CFO

CFO implements calibration-free odds type Bayesian designs for dose finding in phase I clinical trials, leveraging historical cumulative data across various dose levels to identify the maximum tolerated dose (MTD) while maintaining an optimal balance between efficiency and safety.

Key Features:

• Fundamental CFO design: Implements the core calibration-free odds type Bayesian approach that uses historical cumulative data across dose levels for MTD identification.
• Two-dimensional CFO (2dCFO): Extends the CFO design for drug-combination trials.
• Accumulative CFO (aCFO): Accrues all dose information for decision making.
• Time-to-event CFO (TITE-CFO): Addresses late-onset toxicity in trials.
• Fractional CFO (fCFO): Addresses late-onset toxicity through fractional approaches.
• Hybrid designs: Includes TITE-aCFO and f-aCFO, which integrate both late-onset toxicity and all dose information for decision making.
• Randomized CFO (rCFO): Integrates exploration and exploitation mechanisms from reinforcement learning into the CFO design.
• Core functions: Provides functions for determining subsequent cohort doses, selecting the MTD, and conducting simulations to evaluate design operating characteristics.

Scientific Applications:

• Phase I clinical trials: Identifies the maximum tolerated dose (MTD) in dose-finding studies for single-agent and drug-combination therapies.
• Late-onset toxicity assessment: Handles scenarios where toxicity outcomes are not immediately observable through TITE-CFO, fCFO, and hybrid variants.
• Drug-combination trials: Determines optimal dose combinations in two-dimensional dose-finding studies using 2dCFO.
• Design evaluation: Conducts simulations to evaluate operating characteristics of different CFO-type designs under various trial scenarios.

Methodology:

CFO-type designs employ a data-driven decision-making Bayesian approach that leverages historical cumulative data across various dose levels. The design mimics "tug-of-war" games from game theory, where one force pushes the dose down while another pushes it up. The rCFO variant integrates exploration and exploitation mechanisms from reinforcement learning. Extensive simulations validate design performance metrics against state-of-the-art methods.