A causal modelling framework for reference-based imputation and tipping point analysis in clinical trials with...

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Causal Inference

Background:

• Handling treatment discontinuation in clinical trials is crucial for accurate estimation.
• Existing reference-based imputation methods lack formal justification.
• Estimands may follow treatment policy or hypothetical strategies, especially with rescue medication.

Purpose of the Study:

• To present a causal model for estimating treatment effects in the presence of treatment discontinuation.
• To formally justify and contextualize reference-based imputation methods within a causal framework.
• To provide a flexible framework for sensitivity analysis regarding post-discontinuation treatment effects.

Main Methods:

• Developed a potential outcomes framework with explicit assumptions on maintained causal effects post-discontinuation.
• Utilized mathematical arguments and simulation studies to evaluate imputation methods.
• Applied the causal model to data from two longitudinal clinical trials.

Main Results:

• Demonstrated that "jump to reference," "copy reference," and "copy increments in reference" are special cases of the proposed causal model.
• Showcased the causal model's flexibility for tipping point sensitivity analysis.
• Illustrated the practical application of the framework in real-world clinical trial data.

Conclusions:

• The proposed causal model offers a formally justified and transparent approach to handling treatment discontinuation in drug trials.
• Reference-based imputation methods can be understood as specific instances of this causal model.
• The framework facilitates robust sensitivity analyses for estimating causal treatment effects.