Network Meta Analysis

Network Meta-Analysis in R

Overview

Network meta-analysis (NMA) methods for comparing multiple treatments simultaneously using direct and indirect evidence. Covers network structure assessment, frequentist and Bayesian NMA approaches, consistency evaluation, treatment rankings, and visualization techniques.

Network Structure and Data Preparation

Pairwise Data Format

library(netmeta)

# Standard pairwise format for contrast-based NMA
pairwise_data = 2.0)
# net_reg 
    mutate(
      Fixed = paste0(round(Effect_Fixed, 2), " (",
                    round(CI_Lower_Fixed, 2), ", ",
                    round(CI_Upper_Fixed, 2), ")"),
      Random = paste0(round(Effect_Random, 2), " (",
                     round(CI_Lower_Random, 2), ", ",
                     round(CI_Upper_Random, 2), ")")
    ) |>
    select(Comparison, Fixed, Random)

  return(summary_df)
}

Key Packages Summary

| Package | Purpose | |---------|---------| | netmeta | Frequentist NMA (contrast-based) | | gemtc | Bayesian NMA with JAGS | | multinma | Bayesian NMA with Stan | | bnma | Bayesian NMA | | pcnetmeta | Patient-centered NMA | | NMAoutlier | Outlier detection in NMA | | nmathresh | Decision thresholds for NMA |

Best Practices

1. Network geometry: Check connectivity before analysis
2. Transitivity: Assess similarity of study populations across comparisons
3. Consistency: Always assess local and global inconsistency
4. Heterogeneity: Report tau, I², and consider prediction intervals
5. Ranking: Present uncertainty (CrI for ranks, rankograms)
6. Sensitivity: Conduct analyses excluding high RoB studies
7. Reporting: Follow PRISMA-NMA extension guidelines
8. Model selection: Compare fixed vs random effects, check model fit

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: choxos
• Source: choxos/BiostatAgent
• License: MIT

Install and usage instructions live in the source repository linked above.