Introduction In pediatric oncology, therapeutic strategies are limited for relapsed patients. Matching treatment to molecular abnormalities appears as a promising strategy. Due to the heterogeneity of the detected molecular abnormalities and the low number of pediatric patients, Gustave Roussy with the ITCC network has launched a platform trial of 10 phase I/II parallel arms. In each arm, the design integrates a Continual Reassessment Method (CRM) to identify the recommended phase 2 dose (RP2D) defined as the dose associated with 25 % of severe toxicity and an Ensign three-stage design to test whether preliminary estimate of the response rate is below 10 % versus higher than 30 % at the 10 % level with a 90 % power. The two stages are intricated as patients treated at the identified MTD will contribute to the activity assessment and vice versa. The objective of this presentation is to report on the operating characteristics of the chosen design, to illustrate the limits of the phase II cohorts plugged in phase I and the risk of missing an active dose due to erroneous dose selection. Methods Each arm is run independently. We performed a series of simulations to assess the performance of the design of the study. We investigated the probability to identify an active drug for various dose-toxicity and dose-activity relationship. In particular, we considered the case of narrow therapeutic index (doses below the MTD are not active), the case of plateau in activity and the case of active doses deemed too toxic. In each case, after identifying the MTD and applying the Ensign design, we performed the test and evaluated the power. We also re-estimated the risk of toxicity after completion of both the phase I and the phase II parts. Finally, the expected number of patient to be enrolled in this design, the percentage of correct dose recommendation and the percentage of arms declared positive will be provided for selected scenarios. Results For one arm, the percentage of correct MTD selection was 57 %, regardless of the dose-activity relation. If the MTD is active and associated with a 40 % response rate, a mean number of 32 patients were enrolled and the percentage of arm declared active was 86 %. Conversely, if the MTD is not active and associated with a 10 % response rate, a mean number of 26 patients were enrolled and the percentage of arms declared active reached 12 %. Moreover, if 50 % of arms were declared active, the total number of patients to be enrolled was 290. Finally, if all arms were stopped for lack of activity, the overall type error I was 28 %. So far, 77 patients have been enrolled and the key aspects of the statistical implementation will be provided. Conclusion This is the largest phase I/II pediatric program based on tumor biology running worldwide. This design enables to propose multiple personalized therapeutic options to improve comprehension and treatment of pediatric cancers. However, integration of phase I and II is appealing but raises practical difficulties due to variation in analysis populations for toxicity and activity or observation of responses at doses higher or lower than the RP2D. We showed that finding the correct MTD is key to select active treatments and more patients should be dedicated to dose finding.