A Revised Framework to Evaluate the Consistency Assumption Globally in a Network of Interventions

Area of Science:

• Network meta-analysis
• Statistical modeling
• Comparative effectiveness research

Background:

• The unrelated mean effects (UME) model is used for global consistency evaluation in intervention networks.
• The original UME model has limitations in handling multiarm trials and omitted comparisons.

Purpose of the Study:

• To refine the UME model for accurate accommodation of multiarm trials.
• To ensure all observed comparisons are estimated in complex intervention networks.
• To improve the detection of global inconsistency in network meta-analysis.

Main Methods:

• Proposed a refined UME model addressing limitations of the original UME model.
• Utilized scatterplots of posterior mean deviance contributions and Bland-Altman plots.
• Applied both refined and original UME models to two networks with multiarm trials.

Main Results:

• The original UME model omitted over 20% of observed comparisons in tested networks.
• Complementary plots and model fit measures indicated potential inconsistency in both networks.
• The refined UME model successfully accommodated multiarm trials and visualized all evidence.

Conclusions:

• The refined UME model properly incorporates multiarm trials and visualizes all evidence in complex networks.
• Complementary plots aid in drawing informed conclusions about global inconsistency.
• The refined model enhances the analysis of intervention networks with multiarm studies.