Estimating and visualising the trade-off between benefits and harms on multiple clinical outcomes in network meta-analysis

Abstract

Background: The relative treatment effects estimated from network meta-analysis can be employed to rank treatments from the most preferable to the least preferable option. These treatment hierarchies are typically based on ranking metrics calculated from a single outcome. Some approaches have been proposed in the literature to account for multiple outcomes and individual preferences, such as the coverage area inside a spie chart, that, however, does not account for a trade-off between efficacy and safety outcomes. We present the net-benefit standardised area within a spie chart, SAWIS to explore the changes in treatment performance with different trade-offs between benefits and harms, according to a particular set of preferences. Methods: We combine the standardised areas within spie charts for efficacy and safety/acceptability outcomes with a value λ specifying the trade-off between benefits and harms. We derive absolute probabilities and convert outcomes on a scale between 0 and 1 for inclusion in the spie chart. Results: We illustrate how the treatments in three published network meta-analyses perform as the trade-off λ varies. The decrease of the SAWIS quantity appears more pronounced for some drugs, e.g. haloperidol. Changes in treatment performance seem more frequent when SUCRA is employed as outcome measures in the spie charts. Conclusions: SAWIS should not be interpreted as a ranking metric but it is a simple approach that could help identify which treatment is preferable when multiple outcomes are of interest and trading-off between benefits and harms is important.