A random walk rule for phase I clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacometrics

Background:

• Sequential dose allocation in clinical trials requires robust and ethical methods.
• Current methods may lack comprehensive theoretical underpinnings or require extensive simulations.

Purpose of the Study:

• To introduce and analyze a family of random walk rules for dose-finding studies.
• To demonstrate the ability to center dose assignments around a target quantile.
• To provide a theoretical framework that simplifies study design and analysis.

Main Methods:

• Developed a family of random walk algorithms for sequential dose assignment.
• Derived maximum likelihood estimators and their variances under a two-parameter logistic model.
• Compared random walk estimators with nonparametric alternatives.
• Analyzed finite and asymptotic properties of a specific random walk rule using exact theory.

Main Results:

• Random walk rules allow for unimodal centering of dose assignments around a target quantile.
• The proposed methods are simple to implement and possess fully developed distribution theory.
• Exact theory for small sample properties obviates the need for extensive simulations.
• Performance of the proposed rule favorably compares to the continual reassessment method in small samples.

Conclusions:

• Random walk rules offer a theoretically sound and practical approach to dose-finding studies.
• These methods simplify trial design and analysis by providing complete distribution theory.
• The proposed approach is efficient and performs well, especially in small sample scenarios.