This report examined pre‑infusion cerebrospinal fluid (CSF proteomics) from patients evaluated before chimeric antigen receptor T‑cell (CAR‑T) therapy to assess whether baseline disease‑related molecular variation influences biomarker interpretation for immune effector cell–associated neurotoxicity syndrome (ICANS).
The authors reasoned that disease‑specific CSF proteomic structure present prior to infusion could confound or inform efforts to develop CSF biomarkers for post‑CAR‑T neurotoxicity risk stratification.
The analysis included CSF samples from 28 patients with diffuse large B‑cell lymphoma (DLBCL) and 9 patients with multiple myeloma (MM). Clinical ICANS events were recorded: ICANS occurred in 7 of the 28 DLBCL patients and in none of the 9 MM patients in this cohort.
Other cohort details such as patient demographics, prior therapies, exact CAR‑T products used, or timing of CSF collection relative to infusion were not reported in the source text and therefore are not included here.
Unsupervised principal component analysis (PCA) of the pre‑infusion CSF proteomic profiles showed substantial overlap between the DLBCL and MM groups, indicating shared variance across diagnoses. However, supervised orthogonal partial least squares discriminant analysis (orthoPLS‑DA) identified significant disease‑associated discrimination between the two diagnostic groups.
The orthoPLS‑DA discrimination was supported by permutation testing, indicating that the observed separation exceeded that expected by chance under the tested model. The source did not provide full methodological parameters (for example, preprocessing steps, normalization methods, or exact permutation counts), so those specifics are not reported here.
Proteins that contributed most to the orthoPLS‑DA separation were reported to be enriched for signatures related to plasma cell biology, extracellular components, and metabolic processes. This enrichment suggests that baseline CSF proteomic differences reflect underlying disease biology and tissue involvement patterns specific to MM versus DLBCL.
The source did not enumerate individual protein names or provide quantitative effect sizes for specific analytes in the public summary; such details would be available in the underlying proteomics dataset, which the authors state can be obtained from the corresponding author upon reasonable request.
When samples were analyzed in a binary framework stratified by ICANS occurrence (ICANS‑positive vs ICANS‑negative), MM cases aligned with the ICANS‑negative group. This is consistent with the observed absence of ICANS events among the 9 MM patients in the cohort.
In a three‑group analysis separating DLBCL ICANS‑positive, DLBCL ICANS‑negative, and MM, MM remained distinct from both DLBCL subgroups. That finding indicates that although MM cases resembled the ICANS‑negative profile in a two‑group comparison, they still exhibited disease‑specific proteomic features that differentiated them from DLBCL regardless of ICANS status.
Taken together, these observations support the conclusion that baseline disease identity imposes molecular structure on pre‑infusion CSF that could influence biomarker signals attributed to neurotoxicity risk.
The principal implication is that development and interpretation of CSF biomarkers for CAR‑T–associated neurotoxicity should account for underlying disease‑related proteomic variation. Biomarker models that do not consider diagnosis‑specific baseline differences may misattribute disease signals to impending ICANS risk or miss signals that are disease‑masked.
Practically, this suggests that biomarker discovery cohorts should include disease‑matched controls or explicitly model diagnosis as a covariate, and that validation should be performed across relevant hematologic malignancies to ensure generalizability.
This retrospective study was approved by the Ethics Committee of Kyushu University (IRB numbers 22213 and 23399) and used an opt‑out informed consent process with anonymized data. The authors state that the proteomics data are available from the corresponding author on reasonable request.
Funding sources declared include JSPS KAKENHI and the Japan Agency for Medical Research and Development (AMED). A competing interest statement notes that Koji Kato has received honoraria and research funding and has advisory relationships with multiple pharmaceutical companies; details reported in the source include Bristol‑Myers Squibb, Chugai, Gilead Sciences, Novartis, Astellas, Daiichi Sankyo, Eisai, Janssen, Nippon‑Shinyaku, and Ono.
Additional methodological details, full protein lists, and quantitative results were not reported in the public summary available from the source; interested readers should contact the corresponding author or consult the deposited dataset as indicated by the authors.
(Posted August 24, 2026.)