Estimating the treatment effect for adherers using multiple imputation.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Causal Inference

Background:

• Randomized controlled trials (RCTs) are crucial for evaluating treatment efficacy and safety.
• International Council on Harmonization (ICH)-E9 (R1) emphasizes considering intercurrent events (ICEs) when defining treatment effect estimands.
• Principal stratum is a key strategy for handling ICEs, but existing methods for AdACE estimation are computationally complex.

Purpose of the Study:

• To implement and evaluate a novel methodology for estimating the AdACE using multiple imputation (MI) and bootstrapping.
• To provide a more accessible computational approach for AdACE estimation compared to existing high-dimensional integration methods.
• To assess the performance of MI-based AdACE estimators through simulation studies.

Main Methods:

• Utilized multiple imputation (MI) to handle missing data and intercurrent events.
• Employed bootstrapping techniques to construct confidence intervals (CIs) for the AdACE.
• Conducted a simulation study to evaluate the consistency and coverage probabilities of the proposed estimators.

Main Results:

• The MI-based AdACE estimators demonstrated consistency.
• Confidence intervals constructed through bootstrapping achieved nominal coverage probabilities.
• The proposed method was successfully applied to a real-world clinical trial dataset.

Conclusions:

• Multiple imputation combined with bootstrapping offers a practical and statistically sound approach for estimating the adherent average causal effect (AdACE).
• This method simplifies the estimation of treatment effects for adherent populations, addressing limitations of previous complex computational techniques.
• The findings support the use of this MI-based approach in clinical trial analysis, particularly for understanding treatment effects in the presence of intercurrent events.