Methodologies

Comparison-adjusted funnel plots for NMA

How comparison-adjusted funnel plots are generated for network meta-analysis and why sparse or heterogeneous networks limit interpretation.

12 min readUpdated June 15, 2026

Purpose

A comparison-adjusted funnel plot adapts small-study effect visualization to a network where studies belong to different direct comparisons. The goal is to inspect whether small-study patterns remain after accounting for comparison structure.

Coventra generates this plot from the netmeta object when the NMA small-study-effects module is requested and the selected network model supports plotting.

Ordering and direction

The comparison-adjusted funnel depends on treatment ordering and effect direction. Coventra can use ranking order when rankings are available; otherwise the package default order may be used.

Reviewers must verify whether lower or higher values are beneficial and whether the plotted direction matches the manuscript interpretation.

Limitations

  • Sparse networks may not provide meaningful visual evidence.
  • Star-shaped networks have limited information about comparison-specific asymmetry.
  • High heterogeneity, intransitivity, or inconsistent outcome definitions can mimic small-study patterns.
  • A generated plot is a screening aid, not proof of publication bias.

Reporting checklist

  1. Report that the funnel plot was comparison-adjusted.
  2. Report ordering and direction of benefit.
  3. State when the plot was unavailable or uninformative because the network was sparse.
  4. Interpret alongside direct evidence, transitivity, and inconsistency diagnostics.