Adoption of Bayesian Optimal Interval (BOIN) design in phase I oncology trials, trends, and challenges.

Abstract

e23006 Background: Traditionally, the 3+3 design (also known as classic design) has been the most widely used design in Phase I trials due to its simplicity and ease of implementation. However, it has several limitations, such as inefficiency in identifying the optimal dose and a high probability of under-dosing patients. The FDA has discouraged reliance on this design, as it does not formally model the probabilistic relationship between dose and toxicity. The Bayesian Optimal Interval (BOIN) design optimizes dose selection, reduces the number of patients exposed to subtherapeutic or overly toxic doses, and improves trial efficiency. The FDA's Oncology Center of Excellence (OCE) and its Project Optimus initiative encourage the adoption of adaptive and model-based dose-escalation methods, making BOIN a preferred alternative. Methods: A search of ClinicalTrials.gov was conducted, filtering for Phase I trials in solid tumors initiated in 2025. Trials were categorized based on dose-escalation design, investigational medicinal product (IMP) type, and other relevant parameters. All phase 1 trials started in 2025 were selected and analyzed. Results: Of the 641 Phase I trials initiated globally in 2025 for solid tumors, the majority employed the 3+3 design, often alongside accelerated titration or dose optimization strategies. Only 68 trials utilized the BOIN design or other adaptive design, which was slightly more prevalent in trials investigating targeted therapies and biologics. Conclusions: Despite growing interest in BOIN, Bayesian methods remain underutilized, with adoption progressing slowly. While the 3+3 design persists as the dominant approach, the FDA urges sponsors to justify its use and encourages the transition to more flexible, statistically rigorous and probabilistically based designs like BOIN to improve both the efficiency and safety of dose-finding in early-phase oncology trials.