A hybrid return to baseline imputation method to incorporate MAR and MNAR dropout missingness

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Longitudinal Data Analysis

Background:

• Missing data are common in longitudinal clinical trials due to intercurrent events (ICEs).
• Standard imputation methods struggle when missingness involves both missing at random (MAR) and missing not at random (MNAR) assumptions.
• Traditional Return to Baseline (RTB) imputation is limited to MNAR scenarios and cannot address mixed missingness patterns.

Purpose of the Study:

• To propose a novel hybrid imputation strategy for handling missing data in longitudinal clinical trials.
• To develop a method capable of simultaneously addressing MAR and MNAR dropout missingness.
• To enhance the efficiency and applicability of imputation techniques in complex clinical trial data.

Main Methods:

• Development of a hybrid Return to Baseline (RTB) imputation strategy.
• Application of a standard multiple imputation approach within the hybrid framework.
• Derivation of an analytic likelihood-based method to optimize imputation efficiency.

Main Results:

• The proposed hybrid RTB imputation successfully handles mixed MAR and MNAR missing data patterns.
• The analytic likelihood approach demonstrates improved efficiency compared to standard methods.
• The strategy provides a robust solution for previously intractable missing data scenarios in clinical trials.

Conclusions:

• The hybrid RTB imputation strategy offers a flexible and effective solution for missing data in longitudinal clinical trials.
• This method advances the handling of complex missingness, including dropouts due to events like COVID-19.
• The developed techniques improve the reliability and accuracy of statistical analyses in the presence of diverse missing data mechanisms.