Background: CD19 chimeric antigen receptor (CAR)-modified T cell therapy has demonstrated clinical efficacy but often associated with severe adverse effects manifested by cytokine release syndrome (CRS). To increase safety and efficacy of CAR T therapy, a 4thgeneration CAR design has been developed and investigated in a multi-center trial in China.Patients and Methods: From July 2013 to July 2016, the 4SCAR19 phase I/II multi-center trial has enrolled 125 patients (pts) with chemo-resistant, CD19-positive, acute B cell lymphoblastic leukemia (B-ALL) eligible for CAR T cell preparation and infusion. Laboratory data and clinical records were carefully evaluated and 102 pts were qualified for statistical evaluation, including 55 children and 47 adults; 27 had received allo-HSCT prior to CAR T therapy. The median age is 9 (2 to 17) and 37 (19 to 70) for pediatric and adult pts, respectively. The median leukemia blast count in the bone marrow (BM) is 14.5%, with BM blast >50% accounting for nearly one third (33 pts). Autologous/donor T cells were apheresis collected and transduced with an apoptosis-inducible, safety-engineered lentivector CAR containing four intracellular signaling domains: CD19-scFv//CD28/CD137/CD27/CD3ζ-iCasp9 (4SCAR19). Pts received conditioning regimens of cyclophosphamide (17), cyclophosphamide/fludarabine (54), other chemotherapy (29) or none (2), followed by CAR-T cell infusion (average 1.05x106cells/kg). The quality of apheresis cells, gene transfer and T cell proliferation efficiencies, and effective CAR T infusion dose were quantitatively monitored. Statistical analysis used COX proportional hazard model involving categorical or continuous covariates, univariates, or multivariates analyses, and survival analysis was based on right-censored data and Kaplan-Meier estimation (KM curve).Results: The compiled data indicate that the quality of CAR T cells positively correlated with overall survival (OS). The median follow-up time was 7 months (range from 1~35 months). Patient (Pt) cohort 1 (<50% BM blast count, 69 pts) and cohort 2 (≥50% BM blast count, 33 pts) achieved complete response (CR) at 91.3% and 75.8%, respectively. The median OS time of cohort 1 and cohort 2 are 485 days (CI: [387, NA] days) and 317 days (CI: [135, NA]), respectively (P=0.03). The average 4SCAR19 lentivector transduction efficiency was 37.3%. While the infusion dose of CAR T cells positively correlated with OS in pediatric pts (p=0.041), it lacked significant correlation in adults (p=0.95), suggesting that other factors rather than CAR T infusion dose play an important role in CAR T therapy in adults. When pts were analyzed based on low (< 5%) versus high (> 5%) BM blasts, the CRS grade showed no significant correlation with disease burden (P = 0.45 for low burden group, and P = 0.06 for high burden group). Of note that total 73 of the 102 pts experienced 0-1 grade CRS and 8 of them had very high BM leukemia load (>80%), suggesting a very low toxicity of the 4SCAR19 T cells. In addition, of the 17 high (> 80%) BM blast pts, only 3 experienced grade 3-4 CRS. For 38 pts with BM blast ≥ 50%, most had grade 1 (30) or grade 2 (13) CRS, and only 5 pts had grade 3, and 3 pts had grade 4 CRS. For low burden pts (0-5% BM blasts), 86% (42 pts) developed low grade CRS (0 or 1), and even pts with BM blasts above 5%, 53% experienced low grade CRS (0 or 1). Further analysis of inflammatory genetic profile reveals that high CRS might correlate with high inflammatory profile, as several pts with high inflammatory gene patterns, while only had residual disease or no detectable leukemia cells (BM blasts 0-0.005%), developed grade 3-4 CRS.Conclusion: The three-year follow-up of the 4SCAR19 T cell therapy further supports that CAR T immunotherapy could benefit not only low leukemia burden pts, but also late-stage, chemo-resistant, very high-burden leukemia pts. Importantly, our study demonstrates a good safety profile of the 4SCAR19 T cells even under high disease burden. While the multicenter trial involves 22 clinical centers, the variable clinical settings do not seem to impact patient outcomes due to the highly standardized CAR T cell preparation protocol and manageable CRS in most. DisclosuresNo relevant conflicts of interest to declare.