Applying Simulation-Guided Trial Design Principles to Prospectively Plan Target Trial Emulation of a Cluster Randomized Trial for Vaccine Coverage
Author(s)
Jay JH Park, PhD1, Richard Yan, MSc2, Shomoita Alam, PhD3, Rebecca Metcalfe, BA, MA, PhD4.
1Core Clinical Sciences, Vancouver, BC, Canada, 2Department of Statistical and Actuarial Science, Simon Fraser University, Vancouver, BC, Canada, 3Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, QC, Canada, 4Centre for Advancing Health Outcomes, University of British Columbia, Vancouver, BC, Canada.
1Core Clinical Sciences, Vancouver, BC, Canada, 2Department of Statistical and Actuarial Science, Simon Fraser University, Vancouver, BC, Canada, 3Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, QC, Canada, 4Centre for Advancing Health Outcomes, University of British Columbia, Vancouver, BC, Canada.
OBJECTIVES: Target trial emulation (TTE) applies clinical trial design principles to plan the analysis of observational data. Most TTE applications have aimed to emulate an individually randomized clinical trial (RCT). There has been little application to cluster RCTs which are often used to evaluate public health interventions targeted at a group of individuals (i.e., clusters). When contamination cannot be avoided due to the proximity of clusters making cluster randomization infeasible, the TTE framework may be beneficial. In this study, we used simulation-guided design principles to plan prospective data collection of a pre-post experimental study that would emulate a village-level cluster RCT of mass distribution of nutritional supplements within an immunization program in Niger.
METHODS: Using baseline census-type data on the pentavalent vaccination rate among 12-24 month old children (primary outcome) and cluster-level covariates, simulations were conducted to plan: prospective data collection to ascertain follow-up vaccine coverage data; and the statistical analysis plan. We applied covariate-constrained random selection of villages in our simulations across varying cluster sizes (50 to 126 clusters per arm). Under intracluster correlation coefficient values of 0.22 and 0.33, quasi-binomial logistic regression, beta-regression, and inverse probability of treatment weighting (IPTW) methods were then compared in terms of type I error rate and power.
RESULTS: Both quasi-binomial regression and naïve analyses showed inflated type I error rates. Although IPTW could control the type I error rate, it was shown to be very conservative with low statistical power for most scenarios. In addition to the type I error rate control, beta regression showed adequate statistical power to detect a minimum effect size (relative risk reduction) of 0.375 under the base case of 0.20 control event rate.
CONCLUSIONS: Adopting simulation-guided design principles and the TTE framework can be a valuable planning tool for group-level interventions when randomization is not possible.
METHODS: Using baseline census-type data on the pentavalent vaccination rate among 12-24 month old children (primary outcome) and cluster-level covariates, simulations were conducted to plan: prospective data collection to ascertain follow-up vaccine coverage data; and the statistical analysis plan. We applied covariate-constrained random selection of villages in our simulations across varying cluster sizes (50 to 126 clusters per arm). Under intracluster correlation coefficient values of 0.22 and 0.33, quasi-binomial logistic regression, beta-regression, and inverse probability of treatment weighting (IPTW) methods were then compared in terms of type I error rate and power.
RESULTS: Both quasi-binomial regression and naïve analyses showed inflated type I error rates. Although IPTW could control the type I error rate, it was shown to be very conservative with low statistical power for most scenarios. In addition to the type I error rate control, beta regression showed adequate statistical power to detect a minimum effect size (relative risk reduction) of 0.375 under the base case of 0.20 control event rate.
CONCLUSIONS: Adopting simulation-guided design principles and the TTE framework can be a valuable planning tool for group-level interventions when randomization is not possible.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
MSR31
Topic
Epidemiology & Public Health, Methodological & Statistical Research, Study Approaches
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference
Disease
Nutrition, Pediatrics, Vaccines