Cross-NMA/NMR
Cross-NMA/NMR is a set of advanced statistical models that combine results from different types of studies, including randomized trials and non-randomized studies, as well as individual patient data and aggregate data, to give a more complete picture of how well treatments work.
At a glance
Use when
Comparing multiple interventions using mixed evidence sources; when IPD and AD are available; when both RCTs and real-world studies exist in the evidence base.
Avoid when
Data are sparse or highly heterogeneous without clear adjustment strategies; when computational resources or expertise are limited; when assumptions of exchangeability are clearly violated.
Inputs
Individual patient data (IPD), aggregate data (AD), randomized controlled trials (RCTs), non-randomized studies (NRS), treatment networks, covariates for meta-regression.
Outputs
Joint treatment effect estimates, ranked probabilities of effectiveness, regression coefficients for effect modifiers, model fit statistics, heterogeneity and inconsistency assessments.
How it works
Cross-NMA/NMR is a suite of Bayesian Network Meta-Analysis and Network Meta-Regression models that enable cross-design (RCTs and non-randomized studies) and cross-format (individual patient data and aggregate data) evidence synthesis. It uses a three-level hierarchical model to integrate heterogeneous data sources within a unified Bayesian framework, improving estimation accuracy and network connectivity in comparative effectiveness research.
- Project
- HTx
- Funding
- Horizon 2020
- Project status
- Completed 2024
- HTA domains
- Clinical Effectiveness
- Technology
- Non-specific
- Assumptions
- Exchangeability of evidence across study designs after adjustment; consistency between direct and indirect evidence; proper specification of hierarchical priors; linearity in meta-regression relationships.
- Strengths
- Integrates diverse data types and study designs; improves precision and network coverage; allows for adjustment of confounding in NRS; supports flexible modeling via Bayesian framework.
- Limitations
- Computationally intensive; requires detailed data and statistical expertise; potential for bias if adjustment for design differences is inadequate; sensitivity to prior specification.
- Also known as
- Cross-NMA, Cross-NMR, Bayesian cross-design synthesis
Questions this answers
- › How can we compare multiple treatments when some studies are randomized and others are not?
- › Can we combine individual patient data with summary results in one analysis?
- › How do we account for differences in study design and data format in network meta-analysis?
- › What is the relative effectiveness of treatments when only non-randomized evidence is available for some comparisons?
- › How can we reduce bias when integrating real-world evidence with clinical trial data?
- › Which treatments are most effective across diverse patient populations and study types?
Similar by meaning
Beta record. Based on the original catalogue summary; primary-source enrichment pending.