Incorporating historical information to improve phase I clinical trials.

Area of Science:

• Clinical Trial Design
• Biostatistics
• Pharmacology

Background:

• Historical data can significantly improve phase I clinical trial efficiency and accelerate drug development.
• Model-based designs like the continuous reassessment method (CRM) readily incorporate historical data using a 'skeleton'.
• Model-assisted designs (e.g., Bayesian optimal interval (BOIN), Keyboard, mTPI) have seen limited development in incorporating historical data, leading to misconceptions.

Purpose of the Study:

• To propose a unified framework for incorporating historical data into model-assisted clinical trial designs.
• To address the misconception that model-assisted designs cannot utilize prior information.
• To maintain the simplicity and pre-tabulated dose escalation rules characteristic of model-assisted designs.

Main Methods:

• Developed a unified framework integrating historical data into model-assisted designs.
• Utilized the established 'skeleton' approach combined with the concept of prior effective sample size.
• Ensured the dose escalation/de-escalation rules remain simple and can be tabulated before trial commencement.

Main Results:

• The proposed framework effectively incorporates prior information into model-assisted designs.
• Simulation studies demonstrate improved operating characteristics for model-assisted designs when using the proposed method.
• The performance of the proposed method is comparable to that of model-based designs in utilizing historical data.

Conclusions:

• The proposed framework provides an accessible and effective method for incorporating historical data into model-assisted designs.
• This approach enhances the efficiency and robustness of phase I clinical trials.
• It bridges the gap in utilizing prior information for model-assisted designs, similar to model-based approaches.