Semi-parametric sensitivity analysis for trials with irregular and informative assessment times

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Epidemiology

Background:

• Clinical trials often collect outcome data at pre-specified times post-randomization.
• Variability in actual participant assessment times complicates treatment effect estimation.
• Existing methods for irregular assessment times rely on untestable assumptions, necessitating sensitivity analyses.

Purpose of the Study:

• To develop a novel sensitivity analysis methodology for clinical trials with irregular and informative assessment times.
• To provide a robust approach for estimating treatment effects when data collection deviates from the planned schedule.
• To benchmark the new methodology against the explainable assessment (EA) assumption.

Main Methods:

• Developed a sensitivity analysis anchored to the explainable assessment (EA) assumption.
• Employed an exponential tilting approach, controlled by a sensitivity parameter, to model deviations from the EA assumption.
• Utilized a new influence function-based, augmented inverse intensity-weighted estimator for inference.
• Enabled flexible semiparametric modeling of observed data, decoupled from sensitivity parameter specification.

Main Results:

• The proposed methodology offers a flexible framework for handling irregular assessment times in randomized trials.
• The influence function-based estimator provides a robust approach to treatment effect estimation under sensitivity assumptions.
• The method was successfully applied to a real-world randomized trial involving individuals with uncontrolled asthma.

Conclusions:

• The developed sensitivity analysis methodology enhances the reliability of treatment effect estimation in trials with non-standard assessment schedules.
• The approach provides a valuable tool for researchers to assess the impact of deviations from planned data collection.
• The detailed illustration of implementation facilitates the adoption of this method in future clinical research.