EWOUC ‐ NETS ‐ COM : An Efficient Bayesian Adaptive Design for Combination Treatments With Overdose and Underdose Control

Combination treatments (COM) are increasingly used in oncology to boost efficacy, limit resistance, and reduce side effects, requiring new dose‐finding designs that balance safety and efficacy while accounting for potential drug interactions. To meet this need, we propose EWOUC‐NETS‐COM, a novel Bayesian design extending our Escalation with Overdose and Underdose Control (EWOUC) f ramework. This design jointly models toxicity and efficacy through a bivariate normal structure. It incorporates the Normalized Equivalent Toxicity Score (NETS) as a quasi‐continuous toxicity endpoint, enhancing granularity in toxicity assessment of combination treatment. EWOUC‐NETS‐COM introduces three key innovations: (1) a utility‐based allocation algorithm that adapts to emerging drug synergy; (2) two‐dimensional overdose and underdose control for more informed dose exclusion and early stopping; and (3) continuous dual endpoints to reduce information loss and improve precision. Extensive simulations and application in a real‐world trial demonstrate that EWOUC‐NETS‐COM outperforms existing designs in selecting optimal doses, preserving patient safety, ensuring therapeutic efficacy, and maximizing trial efficiency. This practical and flexible design is well‐suited for guiding dose escalation and de‐escalation in two‐drug combination trials and is readily extendable to more complex multi‐agent regimens.