Missing Data Handling in the Application of Matching-Adjusted Indirect Comparison

Area of Science:

• Health Economics and Outcomes Research
• Biostatistics
• Clinical Trial Methodology

Background:

• Health Technology Assessment (HTA) submissions frequently require indirect treatment comparisons (ITC).
• Population-adjusted indirect comparisons (PAIC), particularly Matching-Adjusted Indirect Comparison (MAIC), are common ITCs using individual patient data (IPD) and aggregate data (AgD) to adjust for covariate distribution differences.
• Existing MAIC methodologies lack clear guidance on handling missing data in outcomes or covariates.

Purpose of the Study:

• To propose novel weighting-based methods for handling missing data within the MAIC framework.
• To integrate these methods seamlessly into the existing MAIC approach for robust treatment effect estimation.
• To evaluate the performance of the proposed methods through extensive simulation studies.

Main Methods:

• Development of weighting strategies incorporating inverse probability of not missing outcomes and adjustments for baseline characteristic differences.
• Application of these weights within the MAIC framework to estimate treatment effects using a weighted difference between IPD and AgD.
• Conducting comprehensive simulation studies to assess the performance and validity of the proposed missing data handling techniques.

Main Results:

• The proposed weighting methods are shown to effectively handle missing data in both outcome variables and covariates during MAIC.
• These methods integrate smoothly into the established MAIC framework, maintaining its core principles.
• Simulation results demonstrate the utility and performance of the developed techniques in producing reliable treatment effect estimates.

Conclusions:

• The proposed weighting-based methods offer a practical solution for addressing missing data challenges in MAIC.
• These advancements improve the accuracy and applicability of population-adjusted indirect comparisons in HTA.
• The study provides valuable methodological guidance for researchers and decision-makers utilizing MAIC with incomplete datasets.