Statistical models for Alzheimer's disease clinical trials: Lessons learned from the DIAN-TU Platform Trial

Area of Science:

• Neuroscience
• Biostatistics
• Clinical Trials

Background:

• Alzheimer's disease (AD) clinical trials frequently feature variable follow-up times and open-label extensions (OLE).
• Traditional statistical models, designed for fixed schedules, are inefficient for these complex data structures.

Purpose of the Study:

• To present and demonstrate advanced statistical modeling techniques for optimizing data analysis in AD trials with irregular and extended follow-up.
• These methods were developed and implemented within the Dominantly Inherited Alzheimer Network Trials Unit platform trial.

Main Methods:

• A Cox proportional hazards model for recurrent events, utilizing all observed worsening events.
• A parametric disease progression model estimating time delays or proportional slowing from expected symptom onset.
• Piecewise linear mixed-effects models addressing the interim period between double-blind and OLE phases, handling missing data and variable off-treatment intervals.

Main Results:

• The proposed models effectively manage complex longitudinal data common in trials with OLE phases.
• These approaches offer enhanced statistical efficiency and a more thorough assessment of treatment effects over extended durations compared to conventional methods.

Conclusions:

• Flexible and efficient statistical models are suitable for rare disease and long-duration AD trials.
• Wider adoption and validation of these methods can increase the power and interpretability of future neurodegenerative disease research.