OBJECTIVES: Network meta-analyses compare multiple treatments. However, the network meta-analysis requires strong quantitative assumptions of consistency (e.g. OR_AB = OR_AC / OR_BC), across the trials included in the analysis, i.e. there are no systematic differences between trials in factors that are treatment effect modifiers that are not accounted analytically. If we cannot except this assumption in full, it is then unclear how to interpret the results of network meta-analyses. There is a risk that in trying to apply a statistical model requiring strong, and often untestable, quantitative assumptions we fail to make best use of available trial data. Argumentation-based synthesis (https://bit.ly/3a5BwDC) may provide a useful summary of evidence. Pairs of treatments are compared with respect to individual endpoints based on the results of individual studies, pairwise meta-analyses or network meta-analyses. These pairwise “arguments” are then synthesized across the network of trials. The synthesis identifies which individual treatments are consistently favoured across studies and endpoints and where there are inconsistencies or contradictions in the set of trial evidence. The decision maker can then focus on these areas of uncertainty and resolve them by differentially weighting effects on particular endpoints according to individuals’ preferences and applying “meta-rules” that weight individual results corresponding to their perceived reliability. The utility of argumentation-based synthesis was explored in a comparison first line treatments for chronic myeloid treatments.

METHODS: The argumentation based synthesis was implemented using An interactive RSHINY app.

RESULTS: The use of the argumentation-based synthesis was reviewed over a series of hypothetical case studies reflecting variation individual preferences with respect to treatment efficacy, risk of specific adverse events, and treatment burden

CONCLUSIONS: The interactive argument-based synthesis allowed individuals to interrogate the available evidence at the individual study and meta-analytic level whilst applying their own preferences when making treatment decisions.