Two-dimensional calibration-free odds (2dCFO) design for phase I drug-combination trials

Abstract

In the field of oncology, combination drug therapy has become a popular approach for treating cancer patients. The use of multiple drugs with different mechanisms of action can lead to synergistic effects and minimize toxicity, resulting in better outcomes for patients. However, the small sample sizes of phase I clinical trials for drug combinations have made it difficult to optimize dosing strategies and evaluate the safety and efficacy of these therapies. To address this issue, researchers have developed experimental designs for these trials that are both efficient and reliable. Most of the existing experimental designs for drug-combination trials are parametric and model-based, requiring tuning parameters or prior knowledge of drug toxicity probabilities. These designs are often limited in their ability to accurately model complex dose-toxicity relationships and can be computationally intensive. To overcome these limitations, I have proposed a new experimental design called the two-dimensional calibration-free odds (2dCFO) design. The 2dCFO design is a non-parametric and model-free approach that relies solely on odds ratios and does not require any assumptions about dose-toxicity curves. This design uses information from not only the current dose but also the neighborhood doses (i.e., left, right, up, and down directions) to make more accurate decisions, making it more efficient than interval-based designs. The 2dCFO design is able to estimate the dose-toxicity relationship using a local odds ratio estimation method, which allows it to accurately determine the optimal dose for each cohort. Extensive simulation results have demonstrated that the 2dCFO design is both accurate and efficient, and exhibits great robustness. The 2dCFO design is able to accurately estimate the dosetoxicity relationship and determine the optimal dose for each cohort, while minimizing the number of patients required for the trial. This design holds promise for expediting drug development and enhancing patient outcomes in the battle against cancer, providing a valuable tool for oncologists and researchers. In conclusion, the 2dCFO design offers a promising approach for optimizing dosing strategies and evaluating the safety and efficacy of combination drug therapies in oncology. Its non-parametric and model-free design, combined with its ability to use information from neighborhood doses, makes it a highly efficient and reliable method for drug-combination trials. The 2dCFO design could potentially accelerate the drug development process, leading to more effective treatments and improved outcomes for cancer patients.