Using Off-Treatment Sequential Multiple Imputation for Binary Outcomes to Address Intercurrent Events Handled by a...

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Pharmaceutical Research

Background:

• The estimand framework offers strategies for intercurrent events in clinical trials.
• The treatment policy strategy aligns with the intention-to-treat principle.
• Handling missing data due to patient withdrawal is crucial for accurate estimation.

Purpose of the Study:

• To explore and compare retrieved dropout models for estimating treatment effects with binary endpoints under a treatment policy strategy.
• To assess the performance of these models via simulation and a practical example.
• To identify the most pragmatic retrieved dropout model for handling intercurrent events like treatment discontinuation.

Main Methods:

• Investigated various retrieved dropout models for imputing missing post-intercurrent event data.
• Compared proposed models against a simple imputation model using a simulation study.
• Applied and evaluated models using Phase 3 clinical trial data for rheumatoid arthritis.

Main Results:

• A simple retrieved dropout model with an intercurrent event indicator proved most pragmatic.
• The model's effectiveness is contingent on having at least 50% of observed post-intercurrent event data.
• Model suitability depends on the availability of observed data post-intercurrent event.

Conclusions:

• Retrieved dropout models offer a viable approach for estimating treatment effects with binary outcomes and intercurrent events.
• The pragmatic choice requires sufficient observed data post-intercurrent event for reliable imputation.
• Careful consideration of data availability is essential before implementing these models in clinical trials.