OBJECTIVES: Differences in the quality and design of published trials make it difficult to recognise when the amassing evidence can be considered sufficient for healthcare decision-making. The power of new evidence to shift the result of an existing meta-analysis in to statistical significance is known as ‘assurance’. We propose a flexible Bayesian framework to undertake evidence-based assurance calculations to appropriately design and power a new trial from meta-analyses.

METHODS: Motivated by a recent network meta-analysis evaluating oral anticoagulants for the prevention of stroke in atrial fibrillation, we assessed the power of a new 4-arm trial to shift the result of an existing meta-analysis in to statistical significance for healthcare decision-making. Bayesian Markov Chain Monte Carlo (MCMC) simulation was used to synthesise existing trial data in meta-analyses. Posterior predictive distributions were obtained to predict the effect of a new trial with disparate sample sizes. Simulated event data for a new trial were combined with the original trial results in meta-analyses to calculate assurance.

RESULTS: We developed and successfully implemented a flexible Bayesian framework to perform evidence-based assurance calculations. For the prevention of stroke and systemic embolism, inclusion of new evidence had little power to shift the healthcare decision in to statistical significance.

CONCLUSIONS: Inclusion of a future, large-scale trial evaluating oral anticoagulants on a head-to-head basis had little impact on the healthcare decision. If future healthcare decisions were to rely on meta-analysis results, it would appear that further trials in this area would not be justified. Evidence-based approaches should be considered in the design of future trials for healthcare interventions to avoid resource and research waste.