Assessing and interpreting treatment effects in longitudinal clinical trials with missing data

Area of Science:

• Biostatistics
• Clinical Trials
• Longitudinal Data Analysis

Background:

• Longitudinal data analysis is crucial for evaluating treatment effects over time.
• Missing data in clinical trials poses significant analytical challenges.
• Last Observation Carried Forward (LOCF) has been a traditional but potentially flawed method for handling missing data.

Purpose of the Study:

• To review characteristics of missing data and methods for handling them in longitudinal clinical trials.
• To propose a robust analytical approach for longitudinal data.
• To compare the utility of mixed-effects modeling versus LOCF.

Main Methods:

• Review of statistical methodologies for longitudinal data analysis.
• Examination of characteristics influencing analytic choices for missing data.
• Evaluation of common missing data handling methods, including LOCF.
• Proposal of a primary analysis using repeated measures, likelihood-based, mixed-effects modeling.

Main Results:

• Repeated measures mixed-effects modeling offers a robust primary analysis strategy.
• LOCF can serve as a secondary, composite measure for efficacy, safety, and tolerability.
• The study illustrates how repeated-measures analyses enhance clinical trial decision-making.

Conclusions:

• Mixed-effects modeling is recommended for primary efficacy analysis in longitudinal clinical trials.
• LOCF should be considered a secondary measure due to its limitations.
• Careful consideration of data characteristics and analytical methods is essential for valid interpretation of longitudinal trial data.