Chimeric antigen receptor (CAR) modified T cells targeted CD19 showed promising clinical outcomes in treatment of B cell malignances such as chronic lymphocytic leukemia (CLL), acute lymphoblastic leukemia (ALL) and other indolent lymphomas. However, the clinical benefit varies tremendously among different trials. This meta-analysis investigated the efficacy (response rates and survival time) of CD19-CAR T cells in refractory B cell malignances in Phase I clinical trials. We searched publications between 1991 and 2014 from PubMed and Web of Science. Pooled response rates were calculated using random-effects models. Heterogeneity was investigated by subgroup analysis and meta-regression. Fourteen clinical trials including 119 patients were eligible for response rate evaluation, 62 patients in 12 clinical trials were eligible for progression-free survival analysis. The overall pooled response rate of CD19-CAR T cells was 73% (95% confidence interval [CI]: 46-94%). Significant heterogeneity across estimates of response rates was observed (p < 0.001, I2=88.3%). ALL patients have higher response rate (93%, 95% CI: 65-100%) than CLL (62%, 95% CI: 27-93%) and lymphoma patients (36%, 95% CI: 1-83%). Meta-regression analysis identified lymphodepletion and no IL-2 administrated T cells as two key factors associated with better clinical response. Lymphodepletion and higher infused CAR T cell number were associated with better prognosis. In conclusion, this meta-analysis showed a high clinical response rate of CD19-CAR T cell-based immunotherapy in treatment of refractory B cell malignancies. Lymphodepletion and increasing number of infused CD19-CAR T cells have positive correlations with the clinical efficiency, on the contrary, IL-2 administration to T cells is not recommended.