Dose-finding designs in pediatric phase I clinical trials: Comparison by simulations in a realistic timeline framework

Objective: Usual dose-finding methods in oncology are sequential. Accrual is suspended after each group of patients to assess toxicity before increasing the dose. An adapted Continual Reassessment Method (CRM) and Rolling 6 (R6) method, designed to avoid this suspension of accrual in pediatric oncology, are compared with the traditional 3 + 3 design. Study design and setting: The competing performances were evaluated in a simulation study integrating the temporal dimension, and a phase I trial was reanalyzed. We compared methods for various interpatient arrival times and dose-toxicity relations, in terms of distribution of final recommendations, number of skipped children and duration of trials. Results: R6 and CRM can be safely implemented to limit trial suspensions, especially when mean interpatient arrival time is short. CRM was found to be more efficient than algorithm-based methods (44% of good recommendations vs. 38%) but moderately increased the risk of overtreatment. The R6 design included more patients at suboptimal doses. The design with the shortest study duration depended on the number of dose to escalate before the target. Conclusion: These new methods can reduce the number of skipped patients, but only provide limited gain in terms of ability to select the right dose. New designs are needed.