Random intercept hierarchical linear model for multi-regional clinical trials

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Statistical Modeling

Background:

• Multi-regional clinical trials (MRCTs) require statistical models that account for regional variations.
• Hierarchical linear models (HLMs) are suitable for MRCTs, acknowledging shared patient factors within regions.

Purpose of the Study:

• To investigate the statistical properties of HLMs with random intercepts in MRCTs.
• To compare the random intercept HLM with a random slope HLM regarding their performance and applicability.

Main Methods:

• Statistical analysis of hierarchical linear models, specifically focusing on random intercept and random slope variations.
• Simulation studies were conducted to evaluate model performance and derive selection criteria.

Main Results:

• The random intercept HLM demonstrates an advantage in controlling type I error rates for overall treatment effects when regional differences are negligible.
• Criteria for selecting between random intercept and random slope HLMs were established based on the magnitude of regional treatment effect variation.

Conclusions:

• The random intercept HLM is a valuable tool for MRCTs, particularly for maintaining statistical integrity when regional treatment effects are similar.
• The study provides practical guidance for choosing appropriate HLM structures in MRCTs based on simulated evidence.