Estimating hypothetical estimands with causal inference and missing data estimators in a diabetes trial case study

Area of Science:

• Biostatistics
• Clinical Trials
• Pharmaceutical Research

Background:

• The ICH E9 addendum offers guidance on estimands but lacks detail on estimation methods.
• Precisely defining treatment effects is crucial for interpretable clinical trial results.

Purpose of the Study:

• To evaluate multiple statistical methods for estimating treatment effects using the hypothetical strategy.
• To demonstrate the application of these methods in a type 2 diabetes clinical trial.
• To compare the performance and practical considerations of different estimation approaches.

Main Methods:

• Analysis of a type 2 diabetes clinical trial data.
• Application of mixed models for repeated measures (MMRM).
• Utilization of multiple imputation (MI).
• Implementation of inverse probability of treatment weighting (IPTW).
• Employing the G-formula and G-estimation techniques.
• Utilizing R packages for statistical analysis.

Main Results:

• Broadly similar estimates and standard errors were observed across the evaluated estimation methods.
• The hypothetical strategy was successfully applied to handle treatment discontinuation and rescue medication.
• Practical implementation details and assumptions for each method were described.

Conclusions:

• Multiple statistical methods can effectively estimate treatment effects under the hypothetical strategy.
• The choice of estimation method involves considerations such as computational efficiency and handling of missing data.
• Further research is needed on modeling intercurrent events and time-varying confounders.