Three-stage network meta-regression for heterogeneous treatment effects
This method helps predict how different patients respond to different treatments by combining data from randomized and non-randomized studies. It uses patient characteristics to explain why treatment effects vary and provides personalized predictions.
At a glance
Use when
Evaluating treatments with variable patient responses; when individual participant data and real-world evidence are available; for personalized medicine applications in HTA.
Avoid when
When only aggregate data from homogeneous populations are available; if prognostic factors are poorly measured or missing; when baseline risk is not expected to modify treatment effects.
Inputs
Aggregate data (AD) and individual participant data (IPD) from randomized and non-randomized studies; patient-level characteristics; baseline outcome risk predictors.
Outputs
Personalized treatment effect estimates across patient subgroups; predicted health outcomes under different treatments; heterogeneous treatment effects adjusted for baseline risk.
How it works
A three-stage modeling approach that (1) develops a prognostic model for baseline risk using large cohort studies (non-randomized IPD), (2) recalibrates the prognostic model using randomized trial data, and (3) incorporates the predicted baseline risk as an effect modifier in a network meta-regression model combining aggregate data (AD) and individual participant data (IPD) from both randomized and non-randomized studies to estimate heterogeneous treatment effects.
- Project
- HTx
- Funding
- Horizon 2020
- Project status
- Completed 2024
- HTA domains
- Clinical Effectiveness
- Technology
- Non-specific
- Assumptions
- Baseline risk modifies treatment effect; prognostic model from cohort data is transportable to trial populations; linearity or specified functional form between baseline risk and treatment effect.
- Strengths
- Integrates diverse data sources (AD, IPD, randomized, non-randomized); enables personalized predictions; improves precision in estimating heterogeneous treatment effects; leverages real-world evidence.
- Limitations
- Dependent on quality and availability of IPD and cohort data; assumes correct specification of prognostic model; potential bias if cohort and trial populations differ systematically.
- Also known as
- three-stage network meta-regression, 3-stage NMR, network meta-regression with prognostic modeling
Questions this answers
- › How do treatment effects differ across patients with different characteristics?
- › Which patients benefit most from a specific treatment?
- › Can we combine randomized and non-randomized data to improve treatment effect predictions?
- › How does baseline risk of disease progression modify treatment effectiveness?
- › What is the best treatment option for a patient with a specific risk profile?
- › How can we use real-world data alongside clinical trials for personalized predictions?
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.