crossnma R Package
The crossnma R package helps researchers combine different types of study data—like results from randomised trials and observational studies, as well as individual and summary-level data—to compare multiple treatments using advanced statistical methods.
At a glance
Use when
Combining IPD and AD from multiple study designs; when adjusting for effect modifiers is important; when evidence networks include non-randomised studies; for robust synthesis in sparse networks
Avoid when
Only aggregate data from RCTs are available and no IPD or NRS are present; when users lack access to JAGS or experience with Bayesian modelling; when computational resources are limited
Inputs
Aggregate data and/or individual participant data from randomised and non-randomised studies, study design information, treatment arms, outcome measures, covariates for adjustment
Outputs
Posterior distributions of treatment effects, estimates of heterogeneity and inconsistency, adjusted treatment comparisons, network meta-regression results, convergence diagnostics
How it works
The crossnma R package implements Bayesian three-level hierarchical models for network meta-analysis and network meta-regression, integrating aggregate data and individual participant data from both randomised controlled trials and non-randomised studies. It interfaces with JAGS for Markov Chain Monte Carlo computation and includes tools for data formatting, model generation, convergence assessment, and result summarization.
- Project
- HTx
- Funding
- Horizon 2020
- Project status
- Completed 2024
- HTA domains
- Clinical Effectiveness
- Categories
- Evidence SynthesisRWE
- Technology
- Non-specific
- Assumptions
- Treatment effects can be modelled within a Bayesian hierarchical framework; data formats can be harmonized; exchangeability across studies holds to some degree; user provides correctly structured input data
- Strengths
- Enables synthesis of diverse evidence types; allows adjustment for covariates using IPD; accounts for risk of bias across designs; provides full Bayesian inference with uncertainty quantification
- Limitations
- Requires familiarity with Bayesian statistics and JAGS; may be computationally intensive; model convergence needs careful assessment; limited support for complex data structures without user adaptation
- Also known as
- crossnma, crossnma package
Questions this answers
- › How can I combine individual patient data and summary data in a treatment comparison?
- › How can I include both randomised trials and observational studies in my network meta-analysis?
- › How do I adjust for patient characteristics when comparing treatments across studies?
- › Can I account for differences in study quality or risk of bias in my analysis?
- › How do I perform network meta-regression with mixed data formats?
- › What is the best way to model treatment effects when studies report different outcome formats?
References & sources
- paperDOI: 10.1186/s12874-023-02130-0 ↗
- deliverableec.europa.eu ↗
- deliverableec.europa.eu ↗
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.