cLRT-Mod: An efficient methodology for pharmacometric model-based analysis of longitudinal phase II dose finding...

Area of Science:

• Pharmacometrics
• Clinical Trial Design
• Statistical Modeling

Background:

• Longitudinal data analysis can improve drug effect detection in phase II dose-finding trials compared to end-of-treatment analyses.
• Existing methods may lack power in scenarios with small sample sizes or weak drug effects.

Purpose of the Study:

• To propose and evaluate cLRT-Mod, a pharmacometric adaptation of MCP-Mod for phase II dose-finding.
• To enhance the detection of dose-response signals and identify optimal doses for confirmatory phases.
• To account for model structure uncertainty in dose-finding trials.

Main Methods:

• Development of cLRT-Mod, integrating nonlinear mixed effect models with MCP-Mod principles.
• Extensive clinical trial simulations under various scenarios (e.g., small sample size, weak effect).
• Comparison of cLRT-Mod with existing methods like MCP-Mod and end-of-treatment multiple contrast tests.

Main Results:

• cLRT-Mod with longitudinal data demonstrated increased statistical power compared to end-of-treatment tests, especially for weak drug effects and small sample sizes.
• The method maintained pre-specified model characteristics and nominal type I error rates.
• Model averaging within cLRT-Mod improved coverage probability and avoided confidence interval underestimation.

Conclusions:

• cLRT-Mod offers a powerful and robust approach for phase II dose-finding trials utilizing longitudinal data.
• The methodology effectively detects dose-response signals and selects appropriate doses while managing model uncertainty.
• Application to a real trial confirmed its practical utility in dose-finding studies.