The therapy itself is a feat of cellular engineering. Clinicians harvest a patient’s own T lymphocytes, genetically reprogram them in the laboratory with a viral vector that encodes a chimeric antigen receptor recognizing CD19—a protein displayed on nearly all malignant B cells—expand the modified cells to billions, and reinfuse them after lymphodepleting chemotherapy. The engineered cells then hunt down and destroy CD19-bearing tumor cells wherever they find them, including, crucially, within the immune-privileged sanctuary of the central nervous system. That last point matters enormously, because the pivotal registration trials for the three commercially approved products—axicabtagene ciloleucel, tisagenlecleucel and lisocabtagene maraleucel—largely excluded patients with CNS involvement. The worry was twofold: that the cells might not cross the blood-brain barrier in sufficient numbers, and that activating potent immune cells near the brain could trigger catastrophic neurotoxicity.

Those concerns kept CNS lymphoma patients, the very people with the fewest alternatives, out of the trials that defined the field. Evidence for them accumulated instead in retrospective cohorts, registry studies and small prospective series, each too limited to settle the question. The new analysis, conducted in accordance with PRISMA guidelines and prospectively registered in PROSPERO, set out to synthesize that fragmented literature into usable estimates. The investigators systematically searched PubMed, Google Scholar, Semantic Scholar, ClinicalTrials.gov and the Cochrane Library from inception to 31 August 2025, screening 1,247 records down to 23 eligible studies—18 retrospective and 5 prospective—comprising 547 adults who had typically received at least two prior lines of therapy.

The pooled survival figures carry real weight for a population with otherwise vanishing options. Twelve-month overall survival reached 64.5 percent, with a 95 percent confidence interval spanning 49.0 to 76.0 percent, and 12-month progression-free survival was 51.0 percent, with a wider interval of 35.0 to 68.0 percent. Given that historical outcomes after CNS lymphoma relapse are often measured in a handful of months, a median expectation of roughly a year progression-free for half of all comers represents a genuine shift. The estimates are comparable to those reported in the landmark trials of CAR T-cell therapy for systemic large B-cell lymphoma, suggesting that the blood-brain barrier, once feared as an impenetrable obstacle, does not negate the therapy’s activity.

Response data, analyzed separately for the two disease subtypes, add nuance. Patients with primary CNS lymphoma, in which the malignancy originates in the brain, achieved a pooled complete response rate of 52 percent, with remarkably consistent results across studies—heterogeneity sat at just 32 percent. Secondary CNS lymphoma, which arises when systemic lymphoma spreads into the nervous system, fared somewhat less well, with a pooled complete response rate of 40 percent and moderate heterogeneity of 76 percent. The authors attribute the gap to disease biology: secondary involvement typically coincides with uncontrolled systemic disease and higher tumor burden, both of which erode the odds that even a powerful cellular therapy can reset. Sensitivity analysis showed that a single large EBMT registry study contributed substantially to that heterogeneity; excluding it dropped the figure to 40 percent.

Safety has always been the central anxiety for this indication, and the analysis addresses it directly. Cytokine release syndrome—the systemic inflammatory cascade unleashed when massive numbers of immune cells activate simultaneously—occurred in 76 percent of PCNSL patients and 77 percent of SCNSL patients, but with zero heterogeneity across studies, meaning the rate was startlingly uniform. Immune effector cell-associated neurotoxicity syndrome, the cerebral edema and encephalopathy spectrum that gave trial designers pause, occurred in 50 percent of PCNSL patients and 46 percent of SCNSL patients. Crucially, most events on both fronts were low to moderate in severity and reversed with standard supportive care: tocilizumab, the interleukin-6 receptor antibody that has become the workhorse for CRS, and corticosteroids for neurotoxicity. The rates are broadly consistent with those seen in systemic lymphoma populations, though the authors caution that the analysis lacked a non-CNS comparator arm for direct comparison.

Methodological rigor underpins the estimates. The team used a random-effects model of proportions with the Freeman-Tukey double arcsine transformation to stabilize variance, quantified heterogeneity with the I² statistic, and performed leave-one-out sensitivity analyses to test robustness. Survival probabilities were extracted from Kaplan-Meier curves where necessary, verified independently by two reviewers, and corresponding authors were contacted for missing data. Risk of bias was formally assessed with the ROBINS-I tool, yielding a sobering distribution: seven studies at low risk, seven at moderate risk and nine at serious risk, with retrospective design, confounding and selective participant selection the recurring culprits. Funnel plot inspection found no obvious asymmetry, but the small number of studies per outcome limits how much can be read into that.

The heterogeneity statistics themselves tell an honest story. Overall survival pooled at moderate heterogeneity of 67.6 percent, and progression-free survival at high heterogeneity of 80.2 percent, reflecting the mosaic of patient characteristics, tumor burdens, prior regimens and CAR T-cell constructs stitched together across the included studies. Some patients received axicabtagene ciloleucel, others tisagenlecleucel or lisocabtagene maraleucel, and several trials—particularly from Chinese centers—tested locally manufactured CD19 or CD19/CD22 bispecific constructs. Median follow-up ranged from 5.1 to 37.5 months, clustering around one year, which is precisely the horizon over which the pooled estimates are calculated and too short to say whether deep remissions will hold at three or five years. Disease-specific survival pooling was also impossible, because most studies reported combined outcomes without adequate subgroup data.

Those limitations do not diminish the clinical signal; they simply bound it. For relapsed or refractory CNS lymphoma, where the previous standard of care offered little more than palliation, a therapy that produces complete responses in roughly half of primary CNS lymphoma patients and keeps half of all patients progression-free at one year, with toxicities that are frequent but manageable, has earned a place in the treatment algorithm. The finding that CAR T cells can operate effectively behind the blood-brain barrier also validates years of mechanistic work suggesting the barrier, while restrictive to antibodies and small molecules, is navigable by activated lymphocytes—whether they arrived there naturally or were engineered to be lethal.

What the field needs next is unambiguous. The authors call for large prospective multicentre trials with standardized outcome reporting and longer follow-up to define durability, late relapse patterns and long-term safety, and to determine which product, bridging strategy and consolidation approach works best for which subtype. Open questions abound: whether sequential CD19/CD22 targeting can forestall antigen-loss relapse, how lymphodepletion intensity shapes CNS penetration, and whether minimal residual disease monitoring can identify failing responders early enough for salvage. Until those trials mature, the pooled evidence positions CD19-directed CAR T-cell therapy as the most credible lifeline yet for a disease that, until recently, offered almost none.

Subject of Research: Efficacy and safety of CD19-directed CAR T-cell therapy for relapsed or refractory primary and secondary central nervous system lymphoma

Article Title: Efficacy and Safety of CAR T‐Cell Therapy for Primary and Secondary Central Nervous System Lymphoma: A Systematic Review and Meta‐Analysis

Article References: Khan, A., Malik, A. A., Kareem, W., Idrees, M., Kamil, K. A., Irshad, K., Baig, Z. M., Yahya, A., Qureshi, W., Ahmad, M., & Bilanda, B. (2026). Efficacy and Safety of CAR T‐Cell Therapy for Primary and Secondary Central Nervous System Lymphoma: A Systematic Review and Meta‐Analysis. eJHaem, 7(5), Article e70407. https://doi.org/10.1002/jha2.70407

Image Credits: AI Generated

DOI: 10.1002/jha2.70407

Keywords: CAR T-cell therapy, central nervous system lymphoma, PCNSL, SCNSL, CD19, meta-analysis, cytokine release syndrome, ICANS, immunotherapy, hematologic malignancy, overall survival, blood-brain barrier