RankÔÇÉHeat Plot: An R Shiny tool for presenting multiple network meta-analysis outcomes

Session Type
Oral presentation
Category
Evidence synthesis innovations and technology
Authors
Veroniki AA1, Tricco A1, Watt J1, Straus S1
1Unity Health Toronto; University of Toronto, Canada
Description

Background: The increased number of outcomes and competing interventions assessed in a network meta-analysis (NMA) increases the difficulty of interpreting results, and hence challenges their use in everyday clinical practice and policy.
Objectives: To facilitate evidence-based decision-making on a clinical topic with multiple outcomes per intervention through a graphical representation of NMA results, the rank-heat plot. We are currently developing the rankheatplot R package and R Shiny app and will disseminate the first version at the conference.
Methods: The rankheatplot R Shiny app can be used in any type of discipline and disease using a NMA of multiple interventions compared in different studies obtained from any type of review (i.e., systematic review, rapid review, overview of reviews). The rank-heat plot was first developed and published in 2016. It has been cited in >120 publications (Google Scholar), and has been used in multiple areas, such as geriatrics, pediatrics, neurology, and oncology. It is already part of the viscomp R package, which presents multiple visualization approaches on component NMA. We developed an interactive web application at https://rankheatplot.com/ to produce the rank-heat plot when the study-level data for each outcome are available, and without performing the NMA analysis outside the tool. The tool can be used for any type of data.
Results: The rank-heat plot allows the fast identification of the most likely best and worst interventions, with respect to their effectiveness and/or safety, in a given outcome. It can also identify interventions that have not been studied for an included outcome. The rankheatplot app currently performs analyses in a frequentist framework importing intervention outcome results from the netmeta R package. In the next version of the tool, we plan to incorporate the option of conducting the analysis in a Bayesian framework, as well as outputs based on clinically important effects (e.g., minimal clinically important difference).
Conclusions: The rank-heat plot summarizes results on all interventions and all outcomes, providing information regarding the ranking of interventions per outcome. The tool can be used to facilitate interpretation and decision-making based on NMA synthesis. No patients were involved in the development of this tool.