Graphical tool for locating inconsistency in network meta-analyses
This method helps identify where inconsistencies occur in network meta-analyses by showing which direct comparisons influence overall results and highlighting areas where direct and indirect evidence disagree.
At a glance
Use when
Conducting network meta-analyses where inconsistency between direct and indirect evidence is suspected or needs to be explored.
Avoid when
When only random-effects models are acceptable, or when the network is too large or sparse for meaningful visualization.
Inputs
Direct comparison effect estimates from clinical trials, network structure of treatments and comparisons, results from fixed-effects network meta-analysis models.
Outputs
Net heat plot visualizing contributions and inconsistency; heat matrix with changes in direct-indirect agreement; clustering results identifying hot spots of inconsistency.
How it works
The method introduces the net heat plot, a graphical tool based on fixed-effects models that visualizes the contribution of each direct comparison to network estimates using regression diagnostics. It displays the impact of relaxing consistency assumptions for each comparison via heat colors and applies clustering to a heat matrix to detect hot spots of inconsistency, enabling identification of influential comparisons and their effects on network estimates.
- HTA domains
- Clinical Effectiveness
- Assumptions
- The method assumes a fixed-effects model framework and that inconsistency can be localized to specific direct comparisons; it also assumes that removing consistency constraints for one comparison can reveal its impact on others.
- Strengths
- Provides a transparent, visual way to explore drivers of inconsistency; identifies both sources and effects of inconsistency; supports further subject-matter investigation.
- Limitations
- Limited to fixed-effects models; interpretation may become complex in large networks; relies on proper specification of the network structure.
- Also known as
- net heat plot
Questions this answers
- › Which direct comparisons are driving inconsistency in the network meta-analysis?
- › How do individual direct comparisons influence network estimates?
- › Where are the hot spots of inconsistency located within the network?
- › What happens to the agreement between direct and indirect evidence when consistency is relaxed for a specific comparison?
- › Which network estimates are affected by a particular inconsistent direct comparison?
- › How are different effect estimates interrelated in the presence of inconsistency?
References & sources
Related methods
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.