Target Aggregate Data Adjustment Method (TADA): A Solution for Performing Transportability Analysis with Only Aggregate-Level Data in the Target Population
Author(s)
Quang Vuong, MSc1, Richard Yan, MSc2, Rebecca Metcalfe, BA, MA, PhD3, Jay J. Park, PhD4;
1Core Clinical Sciences, Vancouver, BC, Canada, 2Simon Fraser University, Department of Statistical and Actuarial Science, Vancouver, BC, Canada, 3Centre for Advancing Health Outcomes, Vancouver, BC, Canada, 4McMaster University, Department of Health Research Methods, Evidence, and Impact, Hamilton, ON, Canada
1Core Clinical Sciences, Vancouver, BC, Canada, 2Simon Fraser University, Department of Statistical and Actuarial Science, Vancouver, BC, Canada, 3Centre for Advancing Health Outcomes, Vancouver, BC, Canada, 4McMaster University, Department of Health Research Methods, Evidence, and Impact, Hamilton, ON, Canada
OBJECTIVES: Transportability analysis is a causal inference framework used to evaluate the external validity of randomized clinical trials or observational studies. Most existing transportability analysis methods require individual patient-level data (IPD) for both the source and the target population, narrowing its applicability when only target aggregate-level data (AgD) is available. Concurrently, accounting for censoring is essential to reduce bias in survival data; yet AgD-based transportability methods in the presence of censoring remain underexplored. Here in this study, we propose a weighting method named “Target Aggregate Data Adjustment” (TADA) to simultaneously address both these challenges.
METHODS: TADA uses a two-stage weighting scheme where the final weights are the product of the inverse probability of censoring weights and participation weights derived using the method of moments; observations in the source study are then weighed according to this scheme to obtain transported treatment effects. To evaluate this method’s ability to reduce bias and support valid inference, we conducted a simulation study of TADA’s performance under different scenarios with varying rates of censoring.
RESULTS: In general, TADA with censoring adjustments demonstrates superior performance in terms of bias and coverage compared to the unadjusted approach across various censoring scenarios. The benefits of TADA increased as the difference in censoring rates between treatment and control groups increased. Simulations showed that at lower overall censoring rates of 20% and 30%, the advantages of TADA are particularly prominent. For instance, under a 20% censoring rate, the adjusted bias is 0.037 versus an unadjusted bias of 0.128. However, TADA’s advantages were less pronounced when overall censoring rates reached 50%.
CONCLUSIONS: In the presence of informative censoring, accounting for censoring via TADA can greatly improve control of bias and estimator coverage compared to the unadjusted approach. However, when censoring rates exceeded 50%, the performance of TADA was less favourable.
METHODS: TADA uses a two-stage weighting scheme where the final weights are the product of the inverse probability of censoring weights and participation weights derived using the method of moments; observations in the source study are then weighed according to this scheme to obtain transported treatment effects. To evaluate this method’s ability to reduce bias and support valid inference, we conducted a simulation study of TADA’s performance under different scenarios with varying rates of censoring.
RESULTS: In general, TADA with censoring adjustments demonstrates superior performance in terms of bias and coverage compared to the unadjusted approach across various censoring scenarios. The benefits of TADA increased as the difference in censoring rates between treatment and control groups increased. Simulations showed that at lower overall censoring rates of 20% and 30%, the advantages of TADA are particularly prominent. For instance, under a 20% censoring rate, the adjusted bias is 0.037 versus an unadjusted bias of 0.128. However, TADA’s advantages were less pronounced when overall censoring rates reached 50%.
CONCLUSIONS: In the presence of informative censoring, accounting for censoring via TADA can greatly improve control of bias and estimator coverage compared to the unadjusted approach. However, when censoring rates exceeded 50%, the performance of TADA was less favourable.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
MSR113
Topic
Methodological & Statistical Research
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference
Disease
No Additional Disease & Conditions/Specialized Treatment Areas