multinma R package
The multinma R package allows for more accurate comparisons of treatments by accounting for differences in patient populations across studies. It uses individual patient data from some studies and combines it with aggregate data from others to adjust for imbalances in important patient characteristics, providing more reliable results for decision making.
At a glance
Use when
Conducting network meta-analysis with imbalanced effect modifiers across studies; when individual patient data are available for some treatments; when estimates are needed for a specific target population
Avoid when
Individual patient data are unavailable for any study; when effect modifier distributions are similar across populations; when computational resources are limited
Inputs
Aggregate trial data, individual patient data from one or more studies, covariate distributions, correlation structures (via copulas), target population characteristics
Outputs
Population-adjusted treatment effect estimates, uncertainty intervals, model fit statistics, predictions in specified target populations
How it works
multinma implements multilevel network meta-regression (ML-NMR), an extension of standard network meta-analysis that integrates individual-level regression models with aggregate data through numerical integration (quasi-Monte Carlo). It accounts for covariate distributions and correlations via copulas, enabling population-adjusted treatment effect estimation in any target population. The method avoids aggregation bias and improves model fit and precision compared to standard random-effects NMA.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation, Patient and Social Aspects
- Assumptions
- Covariate-outcome relationships are correctly specified; individual patient data are representative; copula models adequately capture covariate dependencies
- Strengths
- Enables population-adjusted indirect comparisons in network meta-analysis; reduces uncertainty by explaining heterogeneity; allows prediction in any target population; avoids aggregation bias; handles correlated covariates via copulas
- Limitations
- Requires individual patient data from at least some studies; computational complexity due to numerical integration; results depend on correct specification of covariate effects and correlations
- Also known as
- ML-NMR, multinma R package
Questions this answers
- › How can treatment effects be compared when patient populations differ across studies?
- › How can individual patient data be combined with aggregate data in network meta-analysis?
- › How can effect modification be accounted for in indirect treatment comparisons?
- › How can treatment effects be predicted in a specific target population?
- › How can uncertainty be reduced in network meta-analysis by explaining heterogeneity?
- › How can covariate correlations be modeled in population-adjusted indirect comparisons?
References & sources
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.