Introduction Chimeric antigen receptor T-cell (CAR-T) therapy has revolutionized the treatment of large B-cell lymphomas (LBCL), demonstrating remarkable efficacy. However, CAR-T therapy is also associated with significant toxicities, some arising from relative immunologic incompetence after treatment. A key knowledge gap persists regarding how the immune landscape changes post-CAR-T, defined as immune reconstitution (IR), and the influence of IR on patient outcomes. In this study, we set out to describe patterns and determinants of IR following CAR-T as well as impacts of IR on outcomes. Methods Adult patients with primary or transformed LBCL treated with CD19-CAR-T cells, namely axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), and lisocabtagene maraleucel (liso-cel), between 2017 and 2022 were included in this single center retrospective study. Immune cell levels were measured at various timepoints after CAR-T. Median cell counts and interquartile range (IQR) at pre-specified time windows (using the first available observation) were used to describe temporal IR patterns. Multivariable generalized estimating equation (GEE) models with cubic time variables were used to investigate the association between cell count trajectory and baseline characteristics. A multivariable Cox model with longitudinal cell count as a time-varying covariate was utilized to assess the association between IR and overall survival (OS), adjusted for confounding factors. Kaplan-Meier (KM) curves were generated for visualization. Results A total of 210 patients were included, of whom 153 had at least one CD4+ T cell measurement from time of CAR-T. The median age was 65 (range 20-86) and most were male (65.7%) and diagnosed with diffuse large B-cell lymphoma (79.5%). Axi-cel (49.5%) was the most common product, followed by tisa-cel (28.6%) and liso-cel (21.9%). Lymphodepletion was mostly cyclophosphamide and fludarabine (82.4%). Of the 210 patients, 58.1% achieved a complete response (CR) and 17.1% a partial response by 100 days after CAR-T. Following CAR-T, CD3+ T cells were the most abundant circulating lymphocyte. By 30 days (±15 days) after CAR-T, median (IQR) values of CD3+ T cells, B cells, and NK cells were 445 (162-776), 0 (0-0), and 90 cells/ul (57-146), respectively; by 60 days (±15 days), median (IQR) values were 498 (220-706), 0 (0-0), and 104 cells/ul (69-169), respectively. Proportions of CD4+ and CD8+ T cells remained similar through one year after CAR-T, with CCR7-45RA+ TEMRA and CCR7-45RA- effector memory cells dominating among CD8's, and CCR7-45RA- effector memory cells being the most abundant among CD4's. Regulatory T cells (CD25+CD127-) comprised about 10% of the CD4+ T cell lymphocyte population, with a median (IQR) of 9 cells/ul (4.5-15.5) by day 30 (±15 days) after CAR-T. Finally, receipt of tisa-cel was associated with greater levels of CD4+ EM T cells compared with axi-cel and liso-cel (median [IQR]: 253 [158-406] vs 81 [51-138] and 113 cells/ul [88-154], respectively), by 30 days (±15 days) after CAR-T. According to a multivariable GEE model, patients receiving tisa-cel had significantly higher CD4+ count trajectories compared to those receiving axi-cel or liso-cel ( p<0.001 for both) ( Figure A). In addition, patients with more extensive pre-treatment (i.e., 6+ prior lines) had significantly lower CD4+ count trajectories compared to those who received 2-3 or 4-5 prior lines ( p=0.015 and p=0.004, respectively). A multivariable Cox model, moreover, indicated that a higher CD4+ count (in units of 50 cells/ul) was significantly associated with better OS (adjusted HR: 0.845; 95% CI, 0.745-0.960; p=0.010), adjusted for product and prior lines of therapy. KM curves by patients above or below the median of the last observed CD4+ count further demonstrated this significant association between longitudinal IR and OS ( Figure B). Conclusion In this single-center retrospective study, we extensively characterized the immune environment at time of and following CAR-T. To the best of our knowledge, this is the largest study of its kind. Additionally, we identified several factors that impact IR, which in turn shapes OS. In doing so, we show that CD4+ T lymphocytes are key players in disease response post-CAR-T and suggest that tracking CD4+ recovery following CAR-T may provide both diagnostic and therapeutic benefit.