Visualizing Network Meta-Analyses is not Trivial - A Novel Take on the Network Diagram
Author(s)
Popoff E, Powell L, Johnston K
Broadstreet Health Economics & Outcomes Research, Vancouver, BC, Canada
Presentation Documents
OBJECTIVES: Traditionally, feasibility assessments and results for network meta-analyses (NMAs) have been characterized using network diagrams (often repeated per outcome to reflect differential outcome reporting), trial and patient characteristic summaries, pairwise outcome results, and treatment rankings. This can lead to challenges in interpreting and presenting an NMA, given the need for numerous distinct summary figures to convey the underlying methods and results. There is a need for visualization frameworks that concisely describe several network features simultaneously.
METHODS: Hypothetical sets of trials were used to construct these visualizations. Instead of solid circles used to represent nodes, the node was reimagined where the treatment name comprises the upper half circle of the node and wedges containing a single SUCRA ranking for each outcome comprise the bottom half circles. Traditional trial labels were expanded, allowing for averaged patient characteristics to be displayed underneath. Typically, for any given network of T treatments and O outcomes, there can be up to T*(T – 1)*O/2 unique comparisons. SUCRA wedges offer a simple approach of summarizing this information.
RESULTS: These visualizations provide a concise but comprehensive summary of relevant study features and network structure and a summary of assessed outcomes, suitable for executive summaries and presentations. SUCRAs serve a dual purpose of both treatment rankings and outcome reporting, while patient characteristics express the amount of heterogeneity present in the network. Limitations of this approach include difficulty visualizing non-star-shaped networks where trial labels need to be repeated, and the omission of measures of dispersion.
CONCLUSIONS: This visualization technique presents a novel way to communicate the inputs and outputs of NMAs. This technique works best for star-shaped networks and adaptations may be required for other geometries or applications outside NMAs.
Conference/Value in Health Info
Value in Health, Volume 25, Issue 6, S1 (June 2022)
Code
MSR5
Topic
Methodological & Statistical Research, Study Approaches
Topic Subcategory
Decision Modeling & Simulation, Literature Review & Synthesis, Meta-Analysis & Indirect Comparisons
Disease
No Additional Disease & Conditions/Specialized Treatment Areas