Multilevel Network Meta-Regression (ML-NMR)
ML-NMR is a statistical method used to compare treatments when studies involve different patient populations. It uses detailed patient-level data from some studies to adjust for differences in populations, making indirect comparisons more accurate.
At a glance
Use when
Comparing treatments across different populations using mixed aggregate and individual patient data; conducting network meta-analysis with potential effect modifier imbalance; when extrapolation to a target population is needed
Avoid when
Individual patient data are unavailable; key effect modifiers are unmeasured or poorly measured; when population overlap is minimal
Inputs
Aggregate data and individual patient data (IPD), including effect modifiers and treatment outcomes from multiple studies
Outputs
Adjusted treatment effect estimates in a target population, applicable in network meta-analysis
How it works
ML-NMR is a population adjustment method for anchored indirect comparisons in network meta-analysis that uses individual patient data (IPD) from one or more studies to account for imbalances in effect modifiers across populations. It employs a multilevel regression framework to model treatment effects and effect modification, enabling estimation in any target population and extension to larger treatment networks.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation, Patient and Social Aspects
- Assumptions
- All relevant effect modifiers are measured and included; shared effect modifier assumption holds; sufficient overlap between study populations; correct model specification
- Strengths
- Reduces bias when effect modifiers are balanced; extends to large treatment networks; produces estimates in any target population; robust when all effect modifiers are included
- Limitations
- Biased if important effect modifiers are missing; requires individual patient data; performance depends on between-study population overlap and model correctness
- Also known as
- ML-NMR
Questions this answers
- › How can we compare treatments when studies have different patient populations?
- › What methods reduce bias in indirect comparisons due to effect modifier imbalance?
- › How can individual patient data improve network meta-analysis?
- › When should ML-NMR be preferred over MAIC or STC?
- › Can population adjustment methods be applied to large treatment networks?
- › How robust is ML-NMR when assumptions are violated?
References & sources
Related methods
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.