Considerations to address missing data when deriving clinical trial endpoints from digital health technologies

Area of Science:

• Biostatistics
• Digital Health
• Clinical Trial Methodology

Background:

• Digital health technologies (DHTs) offer remote, continuous physiological and behavioral monitoring.
• Accelerated DHT adoption in clinical trials highlights the need for robust missing data handling.
• Existing methods for missing data in clinical trials are often inadequate for complex DHT data.

Purpose of the Study:

• To identify and review advanced statistical approaches for addressing missing data in DHT-derived clinical trial endpoints.
• To discuss strategies for minimizing missing data through optimized DHT deployment and patient-centered study design.

Main Methods:

• Review of missing data characteristics specific to DHTs (e.g., high-frequency time series).
• Discussion of emerging statistical methods for epoch-level data: within-patient imputation, functional data analysis, deep learning.
• Exploration of imputation and robust modeling for daily summary measures derived from DHTs.

Main Results:

• Current DHT missing data approaches are often unsophisticated, relying on data exclusion.
• Advanced methods like functional data analysis and deep learning show promise for epoch-level data.
• Robust modeling and imputation strategies are suitable for daily summary measures.

Conclusions:

• Novel statistical methods are crucial for valid and reliable digital endpoints in clinical trials.
• Proactive strategies, including optimizing DHT use and incorporating patient perspectives, can minimize missing data.
• This review serves as a foundation for further discussion on managing missing data with DHTs.