Randomization tests in clinical trials with multiple imputation for handling missing data.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Psychiatric Research

Background:

• Traditional statistical inference in clinical trials often relies on population models.
• Missing data is a common challenge in clinical trials, frequently addressed by multiple imputation.
• Randomization-based inference provides a valuable alternative framework for data analysis.

Purpose of the Study:

• To describe the construction of a randomization test suitable for clinical trials incorporating multiple imputation.
• To demonstrate the application of this methodology in real-world clinical trial data.
• To offer a statistically sound approach for analyzing randomized controlled trials with missing data.

Main Methods:

• Development of a randomization test procedure that integrates with multiple imputation techniques.
• Application of Fisher's combining function to individual scores derived from imputed datasets.
• Utilizing data from two clinical trials focused on post-traumatic stress disorder.

Main Results:

• The proposed methodology successfully integrates randomization tests with multiple imputation for handling missing data.
• Fisher's combining function effectively aggregated results across imputed datasets for randomization testing.
• The approach was validated using data from post-traumatic stress disorder trials.

Conclusions:

• Constructing randomization tests in the presence of missing data is feasible using multiple imputation.
• This integrated approach enhances the reliability of statistical inference in clinical trials.
• The method provides a robust framework for analyzing clinical trial data, particularly in psychiatric research.