Multi-Agent Phase I Trials in Childhood Acute Lymphoblastic Leukemia (ALL).

Background: Single agent phase I trials with conventional methodology may not be suitable for children with relapsed leukemia. Accrual of children with ALL in relapse to single-agent phase I trials is poor due to clinical urgency and a > 30% likelihood of complete response (CR) with a variety conventional agents combinations (Br J Haematol. 2005; 131(5): 579) with the option of hematopoietic stem cell transplant in remission. As most drugs are ultimately used in combination, a Phase I trial testing a new agent in combination with conventional agents would seem most useful and might increase accrual. However, with conventional phase I methodologies determination of a maximum tolerated dose is complicated by the toxicities of the accompanying conventional agents and by the background morbidity of relapsed leukemia. Methods: The Children’s Oncology Group (COG) study, AALL01P2, employed vincristine, prednisone, doxorubicin, and pegylated asparaginase for children with ALL in first marrow relapse. We determined the incidence of conventional non-hematologic dose limiting toxicities (DLT’s) and modeled the impact on a hypothetical phase I trial of a candidate agent with no additional toxicity. Results: Among 111 patients on AALL01P2, 19% had conventional non-hematologic DLT’s. Induction therapy was judged clinically acceptable. With a traditional Phase I escalation scheme that accepts 0/3 and 1/6 DLT’s at a dose-level and rejects 2/3 and 2/6 DLT’s, an agent that adds no morbidity would be rejected as too toxic at any dose 30% of the time. Conclusion: Background morbidity confounds identification of an acceptable dose of a non-toxic new agent tested in combination with conventional drugs for recurrent ALL. We propose a modification to the traditional Phase I design that increases the DLT thresholds to 1/3 and 2/6, which effectively compensates for background toxicity and reduces the chance of falsely rejecting an acceptable agent.