in studies with ordinal data.

Main Methods:

• The proposed method extends Rochon's approach for repeated binary responses to the ordinal case.
• Sample-size calculations are based on generalized estimating equations (GEE).
• Inference is based on the Wald test for statistical significance.

Main Results:

• Simulation studies confirmed the validity and utility of the proposed sample-size calculation method.
• The method provides accurate power calculations for studies with correlated ordinal outcomes.
• The approach is applicable to various public health research scenarios.

Conclusions:

• The developed method offers a valuable tool for sample-size determination in studies with correlated ordinal data.
• This approach enhances the design of public health studies by ensuring adequate statistical power.
• The methodology is illustrated with a practical example from an arthritis clinical trial.