Coping with missing data in clinical trials: a model-based approach applied to asthma trials

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Longitudinal Data Analysis

Background:

• Patient drop-out is common in clinical trials, leading to missing data and complicating repeated measures analysis.
• Standard methods for handling missing data, like last observation carried forward, may introduce bias.
• Understanding reasons for drop-out is crucial for appropriate statistical modeling.

Purpose of the Study:

• To review reasons for patient drop-out and their impact on clinical trial analysis.
• To propose novel statistical models for simultaneously analyzing treatment response and the drop-out process.
• To compare the performance of proposed models with standard missing data techniques.

Main Methods:

• Development of a class of models to jointly model response and drop-out.
• Application of a Bayesian framework using non-informative priors and BUGS software.
• Proposal of a time transformation to linearize asymptotic repeated measures patterns.
• Illustration with data from a five-arm asthma clinical trial.

Main Results:

• The proposed Bayesian models provide a robust approach to handling missing data due to drop-out.
• Comparison with standard methods like last observation carried forward demonstrates potential advantages.
• Time transformation simplifies modeling of longitudinal data with asymptotic patterns.

Conclusions:

• Joint modeling of treatment response and drop-out offers a more accurate analysis of clinical trial data.
• Bayesian methods with BUGS are suitable for implementing these complex models.
• The proposed techniques enhance the reliability of findings from clinical trials with missing data.