Purpose Identifying the temporal pattern of recurrence and prognostic biomarkers would further help improve the efficacy of chimeric antigen receptor (CAR) -T therapy. Methods We examined the prognoses of 119 patients after sequential infusion of anti-CD19 and anti-CD22, a cocktail of 2 single-target CAR (CAR19/22) T cells in an open-label, single-center clinical trial (ChiCTR-OPN-16008526). And we, from a 70-biomarker panel, identified candidate cytokines that might predict the treatment failure, including primary non-response (NR) and early relapse (ER). Results In our study, 3 (11.5%) patients with B-cell acute lymphoblastic leukemia (B-ALL) and 9 (12.2%) cases of B-cell non-Hodgkin lymphoma (NHL) failed to respond to sequential CAR19/22 T-cell infusion (NR). A total of 11 (42.3%) B-ALL patients and 30 (52.7%) B-NHL patients had relapses during follow-up. Most recurrence events (67.5%) occurred within six months of sequential CAR T-cell infusion (ER). We found that macrophage inflammatory protein (MIP)-3 alpha was a highly sensitive and specific prognostic predictor for patients with NR/ER and those attaining over-6-month remission. Patients who had higher MIP3 alpha levels after sequential CAR19/22 T-cell infusion had significantly favorable progression-free survival (PFS) than their counterparts with relatively lower MIP3 alpha expression. Our experiments demonstrated that MIP3 alpha could enhance the therapeutic effect of CAR-T cells by promoting T-cell infiltration into and enriching memory-phenotype T cells in the tumor environment. Conclusion This study showed that relapse occurred mainly within six months after sequential CAR19/22 T-cell infusion. Moreover, MIP3 alpha could act as a valuable post-infusion biomarker for identifying patients with NR/ER.