A novel model-assisted design in phase I clinical trials: Bayesian optimal interval design / 中国临床药理学与治疗学

ABSTRACT

AIM: To introduce a novel and flexible model-assisted design for Phase I clinical trials Bayesian optimal interval (BOIN) design, including the process of implementation, practical implementation, and evaluation of its performance. METHODS: BOIN design decides dose escalation/de-escalation by comparing the observed toxicity rate at the current dose with an escalation boundary and a de-escalation boundary that are optimized to minimize the probability of making incorrect decision of dose assignment. The application of the BOIN design is illustrated using a trial example. RESULTS: BOIN combines the advantages of the algorithm-based methods and model-based methods. It enjoys desirable statistical properties -it is optimal, safe, robust and easy to implement. Simulation study shows that the BOIN substantially outperforms the existing designs with higher accuracy to identify the maximum tolerated dose (MTD). CONCLUSION: BOIN design possesses the similar statistical performance to the much more complicated model-based designs. It is simple to implement, and easy to calibrate to meet the safety requirement mandated by regulatory agents. The BOIN design has been widely used in different types of cancers. It is a novel design that holds great potential to substantially improve phase I trials in China.