Intent-to-treat analysis for longitudinal studies with drop-outs

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Longitudinal Data Analysis

Background:

• Intent-to-treat (IT) analysis is crucial for unbiased clinical trial evaluation.
• Traditional methods like Last Observation Carried Forward (LOCF) and random dropout models have limitations in IT settings.
• Handling dropouts in longitudinal clinical trials requires robust statistical approaches.

Purpose of the Study:

• To propose and evaluate a novel multiple imputation method for IT analysis of longitudinal data with dropouts.
• To address the drawbacks of existing methods in the context of IT analysis.
• To improve the accuracy of treatment effect estimation in the presence of missing data.

Main Methods:

• A novel multiple imputation approach using an "as treated" model for imputing missing values after dropout.
• Incorporation of actual or imputed doses post-dropout into the imputation models.
• Analysis of multiply-imputed datasets using IT principles, classifying subjects by randomization group.
• Combining results using Rubin's methods for combining multiply-imputed datasets.
• Utilizing distinct models for imputation and final analysis.

Main Results:

• The proposed method effectively handles missing longitudinal data in the presence of dropouts within an IT framework.
• It provides a more robust estimation of treatment effects compared to traditional methods.
• The application to a Tacrine trial for Alzheimer's disease demonstrates its practical utility.

Conclusions:

• The proposed multiple imputation method offers a significant advancement for intent-to-treat analysis in longitudinal clinical trials.
• This approach enhances the reliability of results when dealing with dropouts and varying treatment exposures.
• It is a valuable tool for analyzing complex clinical trial data, particularly in neurodegenerative disease research.