Safety and efficacy of CAR T-cell therapy in central nervous system lymphoma: a systematic review and meta-analysis

Background: The safety and efficacy of Chimeric Antigen Receptor (CAR) T-cell therapy in Central Nervous System Lymphoma (CNSL) remain uncertain, given the limited representation of CNS involvement in pivotal CAR T-cell trials. This meta-analysis synthesized data from studies evaluating outcomes in both primary (PCNSL) and secondary CNS lymphoma (SCNSL) cohorts treated with CART. Methods: A comprehensive search was performed across PubMed, Embase, and clinical registries up to October 2025. Studies reporting efficacy or toxicity outcomes in CNSL following CD19-targeted CART therapy were included. Proportions were stabilized using the Freeman-Tukey double arcsine transformation and pooled using inverse variance weighting. Between-study heterogeneity was estimated with the DerSimonian-Laird method, and τ² confidence intervals were calculated via the Jackson approach. Publication bias was assessed using Egger's regression. Subgroup and multiple meta-regression analyses evaluated moderators including CNS category (PCNSL vs SCNSL) and publication year. Results: Data from thirty-eight studies were meta-analyzed. The pooled overall response rate (ORR) was 0.75 95% CI: 0.70-0.79 with moderate-to-substantial heterogeneity (I² = 53.3%). Complete response (CR) rate was 0.52 0.46-0.58, and partial response (PR) rate 0.18 0.14-0.22. No significant publication bias was detected (Egger's p = 0.39 for ORR). Meta-regression indicated no significant effect of publication year or lymphoma subtype on response. Cytokine Release Syndrome (CRS) occurred in 83.5% 79.0-88.0, with grade ≥3 CRS in 5.77% 3.0-9.0. Immune Effector Cell-Associated Neurotoxicity Syndrome (ICANS) was reported in 44.9% 36.0-54.0, with severe (grade ≥3) events in 17.4% 12.0-23.0. Heterogeneity was substantial for ICANS (I² = 84.2%) but moderate for CRS (I² = 64.3%). Conclusions: CART therapy achieves robust response rates in CNS lymphoma, with efficacy comparable between PCNSL and SCNSL. Neurotoxicity remains frequent but manageable, and severe CRS events are infrequent. Systematic review registration: https://www.crd.york.ac.uk/prospero/, identifier CRD420251070033.