A non-parametric U-statistic testing approach for multi-arm clinical trials with multivariate longitudinal data

Area of Science:

• Biostatistics
• Clinical Trial Design
• Longitudinal Data Analysis

Background:

• Randomized clinical trials (RCTs) frequently assess multiple longitudinal outcomes to evaluate treatment effectiveness.
• The existing Longitudinal Rank-Sum Test (LRST) is effective for two-arm comparisons but lacks multi-arm applicability.
• Assessing multiple doses against a control group is common in RCTs, necessitating advanced statistical methods.

Purpose of the Study:

• To extend the Longitudinal Rank-Sum Test (LRST) to a multi-arm setting for comprehensive treatment efficacy evaluation.
• To develop a flexible and powerful method for comparing multiple doses against a control group in RCTs.
• To enhance the detection of the most effective dose in multi-arm clinical trials.

Main Methods:

• Development of a novel multi-arm Longitudinal Rank-Sum Test (LRST).
• The method is a U-statistics-based, non-parametric, rank-based approach leveraging temporal data structures.
• Extensive simulations were conducted to evaluate Type I error control and statistical power.

Main Results:

• The multi-arm LRST maintains excellent Type I error control across various scenarios.
• The proposed method demonstrates greater statistical power compared to traditional two-arm LRST with multiplicity adjustments.
• Simulations confirmed the method's robustness and effectiveness in identifying optimal treatment doses.

Conclusions:

• The multi-arm LRST offers a robust and powerful solution for analyzing efficacy in multi-arm clinical trials with longitudinal outcomes.
• This extension addresses the limitations of the two-arm LRST, providing a valuable tool for dose-finding studies.
• The method's practical utility was validated through its application to the Bapineuzumab (Bapi) 301 trial.