The design and analysis of stochastic costeffectivenessstudies for the evaluation of health careinterventions

Many clinical trials are in progress which involve the collection of patient-level data on boththe health outcome and resource use consequences of the health care interventions underevaluation. The overall aim of many such evaluations will be to undertake a cost-effectivenessanalysis, which will often result in a cost-effectiveness ratio summarising the value for moneyof the intervention in question. In this paper, we explore the issues surrounding the design andanalysis of such studies. At the design stage of an analysis, we propose an improved samplesize formula for cost-effectiveness analysis that allows for covariance between cost and effectdifferences. This approach is based on the 'net benefits' approach to the analysis of uncertaintyin cost-effectiveness analysis. At the analysis stage of an evaluation, we explore thedifferences and similarities of the 'net benefit' approach to analysing cost-effectivenessinformation and the traditional approach based on cost-effectiveness ratios. Despite theapparent differences, we show that the two approaches are exactly equivalent when it comesto estimating the probability that the intervention is cost-effective under alternative values ofthe ceiling cost-effectiveness ratio appropriate for decision-making purposes.