Evaluation of statistical designs of phase I/II clinical trials for cancer patients published in 2005
14077 Background: Phase I trials determine the maximal safe dose that could be used in phase II trials. Designs are based on the assumption that efficacy and toxicity increase with dose. Phase I/II trials determine the safety, dosage levels, and response rate. This review addresses statistical issues of phase I/II studies designs. Methods: We reviewed phase I/II clinical trials for cancer patients published in 2005. The main criteria were: type of treatment, statistical design, endpoints, expected efficacy and toxicity, one- or two-steps designs, dose levels, definition of Dose Limiting Toxicity and recommended dose, objective response, survival, patient selection and follow-up. Results: 41 phase I/II trials were found. All but one, targeted a specific type of tumor. 14 studies included combined cytotoxic therapies. 21 studies included a cytotoxic agent combined with a targeted therapy (12) or with radiations (9). Others used monochemotherapy, immunotherapy, vaccine or gene therapy. 23 studies were a two steps design, i.e. a phase I followed by a phase II trial, and used a classical Fibonacci escalation dose model. All others used a one-step design evaluating efficacy and toxicity concomitantly. Among them, 3 had a Fibonacci-like design with a desescalation model and 4 had a randomization to different dose levels. In 1 trial, dose escalation was performed in the same patient. In the 10 remaining studies, 1 evaluated only one dose level and was improperly presented as a phase I/II study, and 9 did not describe any statistical design. DLT was described in only 27 trials. Also, recommended doses for further trials were only provided in 30 studies. Efficacy was evaluated with clinical or radiological response for 34 studies, biological marker was evaluated in 5 cases and time to progression in 2 cases. Conclusion: Most of the phase I/II trials published in 2005 used a classical two steps design with an adapted Fibonacci dose escalation. None of them used new designs such as continual reassessment method (CRM), which have the advantage to incorporate data during the course of the trial, leading to optimization of the study in terms of cost and speed. Methodological progresses are necessary to address issues related to multiple endpoints and to help clinicians to feel comfortable with the CRM. No significant financial relationships to disclose.