Design of a Phase I Drug Combination Study with Adaptive Allocation Based on Dose-Limiting Toxicity Attribution

<b>Background:</b> This article describes the adaptation of a Phase I drug combination method to incorporate dose-limiting toxicity (DLT) attribution in dose assignments. The study is motivated by the Embolden trial (NCT03240211), a Phase Ib, multicenter trial at the UVA Comprehensive Cancer Center evaluating pembrolizumab with pralatrexate (Arm A), decitabine (Arm C), or both (Arm B) in relapsed/refractory peripheral and cutaneous T cell lymphomas. <b>Methods:</b> While Arms A and C used monotherapy dose escalation, Arm B required simultaneous escalation of both agents, integrating drug-specific DLT attribution to guide dosing. <b>Results:</b> We adapted the partial order continual reassessment method (POCRM) to incorporate this attribution, ensuring appropriate de-escalation of the offending agent. Given the trial's complexity, software modifications were necessary to evaluate design performance through simulations. <b>Conclusions:</b> This work underscores the importance of novel dose-finding strategies in early-phase trials and aims to promote their broader adoption for improved trial efficiency and transparency.