Adding intrathecal immune checkpoint inhibitors, delivered into the fluid around the brain and spinal cord, to systemic therapy was not linked to a clear increase in side effects, called adverse events in the review, among patients with leptomeningeal disease, or LMD. The review's strongest safety signal came from an event-level analysis: it reported 72% fewer grade 3 or higher adverse events per patient than with systemic therapy alone. Yet patient-level and time-adjusted analyses did not show statistically clear differences, so the result does not establish a safety advantage for intrathecal treatment.
How the comparison was assembled
The question was whether adjunctive intrathecal immune checkpoint inhibitors, given alongside systemic treatment, differed in safety and efficacy from systemic immune checkpoint inhibitors alone. Safety was the primary outcome, measured through adverse events graded with CTCAE. Overall survival and progression-free survival were secondary outcomes.
The review followed PRISMA and searched PubMed, Embase and Scopus without date limits. The researchers also screened reference lists and clinical-trial registries. Eligible studies had to confirm LMD by cerebrospinal-fluid cytology or MRI, include at least one immune checkpoint inhibitor, and report adverse events and/or survival. Preclinical studies, reviews, incomplete conference abstracts, duplicates and studies without patient counts were excluded.
Of 94 studies screened, 28 were included. Across them, 542 patients were screened and 201 immune checkpoint inhibitor-treated patients were analyzed: 161 received systemic treatment alone and 40 received intrathecal treatment with concurrent systemic therapy. Study-specific adverse-event rates were pooled with a random-effects meta-analysis, and the review also estimated risk at the patient level.
The cohorts were not evenly matched. The systemic group included skin, lung and breast cancers at 45.3%, 22.9% and 19.3%, respectively. The intrathecal group was 97.5% melanoma and 2.5% breast cancer, and all of its patients received PD-1 inhibitors. The authors noted that retrospective intervention studies can be affected by confounding, treatment-selection bias and overlapping therapies.
The strongest signal came from event counts
At the patient level, pooled risk of any adverse event was 0.70 with intrathecal treatment and 0.75 with systemic treatment alone. The relative risk was 0.94, with an interval from 0.54 to 1.63 and p = 0.83. In other words, the estimated chance of having at least one recorded event was similar in the two cohorts.
For grade 3 or higher events, a category used for more serious toxicity, pooled patient-level risk was 0.17 in the intrathecal cohort versus 0.34 in the systemic cohort. The relative risk was 0.50, but the interval ran from 0.22 to 1.13 and p = 0.096. The apparent reduction was therefore not statistically significant.
Event-level analysis produced a different headline number. It reported 72% fewer grade 3 or higher adverse events per patient with adjunctive intrathecal therapy, a rate ratio of 0.28, a 95% confidence interval of 0.12 to 0.64 and p = 0.003. The result persisted when systemic comparators were limited to PD-1-only studies, with a rate ratio of 0.26 and an interval of 0.10 to 0.68.
That signal weakened after exposure time was taken into account. Total adverse-event rates were 0.177 versus 0.201 events per patient-month, and none of the reported time-adjusted comparisons was statistically significant. For grade 3 or higher events, rates were 0.013 versus 0.057 per patient-month, with a relative risk of 0.22 and p = 0.362. A dose-adjusted analysis found no significant association between cumulative immune checkpoint inhibitor exposure and adverse-event outcomes, with p = 0.873 for any event and p = 0.294 for grade 3 or higher events.
Survival results did not settle the question
The efficacy results were no clearer. Median progression-free survival, a measure of time before disease progression, was 6.0 months with systemic treatment and 7.0 months with intrathecal treatment; the difference was not statistically significant, with p = 0.11. Median overall survival was 9.0 versus 5.0 months, also without a statistically significant difference, with p = 0.38. Multivariable adjustment found no significant difference in either PFS or OS.
The authors also described two institutional cases in a separate, IRB-approved retrospective chart review. The experience was mixed: one patient developed mild, corticosteroid-responsive adverse events after a single dose, while the other tolerated multiple doses, had transient neurological improvement and had no adverse events. These cases were descriptive institutional experience, not a controlled comparison.
Why the result remains provisional
The broader evidence was heterogeneous, consisting of case reports, small series, single-arm studies, retrospective cohorts and early-phase trials. Case reports and series were generally adequately reported, but single-arm studies lacked concurrent controls. Retrospective intervention-effect cohorts raised serious concerns about confounding by indication, treatment-selection bias and overlapping therapies.
The safety comparison was also indirect: it used an aggregate systemic comparator rather than a within-study control. Because many analyses reported study-level totals, event rates could not distinguish repeated adverse events in one patient from events spread across several patients. That makes the lower event-level burden harder to interpret as an effect of the intrathecal route.
Taken together, the review supports prospective evaluation of intrathecal immune checkpoint inhibitors as an adjunctive strategy, but it does not settle comparative safety or efficacy. The authors concluded that adjunctive intrathecal treatment was not associated with increased adverse-event risk or high-grade toxicity burden versus systemic treatment alone. The narrow, predominantly melanoma and PD-1-based intrathecal cohort also limits how widely the findings can be applied.
Paper data and sources
Original title: Intrathecal immune checkpoint inhibitors as adjunctive therapy for leptomeningeal disease: A case series and systematic review.
Authors: Manav Daftari, Christian K Ramsoomair, Anurag Aka et al.
Journal/Repository: Neurosurgical review
Status: Peer-reviewed
First online: 2026-08-21
DOI: 10.1007/s10143-026-04454-z
Original paper · Full text