Component heat plot

Pairwise component combination heat map

NMA
A matrix of pairs of components, each cell summarizing the relative effects of interventions that contain both.
NMAProposed

Component heat plot example

Component heat plot for an artificial network of multicomponent interventions (components A to G) for major adverse cardiovascular events. Each cell gives the median relative effect of interventions that include both components, with the number of interventions in parentheses; X marks combinations never observed. Data: viscomp::nmaMACE (artificial).
Family
Component, dose-response, and threshold NMA
Purpose
Explore which components and pairs of components are associated with larger effects, before formal CNMA.
Inputs
A network meta-analysis of multicomponent interventions.
Software
R viscomp::heatcomp()

What it shows

Seitidis and colleagues developed a set of graphics for exploring networks of multicomponent interventions. The component heat plot arranges components in rows and columns. Each cell summarizes (by median or mean) the NMA relative effects of all interventions that contain both components; diagonal cells summarize single components. Color intensity highlights promising pairs, and crossed cells show pairs never studied together.

How to read it

  • Rows and columns: components.
  • Cells: summary of the relative effects of interventions containing both components; parentheses give how many interventions contribute.
  • Color: magnitude of the summary effect; here odds ratios below 1 (green) mean fewer events.
  • X: combination not present in the network.

Interpretation

Component E stands out: interventions containing it have the lowest median odds ratios for major adverse cardiovascular events (0.46 alone, 0.67 with B, 0.70 with A), whereas its combination with C sits near 1. Many pairs were never studied together (X), and most cells rest on one to three interventions. The plot points to combinations worth modeling formally in a component NMA, not to estimates to report.

Pitfalls

  • Summaries of intervention effects are descriptive and confounded by the other components present.
  • Cells based on one intervention are fragile.
  • The display is exploratory; it does not replace a CNMA model.

Code

library(netmeta)
library(viscomp)

# Example network of multicomponent interventions (components A to G plus
# usual care) for major adverse cardiovascular events, shipped with viscomp
data(nmaMACE)

# Heat map of pairs of components: each cell summarizes the relative effects
# of interventions that include both components (diagonal: single components);
# numbers in parentheses count the interventions involved
heatcomp(nmaMACE, median = TRUE, random = TRUE)

References

  • Seitidis G, Tsokani S, Christogiannis C, et al. Graphical tools for visualizing the results of network meta-analysis of multicomponent interventions. Res Synth Methods. 2023;14:382-395. doi:10.1002/jrsm.1617