Meta-Analytic-Predictive (MAP) prior
The MAP prior method uses data from past clinical trials to create a starting point for analyzing new trial results, especially when looking at control groups. It helps determine how much weight to give historical data based on how similar past trials are to the current one.
At a glance
Use when
Designing or analyzing nonconfirmatory trials (e.g., Phase IV, proof-of-concept) with available historical control data; when sample sizes are small and borrowing strength is beneficial
Avoid when
Historical trials are clinically or methodologically dissimilar; when confirmatory trial standards require strict independence of data; when between-trial heterogeneity is high and poorly characterized
Inputs
Historical control data from previous trials (e.g., sample sizes, outcome estimates, between-trial heterogeneity)
Outputs
A prior distribution for the control parameter in a new trial, characterized by its effective sample size and uncertainty
How it works
The MAP prior is a Bayesian method that synthesizes historical control data across multiple trials using a meta-analytic-predictive model. It models between-trial heterogeneity of control parameters via random effects, with the between-trial variance determining the effective sample size of the prior. The method bounds the influence of historical data by a prior maximum sample size, defined as the ratio of within- to between-trial variance, ensuring robustness against overuse of historical information.
- HTA domains
- Clinical Effectiveness, Safety
- Assumptions
- Control parameters across trials are exchangeable; between-trial heterogeneity can be modeled as random effects; similarity of patient populations and trial designs across studies
- Strengths
- Efficiently incorporates relevant historical data; quantifies uncertainty via effective sample size; improves trial efficiency especially in early-phase or nonconfirmatory settings
- Limitations
- Relies on the assumption of exchangeability; difficulty in quantifying between-trial variance; conclusions may be sensitive to prior assumptions
- Also known as
- MAP prior, meta-analytic predictive approach
Questions this answers
References & sources
Similar by meaning
- Bayesian random-effects meta-analysis with empirical heterogeneity priors for HTA
- Three-stage network meta-regression for heterogeneous treatment effects
- Threshold Analysis for NMA
- Two-stage network meta-regression for heterogeneous treatment effects
- Parametric G-computation for indirect treatment comparison
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.