Infusion of T cells expressing a GD2 specific chimeric antigen receptor (CAR) in patients with neuroblastoma can induce sustained complete remissions. Introducing costimulatory endodomains into the CAR may improve the antitumor potential of T cells; therefore, we conducted a Phase I study with adaptive trial design of T cells expressing a GD2 CAR with both CD28 and OX40 costimulatory endodomains. We tested the safety, persistence and antitumor efficacy of the CAR T in 11 patients with neuroblastoma, in 3 cohorts. The 1st cohort received escalating doses of thawed GD2 CAR T cells that had been grown with interleukin-2 (IL2). The 2nd cohort received lymphodepletion with fludarabine and cyclophosphamide followed by escalating doses of fresh cells grown with interleukin-7 and -15 (IL7/15). The 3rd cohort received 2 doses of the PD1 inhibitor, pembrolizumab following infusion of CAR T as per cohort 2. CAR expression was similar in cells irrespective of cohort (median 65%; range 43%-83%). In vivo CAR T cell expansion was greater in cohorts 2 and 3 than in cohort 1 at week 1 post-infusion (p<0.01), as was persistence at week 4 (p<0.02), although CAR T cell numbers in all patients declined after their peak at +7-14 days. The treatments were tolerated well and no dose limiting toxicities were observed. Antitumor responses were modest: at 6 weeks 1 patient had mixed response, 4 had stable disease and 6 patients progressed. Thus, T cells expressing a GD2 CAR with CD28 and OX40 costimulatory endodomains are safe, expand and are detectable for >4 weeks in vivo but early antitumor responses were disappointing even after lymphodepletion and co-administration of PD1 Ab.