Empirical estimation of disutilities and decision thresholds for composite endpoints

The evaluation of health benefits and harms of an intervention with GRADE Evidence to Decision (EtD) frameworks includes judgments if the effects are “trivial”, “small”, “moderate” or “large”. Such judgments ideally require the a priori establishment of decision thresholds (DTs), whose empirical derivation for single outcomes has been previously described. In this article, we provide a methodological approach to estimate DTs for composite endpoints based on disutilities. We generated an approach that involves the computation of pooled disutilities, in which the disutility of each outcome comprised in the composite endpoint is weighted by the respective relative frequency. We also applied a modelling approach based on probability distributions to account for uncertainty in the estimates. We present a practical example of the determination of DTs associated with the development of at least one adverse event following treatment with intranasal medications for rhinitis that we used in the Allergic Rhinitis in Asthma (ARIA) guidelines. We provide the methodological steps of a modelling-based approach to compute pooled disutilities and, as a result, DTs for composite endpoints. We have developed a webtool (https://compositedt.med.up.pt/) that allows for a simple implementation of the proposed approach. Applying our approach to a practical example, we concluded that rhinitis nasal medications, compared to placebo, were associated with a trivial harm from adverse events. We propose an approach for estimating DTs for composite endpoints, which may be particularly valuable whenever composite endpoints are used for clinical research, clinical practice and decision-making. •An approach has been developed to compute DTs for composite endpoints.•The approach is based on weighted disutilities and allows to consider uncertainty.•We provide detailed stepwise guidance and a webtool to implement this approach.