Multiply robust estimators in longitudinal studies with missing data under control-based imputation.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Longitudinal Data Analysis

Background:

• Longitudinal studies frequently encounter missing data, complicating treatment effect estimation.
• Regulatory guidance (ICH E9(R1) addendum) emphasizes defining treatment effects considering intercurrent events.
• Jump-to-reference (J2R) is a key scenario for evaluating treatment effects in the presence of intercurrent events.

Purpose of the Study:

• To develop novel statistical estimators for average treatment effect under the J2R framework.
• To provide robust methods for handling missing data and intercurrent events in longitudinal studies.
• To introduce an efficiently estimated, multiply robust estimator for treatment effect evaluation.

Main Methods:

• Developed a potential outcomes framework for J2R analysis.
• Derived identification formulas utilizing different observed data distributions.
• Proposed a novel estimator based on the efficient influence function for enhanced robustness.

Main Results:

• The proposed estimators provide valid assessment of average treatment effect under J2R.
• The novel estimator demonstrates multiple robustness properties, achieving consistency under various nuisance parameter specifications.
• Simulation studies and an antidepressant trial validated the performance of the new methods.

Conclusions:

• The new estimators offer robust and flexible approaches for analyzing longitudinal data with intercurrent events.
• These methods align with current regulatory expectations for treatment effect estimation.
• The findings contribute to more reliable clinical trial data interpretation and decision-making.