Improved GSimp: A Flexible Missing Value Imputation Method to Support Regulatory Bioequivalence Assessment

Area of Science:

• Pharmacokinetics and Pharmacodynamics
• Biostatistics
• Clinical Trial Design

Background:

• Missing values are frequent in in vivo bioequivalence (BE) studies, complicating BE assessment.
• Existing imputation methods often cater to specific missing data types (e.g., Missing Not at Random - MNAR), limiting their applicability.
• The original Gibbs sampler-based imputation (GSimp) showed promise for MNAR data but lacked flexibility for mixed missing data types.

Purpose of the Study:

• To introduce an improved Gibbs sampler-based imputation method (Improved GSimp) for handling mixed missing data types in bioequivalence studies.
• To enhance imputation accuracy and flexibility for bioequivalence assessment in the presence of missing values.
• To provide a robust statistical tool for bioequivalence studies with complex missing data patterns.

Main Methods:

• Development of an "Improved GSimp" algorithm designed to accommodate various missing data types (e.g., MNAR, Missing Completely at Random - MCAR).
• Simulation studies were conducted to mimic diverse missing data scenarios, including mixed types and varying proportions of missing values.
• Performance comparison using Normalized Root Mean Square Error (NRMSE) against existing methods like the original GSimp and "half of minimal value" imputation.

Main Results:

• The Improved GSimp demonstrated superior imputation accuracy across all simulated scenarios compared to other evaluated methods.
• The method effectively handled mixed types of missing data, a common challenge in bioequivalence studies.
• Simulation results confirmed the robustness and enhanced performance of the Improved GSimp.

Conclusions:

• The Improved GSimp offers a more flexible and accurate solution for imputing missing values in bioequivalence studies with mixed missing data types.
• This advanced imputation technique can improve the reliability and validity of bioequivalence assessments.
• The findings support the adoption of Improved GSimp for bioequivalence studies facing missing data challenges.