Enhancing phase I dose-finding trials design through dynamic borrowing information and handling late-onset toxicity

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Regulatory agencies increasingly consider historical controls to improve clinical trial efficiency.
• Phase I dose-finding trials face challenges with efficiency and delayed toxicity.
• Novel anti-cancer drugs often exhibit delayed toxicity, complicating dose-escalation.

Purpose of the Study:

• To enhance the statistical operating characteristics of Phase I dose-finding trials.
• To propose a novel model-assisted design method, MEM-Keyboard, for dose-finding.
• To extend the method to handle late-onset toxicity using historical data (MEM-TITE-Keyboard).

Main Methods:

• The proposed MEM-Keyboard design utilizes multisource exchangeability models (MEMs).
• MEMs allow dynamic borrowing of information from multiple supplemental data sources, including historical trial data.
• The MEM-TITE-Keyboard extends this by incorporating historical data to manage late-onset toxicity.

Main Results:

• Simulation studies demonstrate improved probability of correctly selecting the maximum tolerated dose (MTD).
• The methods offer an acceptable level of risk compared to designs ignoring information borrowing and late toxicity.
• Enhanced efficiency in early clinical trials is indicated.

Conclusions:

• MEM-Keyboard and MEM-TITE-Keyboard are practical tools for identifying MTD.
• These methods can accelerate drug development by improving early clinical trial design.
• The use of historical data in dose-finding trials is validated.