Pediatric trials present many design and analysis challenges arising from special ethical issues and the relatively small accessible patient population. Two recent trial innovations, outcome adaptive randomization and adaptive designs that utilize historical data, may help addressing the challenges. We combine the two ideas and evaluate the utilities. A real life trial provides a motivating example and context for simulation studies. We used a doubly adaptive biased coin design (DBCD) targeting randomization probabilities that minimize non-favorable outcomes and modified it to accommodate the constraints of the trial. Monte Carlo computer simulation was used to compare the design with fixed randomization designs including the original design for operating characteristics such as % of the study subjects treated with the superior treatment arm, % with non-favorable primary outcome, % stopped early correctly and % stopped wrongfully. Accrual and outcome data was simulated using the trial data for different accrual speed and treatment effects. We show that outcome adaptive randomization, if implemented efficiently, is applicable to multi-site comparative trials with modest yet reliable benefits.