Implementation of tripartite estimands using adherence causal estimators under the causal inference framework

Area of Science:

• Clinical Trials Methodology
• Causal Inference
• Biostatistics

Background:

• Intercurrent events (ICEs) and missing data complicate treatment effect assessment in clinical trials.
• Standard analyses may not fully capture the nuances of treatment outcomes when ICEs occur.
• Defining relevant estimands is crucial for meaningful data analysis.

Purpose of the Study:

• To detail the estimation and interpretation of tripartite estimands within a causal inference framework.
• To provide a method for understanding the totality of treatment effects, considering various scenarios of ICEs.
• To illustrate the application of tripartite estimands in a real-world clinical trial setting.

Main Methods:

• Utilizing a causal inference framework to define and estimate tripartite estimands.
• The tripartite estimands assess treatment differences in: proportion with ICEs due to adverse events, proportion with ICEs due to lack of efficacy, and primary outcome for treatment adherents.
• Application demonstrated in a Phase 3 basal insulin trial for type 1 diabetes.

Main Results:

• The manuscript provides a methodological discussion on estimating tripartite estimands.
• Interpretation guidelines for tripartite estimates are presented.
• The approach offers a more complete picture of treatment effects compared to traditional methods.

Conclusions:

• Tripartite estimands offer a robust framework for analyzing clinical trial data with intercurrent events and missing values.
• This method enhances the understanding of treatment effects for diverse patient outcomes.
• The causal inference approach provides valuable insights for stakeholders in clinical research.