On model selections for repeated measurement data in clinical studies

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Longitudinal Data Analysis

Background:

• Repeated measurement designs are common in randomized controlled trials (RCTs) for long-term efficacy.
• Traditional t-tests at fixed time points underutilize longitudinal data.
• Mixed-effects models offer a way to analyze all available data but may not be optimal for specific research questions.

Purpose of the Study:

• To propose a novel analytic strategy for longitudinal clinical trial data.
• To enhance the mixed-effects model by integrating a model selection scheme.
• To improve the utilization of all available data for robust treatment effect assessment.

Main Methods:

• Coupling mixed-effects models with a model selection scheme for longitudinal data.
• Developing new test statistics that leverage optimal model information.
• Evaluating performance via extensive Monte Carlo simulations under various missing data mechanisms.

Main Results:

• The proposed method is more powerful than the t-test for detecting treatment effects at the last time point.
• It outperforms standard mixed-effects models in testing overall treatment effects across time.
• The framework demonstrates superior robustness and flexibility in handling missing data compared to competing methods.

Conclusions:

• The proposed analytic strategy offers a powerful and flexible approach for longitudinal clinical trial data analysis.
• It effectively utilizes all available data and provides robust results, particularly with missing data.
• The method is applicable to real-world clinical trials, as shown in an analysis of a testosterone intervention study.