A SIMULATION STUDY OF AREA-UNDER-THE CURVE MATCHING-ADJUSTED INDIRECT COMPARISON FOR ROBUST LONGITUDINAL TREATMENT EVALUATION
Author(s)
David Bin-Chia Wu, PhD1, Lin YuJr, MSc2, Luming Shi, MSc3.
1Market Access (Asia Pacific), Johnson and Johnson Innovative Medicine, Singapore, Singapore, 2Digital Medical Center, Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan, Taipei, Taiwan, 3Cochrane Singapore, Singapore Clinical Research Institute, Singapore, Singapore.
1Market Access (Asia Pacific), Johnson and Johnson Innovative Medicine, Singapore, Singapore, 2Digital Medical Center, Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan, Taipei, Taiwan, 3Cochrane Singapore, Singapore Clinical Research Institute, Singapore, Singapore.
OBJECTIVES: Matching-adjusted indirect comparison (MAIC) is increasingly used to support health technology assessment (HTA) and payer decision-making in the absence of head-to-head evidence. For continuous longitudinal outcomes, conventional analyses typically focus on a single assessment timepoint, potentially underrepresenting cumulative patient benefit when treatment effects evolve over time. This study evaluated the robustness and decision relevance of an area-under-the-curve (AUC)-based MAIC approach to inform supportive sensitivity analysis for longitudinal comparative effectiveness assessment.
METHODS: A simulation study was conducted within an anchored frequentist MAIC framework comparing two hypothetical treatments through a common comparator. Individual patient data were generated for one trial and aggregate data for the comparator trial. Baseline imbalance in prognostic factors and treatment-effect modifiers was introduced and adjusted using standard MAIC weighting. Continuous disease severity scores were simulated across six follow-up visits under stable, delayed-onset, and waning treatment-effect patterns with varying dropout. Performance across 1,000 simulation replicates was evaluated using bias, root mean squared error (RMSE), confidence interval coverage.
RESULTS: The frequentist AUC-based MAIC demonstrated robust performance across clinically relevant scenarios where mean bias remained below 10% of the true treatment effect in most scenarios, including delayed-onset and waning-effect settings. RMSE remained within 8% of the true effect under moderate baseline imbalance and up to 20% dropout. Confidence interval coverage ranged from 71% to 80% across most scenarios. The performance of the estimator remained stable despite varying treatment-effect trajectories, demonstrating good operating characteristics across the simulated scenarios.
CONCLUSIONS: The frequentist AUC-based MAIC provides a practical and complimentary framework for evaluating longitudinal treatment effects by summarizing cumulative patient benefit rather than isolated timepoint differences. Despite the moderate variability of coverage under a few scenarios, the method maintained acceptable estimation accuracy and operational feasibility, supporting its application in HTA evidence generation when treatment effects vary over time and direct comparative evidence is unavailable for reimbursement decisions.
METHODS: A simulation study was conducted within an anchored frequentist MAIC framework comparing two hypothetical treatments through a common comparator. Individual patient data were generated for one trial and aggregate data for the comparator trial. Baseline imbalance in prognostic factors and treatment-effect modifiers was introduced and adjusted using standard MAIC weighting. Continuous disease severity scores were simulated across six follow-up visits under stable, delayed-onset, and waning treatment-effect patterns with varying dropout. Performance across 1,000 simulation replicates was evaluated using bias, root mean squared error (RMSE), confidence interval coverage.
RESULTS: The frequentist AUC-based MAIC demonstrated robust performance across clinically relevant scenarios where mean bias remained below 10% of the true treatment effect in most scenarios, including delayed-onset and waning-effect settings. RMSE remained within 8% of the true effect under moderate baseline imbalance and up to 20% dropout. Confidence interval coverage ranged from 71% to 80% across most scenarios. The performance of the estimator remained stable despite varying treatment-effect trajectories, demonstrating good operating characteristics across the simulated scenarios.
CONCLUSIONS: The frequentist AUC-based MAIC provides a practical and complimentary framework for evaluating longitudinal treatment effects by summarizing cumulative patient benefit rather than isolated timepoint differences. Despite the moderate variability of coverage under a few scenarios, the method maintained acceptable estimation accuracy and operational feasibility, supporting its application in HTA evidence generation when treatment effects vary over time and direct comparative evidence is unavailable for reimbursement decisions.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
SA81
Topic
Methodological & Statistical Research, Study Approaches
Topic Subcategory
Meta-Analysis & Indirect Comparisons
Disease
No Additional Disease & Conditions/Specialized Treatment Areas