Dose-finding designs using a novel quasi-continuous endpoint for multiple toxicities.

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Phase I oncology trials aim to find safe drug doses using dose-limiting toxicity (DLT).
• DLT, a binary endpoint, may discard valuable toxicity data, especially for molecularly targeted therapies.
• Existing methods may not fully capture the complexity of multiple toxicities.

Purpose of the Study:

• To propose a quantitative and comprehensive toxicity score, the total toxicity profile (TTP).
• To develop a novel dose-finding design, the quasi-likelihood continual reassessment method (CRM), incorporating the TTP score.
• To evaluate the performance of the TTP-based CRM against existing designs.

Main Methods:

• Defined TTP as the Euclidean norm of weighted toxicities, reflecting clinical importance.
• Developed a frequentist quasi-likelihood CRM with a logistic dose-toxicity model.
• Conducted simulations comparing the proposed design with three other TTP-based designs under various scenarios.

Main Results:

• All TTP-based designs demonstrated good performance in identifying recommended doses across most scenarios.
• The proposed quasi-likelihood CRM showed excellent overdosing control.
• For a sample size of 36, the quasi-likelihood CRM achieved 80%-90% correct selection rates.

Conclusions:

• The total toxicity profile (TTP) offers a more comprehensive measure of toxicity than traditional DLT.
• The quasi-likelihood CRM incorporating TTP is an effective and accurate dose-finding design.
• TTP-based designs are a promising alternative to conventional methods, particularly for molecularly targeted agents.