1. Introduction
Chimeric antigen receptor T-cell (CAR-T) therapy has revolutionized the treatment of relapsed and refractory hematological malignancies, demonstrating significantly superior efficacy compared with traditional salvage regimens [ 1 , 2 , 3 ]. Randomized clinical trials, including ZUMA-7 and TRANSFORM, have confirmed the benefits of CAR-T in the treatment of diffuse large B-cell lymphoma, challenging the traditional role of autologous stem cell transplantation (ASCT) [ 4 , 5 ].
Despite its therapeutic success, CAR-T is associated with specific toxicity. Immune cell-associated neurotoxicity syndrome (ICANS) occurs in a significant proportion of patients and is usually reversible [ 6 , 7 ]. However, emerging evidence suggests that cognitive impairment may persist beyond the acute phase, as evidenced by patient surveys and experimental studies demonstrating long-term neuroinflammatory changes [ 8 , 9 , 10 ]. In comparison, ASCT has well-described long-term complications, including cognitive dysfunction, but systematic risk comparisons between CAR-T and ASCT are lacking [ 11 , 12 ].
To fill this gap, we conducted a large cohort study using propensity score matching to compare the incidence and time course of cognitive impairment and associated neurological complications in patients after CAR-T compared with ASCT. The results have practical implications for treatment selection, patient counseling, and long-term care planning for oncohematological patients.
2. Materials and Methods
2.1. Study Design and Data Source
We conducted a retrospective cohort study using the TriNetX US Collaborative Network, which contains anonymized electronic medical records from over 100 healthcare organizations [ 13 ]. The study period extended from January 2014 to April 2025. The study was reported in accordance with the STROBE statement and its RECORD extension for routinely collected health data [ 14 , 15 ]. Institutional ethics approval was not required, as only anonymized, HIPAA-compliant aggregate data were analyzed. The study protocol was retrospectively deposited on the Open Science Framework (OSF DOI: 10.17605/OSF.IO/2BXR7; https://osf.io/2bxr7 ; accessed 25 May 2026) as a transparency record of the prespecified cohort and outcome definitions, covariates, and analysis plan, and to tag explicitly the analyses added in response to peer review; the absence of prospective registration is discussed in Section Limitations.
2.2. Patient Identification
CAR-T Cohort. Patients were identified using drug names and Healthcare Common Procedure Coding System (HCPCS) codes for all FDA-approved CAR-T products: tisagenlecleucel (Kymriah), axicabtagene ciloleucel (Yescarta), lisocabtagene maraleucel (Breyanzi), brexucabtagene autoleucel (Tecartus), idecabtagene vicleucel (Abecma), and ciltacabtagene autoleucel (Carvykti). The index date was the date of CAR-T infusion. Complete codes are provided in Supplementary S2 .
ASCT Cohort. Patients were identified using CPT code 38241 and ICD-10-PCS code 30233G0. The index date was the date of stem cell infusion. To address calendar-era differences in supportive care, coding practice, and patient selection, ASCT recipients were restricted to an index date on or after August 2017 (the first FDA CAR-T approval), yielding a contemporaneous comparator ( Figure 1 ) [ 16 ].
2.3. Eligibility Criteria
Inclusion criteria: age ≥ 18 years at index date; diagnosis of lymphoma or multiple myeloma; receipt of CAR-T therapy or ASCT during the study period; availability of records for at least 6 months before index; and at least 30 days of follow-up after index.
Exclusion criteria: previous cognitive impairment, dementia, or related conditions within 1 year before index; incomplete data; ASCT index date before August 2017; and receipt of both CAR-T and ASCT during the study period.
2.4. Outcomes
Primary Outcome. Composite cognitive impairment, defined as the first post-index occurrence of any of the following ICD-10 codes: R41.81, R41.89, R41.3, R41.840, R41.82, F03.90, and R45.81. These codes were pooled because individual cognitive codes are uncommon and because clinical cognitive impairment manifests heterogeneously across memory, attention, and executive domains; the composite increases statistical power while individual-component analyses confirm that the signal is not driven by any single code. A prespecified sensitivity definition removed R41.82 (altered mental status)—the least specific component—to test specificity for chronic impairment. Delirium and acute confusional states were excluded from the primary composite to maintain specificity for cognitive rather than encephalopathic outcomes.
Secondary Outcomes. Neurological dysfunction, individual neurological complications, mood and stress-related disorders, functional outcomes (falls, mobility), secondary malignancies (acute myeloid leukemia, myelodysplastic syndrome, secondary T-cell lymphoma), coagulation abnormality, venous thromboembolism, and negative control outcomes (gallstones, kidney stones, appendicitis). The negative controls were chosen because they are common enough in this oncology population to provide statistical power, have no plausible inflammatory, cytotoxic, or neuro-immune pathway shared with either CAR-T or ASCT, and are typically detected through acute symptomatic presentation rather than active surveillance, so that equal incidence between arms supports absence of differential surveillance, detection, or coding bias. Complete ICD-10 definitions are provided in Supplementary Table S1 .
2.5. Propensity Score Matching
Patients were matched 1:1 using a nearest-neighbor algorithm without replacement (caliper 0.2 × SD of the logit of the propensity score). Covariates included demographics (age, sex, race), primary diagnosis, central nervous system involvement, comorbidities, mental health conditions, prior and concomitant disease-directed treatments (systemic corticosteroids, rituximab, bendamustine, daratumumab, lenalidomide, carfilzomib, bortezomib, polatuzumab vedotin, brentuximab vedotin, intrathecal chemotherapy, antineoplastic radiation therapy), and laboratory values (body mass index; lactate dehydrogenase [LDH]). LDH was modeled as a categorical variable (≤250, 251–500, ≥500 U/L) rather than continuous, because extreme values produced residual imbalance (continuous LDH SMD 0.141); categorization removed this imbalance. Balance was assessed by standardized mean differences (SMD < 0.1; Table 1 ). Full details are in Supplementary S4 .
2.6. Statistical Analysis
Baseline characteristics were compared using chi-square and t -tests. Cumulative incidence was estimated as 1−Kaplan–Meier survival and compared with log-rank tests; Cox proportional-hazards models provided hazard ratios (HR) with 95% confidence intervals (CI). Cumulative incidence and the cumulative (index-to-landmark) hazard ratio were reported at fixed landmarks ( Table 2 a). The proportional-hazards assumption was evaluated with the Schoenfeld residual global test; because it was violated, the single Cox HR was retained only as a time-averaged summary, and the restricted mean survival time (RMST) difference in cognitive-impairment-free days at 1, 2, and 3 years was used as the principal, assumption-free effect measure ( Table 2 c) [ 17 , 18 , 19 , 20 ]. RMST and its variance were derived from the Kaplan–Meier estimates, with patient-level risk sets reconstructed by the method of Guyot et al. [ 20 ]. Patients were censored at death, loss to follow-up (>180 days without an encounter), or end of study (1 April 2025). To address delayed ICANS, landmark analyses re-anchored follow-up at Day 30 (primary) and Day 90 ( Table 2 b); analyses excluding events within 7, 14, and 21 days are reported in Supplementary Table S2 [ 16 , 21 , 22 ]. Subgroup analyses were stratified by age, sex, race, diagnosis, and CAR-T product. E-values were calculated to assess robustness to unmeasured confounding [ 23 , 24 ]. All tests were two-sided with α = 0.05.
3. Results
3.1. Baseline Characteristics
Of 37,830 patients meeting inclusion criteria in the TriNetX network, 3067 CAR-T therapy patients were successfully matched to 3067 ASCT patients ( Figure 1 ). Groups were well-balanced in demographics, diagnoses, comorbidities, prior treatments, and baseline laboratory values (all SMD < 0.1; Table 1 ). The median age was 60 years, and 61% were male. The most common diagnoses were diffuse large B-cell lymphoma (57%) and multiple myeloma (33%). The median follow-up was 634 days for CAR-T and 713 days for ASCT, reflecting the more recent introduction of CAR-T.
3.2. Primary Outcome: Cognitive Impairment
CAR-T patients had a higher cognitive impairment risk compared with ASCT recipients (HR 1.58; 95% CI 1.39–1.80; p < 0.001; E-value 2.54). The cumulative incidence curves showed an early and sustained difference between groups ( Figure 2 ): at 180 days, the incidence was 13.7% versus 6.9% (absolute difference 6.8%), and at 1000 days 21.3% versus 14.5% (absolute difference 6.8%) ( Table 2 a). Because the proportional-hazards assumption was violated (Schoenfeld p < 0.001), the difference was summarized using RMST: CAR-T recipients remained free of cognitive impairment for 24.8 fewer days at 1 year (95% CI −30.1 to −19.6) and 53.4 fewer days at 2 years (−65.2 to −41.6; both p < 0.001), with the gap widening over time ( Table 2 c; Figure S2 ).
3.3. Time-Based Analyses
The highest risk was observed during the first 30 days (HR 4.22; 95% CI 3.23–5.53; E-value 7.91) and remained significant at 90 days (HR 2.65), 180 days (HR 2.33), 500 days (HR 1.96), 1000 days (HR 1.84), 1500 days (HR 1.79), 2000 days (HR 1.77), and 2500 days (HR 1.76; all p < 0.001) ( Table 2 ).
3.4. Sensitivity Analyses
In immortal-time-corrected landmark analyses re-anchoring follow-up beyond the acute ICANS window, the association persisted at the Day 30 landmark (HR 1.30; 95% CI 1.10–1.52) and the Day 90 landmark (HR 1.25; 95% CI 1.05–1.48) ( Table 2 b). Simpler analyses excluding events within 7, 14, and 21 days were concordant (HR 1.70, 1.43, and 1.33, respectively; all p < 0.001; Table S2 ). The persistence of a significant association beyond the acute window indicates that the elevated risk is not attributable solely to acute ICANS.
3.5. Secondary Outcomes
Cognitive Outcomes. CAR-T recipients had a higher risk of overall cognitive impairment (HR 1.58; 95% CI 1.39–1.80; p < 0.001), which remained significant after removing altered mental status (R41.82) from the composite (HR 1.61; 95% CI 1.41–1.84), with selective increases in memory impairment (HR 1.38; 95% CI 1.04–1.84; p = 0.024) and cognitive dysfunction (HR 1.90; 95% CI 1.37–2.63; p < 0.001). Dementia diagnoses were not significantly increased ( Table 3 ).
Neurological Dysfunction. Overall neurological dysfunction was 69% higher in the CAR-T group (HR 1.69; 95% CI 1.50–1.89; p < 0.001; E-value 2.76). Encephalopathy was the most common complication (HR 2.04; 95% CI 1.73–2.40; p < 0.001). Delirium (HR 1.66; 95% CI 1.25–2.21; p < 0.001) and seizures (HR 1.62; 95% CI 1.25–2.11; p < 0.001) were also significantly more common ( Figure 3 , Table 3 ).
Mood and Stress-Related Disorders. Results showed a 17% higher risk in the CAR-T group (HR 1.17; 95% CI 1.08–1.27; p < 0.001). Sleep disorders showed the most pronounced difference (HR 1.29; 95% CI 1.17–1.43; p < 0.001). The incidence of depressive episodes and anxiety disorders was similar between groups.
Hematologic Outcomes. CAR-T recipients had a significantly lower risk of secondary acute myeloid leukemia (HR 0.46; 95% CI 0.38–0.55; p < 0.001; E-value 3.80) and secondary T-cell lymphoma (HR 0.16; 95% CI 0.05–0.55; p = 0.003). No significant difference was observed in myelodysplastic syndrome (HR 0.94; p = 0.62). Coagulation abnormality was more common (HR 1.31; p = 0.028), whereas venous thromboembolism did not differ (HR 1.01; p = 0.95) ( Table 3 , Figure 3 ).
Functional Outcomes. Fall-related injuries were more common in CAR-T patients (HR 1.18; 95% CI 1.03–1.35; p = 0.016). Mobility impairment was not significantly different (HR 0.94; 95% CI 0.75–1.17; p = 0.57).
3.6. Subgroup Analysis
Subgroup hazard ratios remained stable across demographic and clinical categories ( Figure 4 ). The association between CAR-T therapy and cognitive impairment did not differ significantly by age (HR 1.80 for 18–65 years; HR 1.92 for >65 years), gender (HR 1.76 for female; HR 1.91 for male), race, disease type, or CAR-T product type, indicating a consistent association across all subgroups studied.
3.7. Negative Control Analysis
To assess specificity and residual bias, we examined negative control outcomes theoretically unrelated to treatment choice. Analyses of gallstones (HR 1.18; 95% CI 0.96–1.46; p = 0.12), kidney stones (HR 1.14; 95% CI 0.90–1.45; p = 0.27), and appendicitis (HR 0.92; 95% CI 0.50–1.68; p = 0.79) showed no significant differences, supporting the validity of the observed significant associations.
4. Discussion
This large propensity score-matched study found that CAR-T therapy is associated with a higher risk of long-term cognitive impairment compared with ASCT. The risk was highest in the first month, consistent with acute ICANS. However, even in immortal-time-corrected landmark analyses and when early events were excluded, the increased risk remained throughout follow-up exceeding 2500 days, suggesting both acute and chronic components. Expressed as RMST, CAR-T recipients lost on average roughly 25 and 53 cognitive-impairment-free days over the first 1 and 2 years—an absolute, time-aligned measure that is clinically interpretable in addition to being statistically significant [ 17 , 19 ].
Preclinical studies offer hypotheses that are consistent with these associations: acute ICANS involves cytokine-mediated blood–brain barrier disruption and endothelial activation [ 25 ], and animal models show persistent microglial reactivity, oligodendrocyte dysfunction, and impaired hippocampal neurogenesis after CAR-T [ 26 ]. However, our study includes no neuroimaging, biomarker, or pathological data and therefore cannot establish a mechanism. The findings should be interpreted as associations; we cannot distinguish chronic CAR-T–specific neurotoxicity from persistent ICANS sequelae, differential detection or surveillance, or survivorship effects. Preclinical reports such as Geraghty et al. [ 26 ] are cited only as hypotheses consistent with our observations, not as mechanistic evidence from the present data.
CAR-T products differ in efficacy and toxicity profiles. In our subgroup analysis the association was directionally present for all products, with numerically higher point estimates for ciltacabtagene autoleucel and lisocabtagene maraleucel, although confidence intervals overlapped ( Figure 3 ). Product type is strongly collinear with disease (axicabtagene-, lisocabtagene-, and tisagenlecleucel for lymphoma; idecabtagene- and ciltacabtagene-vicleucel for myeloma; brexucabtagene autoleucel for mantle-cell/ALL), so product-specific effects cannot be fully separated from disease [ 27 , 28 ]. ASCT was selected as an active comparator because it is the most clinically relevant alternative—pivotal trials (ZUMA-7, TRANSFORM) compared CAR-T against salvage chemotherapy followed by ASCT—and because an active comparator reduces confounding by indication relative to an untreated reference [ 4 , 5 ]. Nonetheless, channeling bias is possible: in the CAR-T era, CAR-T is often used after ASCT failure or as second-line therapy, so the two groups may differ in ways residual to matching. Restricting ASCT to the contemporaneous era partly mitigates, but does not eliminate, this concern [ 16 ].
Our study found lower rates of secondary acute myeloid leukemia in CAR-T recipients compared with ASCT recipients, likely reflecting differences in conditioning regimens. These findings provide some reassurance regarding the hematologic safety of CAR-T, although continued follow-up for late events is needed [ 29 ].
Limitations
Limitations include reliance on diagnostic codes rather than formal neuropsychological testing (e.g., MoCA, MMSE, or the ICCTF cognitive battery), which may misclassify cognitive outcomes; because ascertainment was identical in both arms within the same network, such misclassification is expected to be non-differential and to bias estimates toward the null, implying the true association is at least as large as observed. The platform did not permit a confirmed-diagnosis definition requiring repeated, temporally separated coding, which would further enhance specificity [ 30 ]. Important confounders that TriNetX cannot capture—education and cognitive reserve, socioeconomic status, treatment preference, and performance status—were unmeasured; we quantified robustness to such unmeasured confounding using E-values [ 23 ]. Other limitations include potential residual confounding despite excellent post-matching balance and shorter follow-up in the CAR-T cohort, which we partly addressed with time-aligned RMST and landmark analyses. We also note that the study was not prospectively registered before data extraction; the protocol has been retrospectively deposited on the Open Science Framework as a transparency measure, but retrospective deposition does not provide the same safeguard against outcome switching or selective reporting that prospective registration would. Nevertheless, the large sample, rigorous methodology, negative-control outcomes, and time-stratified analyses enhance validity.
5. Conclusions
CAR-T therapy is associated with a higher risk of cognitive impairment that extends well beyond the acute phase. Because these are observational, code-based associations rather than mechanistic findings, they highlight the need for prospective studies with standardized cognitive testing, biomarker collection, and interventional trials targeting neuroinflammation to reduce long-term cognitive risks in CAR-T recipients [ 26 , 31 ].
Acknowledgments
The authors thank the TriNetX research network for providing access to the data used in this study.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Supplementary Materials
The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/diagnostics16121862/s1 , STROBE Checklist; Supplementary Methods (S1–S6), S1. Data Source and Study Population, S2. Detailed Cohort Definitions, S3. Outcome Definitions and ICD-10 Codes, S4. Propensity Score Matching Methodology, S5. Statistical Analysis Plan, S6. Sensitivity, Robustness, and Negative-Control Analyses; Figure S1: Graphical representation of study design; Figure S2: Restricted Mean Survival Time (RMST) for Cognitive-Impairment-Free Survival; Table S1: Complete ICD-10 code definitions for all outcomes; Table S2: Sensitivity Analyses Excluding Events Within the Acute ICANS Windows.
Author Contributions
Conceptualization, A.B., P.-H.C., H.-J.J. and C.-H.L.; methodology, P.-H.C., W.-C.C., L.-T.K. and T.Y.-J.H.; formal analysis, A.B., H.-Y.C. and L.-T.K.; data curation, A.B., P.-H.C. and W.-C.C.; validation, W.-C.C., H.-Y.C., T.Y.-J.H. and M.-S.D.; writing—original draft preparation, A.B. and P.-H.C.; writing—review and editing, H.-Y.C., L.-T.K., T.Y.-J.H., M.-S.D., H.-J.J. and C.-H.L.; supervision, M.-S.D., H.-J.J. and C.-H.L.; project administration, C.-H.L., A.B. and P.-H.C. contributed equally to this work as co-first authors. H.-J.J. and C.-H.L. contributed equally as co-corresponding authors. All authors have read and agreed to the published version of the manuscript.
Institutional Review Board Statement
Ethical review and approval were waived for this study because it used fully de-identified, aggregate data from the TriNetX Research Network in compliance with the Health Insurance Portability and Accountability Act (HIPAA) and was therefore exempt from institutional review board oversight.
Informed Consent Statement
Patient consent was waived because the study used only de-identified data obtained from the TriNetX Research Network, in accordance with HIPAA-compliant procedures.
Data Availability Statement
The data presented in this study are available from the TriNetX Research Network ( https://trinetx.com , accessed on 14 June 2026). Access to TriNetX data requires institutional membership and compliance with the network’s data use agreements. Summary data supporting the findings of this study are available from the corresponding authors upon reasonable request.
Conflicts of Interest
The authors declare no conflicts of interest.
References
1. June C.H. O’Connor R.S. Kawalekar O.U. Ghassemi S. Milone M.C. CAR T cell immunotherapy for human cancer Science 2018 359 1361 1365 10.1126/science.aar6711 29567707
2. Munshi N.C. Anderson L.D. Jr. Shah N. Madduri D. Berdeja J. Lonial S. Raje N. Lin Y. Siegel D. Oriol A. Idecabtagene vicleucel in relapsed and refractory multiple myeloma N. Engl. J. Med. 2021 384 705 716 10.1056/NEJMoa2024850 33626253
3. Locke F.L. Miklos D.B. Jacobson C.A. Perales M.A. Kersten M.J. Oluwole O.O. Ghobadi A. Rapoport A.P. McGuirk J. Pagel J.M. Axicabtagene ciloleucel as second-line therapy for large B-cell lymphoma N. Engl. J. Med. 2022 386 640 654 10.1056/NEJMoa2116133 34891224
4. Westin J.R. Oluwole O.O. Kersten M.J. Miklos D.B. Perales M.A. Ghobadi A. Rapoport A.P. Sureda A. Jacobson C.A. Farooq U. Survival with axicabtagene ciloleucel in large B-cell lymphoma N. Engl. J. Med. 2023 389 148 157 10.1056/NEJMoa2301665 37272527
5. Kamdar M. Solomon S.R. Arnason J. Johnston P.B. Glass B. Bachanova V. Ibrahimi S. Mielke S. Mutsaers P. Hernandez-Ilizaliturri F. Lisocabtagene maraleucel versus standard of care with salvage chemotherapy followed by autologous stem cell transplantation as second-line treatment in patients with relapsed or refractory large B-cell lymphoma (TRANSFORM): Results from an interim analysis of an open-label, randomised, phase 3 trial Lancet 2022 399 2294 2308 35717989 10.1016/S0140-6736(22)00662-6
6. Lee D.W. Santomasso B.D. Locke F.L. Ghobadi A. Turtle C.J. Brudno J.N. Maus M.V. Park J.H. Mead E. Pavletic S. ASTCT consensus grading for cytokine release syndrome and neurologic toxicity associated with immune effector cells Biol. Blood Marrow Transplant. 2019 25 625 638 10.1016/j.bbmt.2018.12.758 30592986 PMC12180426
7. Santomasso B.D. Park J.H. Salloum D. Riviere I. Flynn J. Mead E. Halton E. Wang X. Senechal B. Purdon T. Clinical and biological correlates of neurotoxicity associated with CAR T-cell therapy in patients with B-cell acute lymphoblastic leukemia Cancer Discov. 2018 8 958 971 10.1158/2159-8290.CD-17-1319 29880584 PMC6385599
8. Gust J. Hay K.A. Hanafi L.A. Li D. Myerson D. Gonzalez-Cuyar L.F. Yeung C. Liles W.C. Wurfel M. Lopez J.A. Endothelial activation and blood-brain barrier disruption in neurotoxicity after adoptive immunotherapy with CD19 CAR-T cells Cancer Discov. 2017 7 1404 1419 10.1158/2159-8290.CD-17-0698 29025771 PMC5718945
9. Ruark J. Mullane E. Cleary N. Cordeiro A. Bezerra E.D. Wu V. Voutsinas J. Shaw B.E. Flynn K.E. Lee S.J. Patient-reported neuropsychiatric outcomes of long-term survivors after chimeric antigen receptor T cell therapy Biol. Blood Marrow Transplant. 2020 26 34 43 10.1016/j.bbmt.2019.09.037 31605820 PMC6951812
10. Karschnia P. Jordan J.T. Forst D.A. Arrillaga-Romany I.C. Batchelor T.T. Baehring J.M. Clement N.F. Gonzalez Castro L.N. Herlopian A. Maus M.V. Clinical presentation, management, and biomarkers of neurotoxicity after adoptive immunotherapy with CAR T cells Blood 2019 133 2212 2221 10.1182/blood-2018-12-893396 30808634
11. Sharafeldin N. Bosworth A. Patel S.K. Chen Y. Morse E. Mather M. Sun C.L. Francisco L. Forman S.J. Wong F.L. Cognitive functioning after hematopoietic cell transplantation for hematologic malignancy: Results from a prospective longitudinal study J. Clin Oncol. 2018 36 463 475 10.1200/JCO.2017.74.2270 29252122
12. Kelly D.L. Buchbinder D. Duarte R.F. Auletta J.J. Bhatt N. Byrne M. DeFilipp Z. Gabriel M. Mahindra A. Norkin M. Neurocognitive Dysfunction in Hematopoietic Cell Transplant Recipients: Expert Review from the Late Effects and Quality of Life Working Committee of the Center for International Blood and Marrow Transplant Research and Complications and Quality of Life Working Party of the European Society for Blood and Marrow Transplantation Biol. Blood Marrow Transplant. 2018 24 228 241 10.1016/j.bbmt.2017.09.004 28939455 PMC5768142
13. TriNetX. TriNetX Research Network Available online: https://trinetx.com (accessed on 1 April 2025)
14. von Elm E. Altman D.G. Egger M. Pocock S.J. Gøtzsche P.C. Vandenbroucke J.P. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies Lancet 2007 370 1453 1457 10.1016/S0140-6736(07)61602-X 18064739
15. Benchimol E.I. Smeeth L. Guttmann A. Harron K. Moher D. Petersen I. Sørensen H.T. von Elm E. Langan S.M. RECORD Working Committee The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement PLoS Med. 2015 12 e1001885 10.1371/journal.pmed.1001885 26440803 PMC4595218
16. Suissa S. Immortal time bias in pharmaco-epidemiology Am. J. Epidemiol. 2008 167 492 499 10.1093/aje/kwm324 18056625
17. Uno H. Claggett B. Tian L. Inoue E. Gallo P. Miyata T. Schrag D. Takeuchi M. Uyama Y. Zhao L. Moving beyond the hazard ratio in quantifying the between-group difference in survival analysis J. Clin. Oncol. 2014 32 2380 2385 10.1200/JCO.2014.55.2208 24982461 PMC4105489
18. Royston P. Parmar M.K.B. Restricted mean survival time: An alternative to the hazard ratio for the design and analysis of randomized trials with a time-to-event outcome BMC Med. Res. Methodol. 2013 13 152 10.1186/1471-2288-13-152 24314264 PMC3922847
19. Zhao L. Claggett B. Tian L. Uno H. Pfeffer M.A. Solomon S.D. Trippa L. Wei L.J. On the restricted mean survival time curve in survival analysis Biometrics 2016 72 215 221 10.1111/biom.12384 26302239 PMC5114026
20. Guyot P. Ades A.E. Ouwens M.J.N.M. Welton N.J. Enhanced secondary analysis of survival data: Reconstructing the data from published Kaplan-Meier survival curves BMC Med. Res. Methodol. 2012 12 9 10.1186/1471-2288-12-9 22297116 PMC3313891
21. Mi X. Hammill B.G. Curtis L.H. Lai E.C. Setoguchi S. Use of the landmark method to address immortal person-time bias in comparative effectiveness research: A simulation study Stat. Med. 2016 35 4824 4836 10.1002/sim.7019 27350312
22. Hernán M.A. Sauer B.C. Hernández-Díaz S. Platt R. Shrier I. Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses J. Clin. Epidemiol. 2016 79 70 75 27237061 10.1016/j.jclinepi.2016.04.014 PMC5124536
23. VanderWeele T.J. Ding P. Sensitivity Analysis in Observational Research: Introducing the E-Value Ann. Intern. Med. 2017 167 268 274 10.7326/M16-2607 28693043
24. Mathur M.B. Ding P. Riddell C.A. VanderWeele T.J. Web Site and R Package for Computing E-values Epidemiology 2018 29 e45 e47 10.1097/ede.0000000000000864 29912013 PMC6066405
25. Shimabukuro-Vornhagen A. Gödel P. Subklewe M. Stemmler H.J. Schlößer H.A. Schlaak M. Kochanek M. Böll B. von Bergwelt-Baildon M.S. Cytokine release syndrome J. Immunother. Cancer 2018 6 56 10.1186/s40425-018-0343-9 29907163 PMC6003181
26. Geraghty A.C. Acosta-Alvarez L. Rotiroti M.C. Dutton S. O’Dea M.R. Kim W. Trivedi V. Ravi K. Mancusi R. Ni L. Immunotherapy-related cognitive impairment after CAR T cell therapy in mice Cell 2025 188 3238 3258.e25 10.1016/j.cell.2025.03.041 40359942 PMC12176077
27. Cohen A.D. Parekh S. Santomasso B.D. Gallego Perez-Larraya J. van de Donk N.W.C.J. Arnulf B. Mateos M.V. Lendvai N. Jackson C.C. De Braganca K.C. Incidence and management of CAR-T neurotoxicity in patients with multiple myeloma treated with ciltacabtagene autoleucel in CARTITUDE studies Blood Cancer J. 2022 12 32 10.1038/s41408-022-00629-1 35210399 PMC8873238
28. Abramson J.S. Solomon S.R. Arnason J. Johnston P.B. Glass B. Bachanova V. Ibrahimi S. Mielke S. Mutsaers P. Hernandez-Ilizaliturri F. Lisocabtagene maraleucel as second-line therapy for large B-cell lymphoma: Primary analysis of the phase 3 TRANSFORM study Blood 2023 141 1675 1684 36542826 10.1182/blood.2022018730 PMC10646768
29. Cordeiro A. Bezerra E.D. Hirayama A.V. Yeung C.C.S. Wu V. Voutsinas J. Liu Q. Acharya U.H. Pender B.S. Hawkins R.M. Late Events after Treatment with CD19-Targeted Chimeric Antigen Receptor Modified T Cells Biol. Blood Marrow Transplant. 2020 26 26 33 10.1016/j.bbmt.2019.08.003 31419568 PMC6953906
30. Wilkinson T. Ly A. Schnier C. Rannikmäe K. Bush K. Brayne C. Quinn T.J. Sudlow C.L.M. UK Biobank Neurodegenerative Outcomes Group Dementias Platform UK Identifying dementia cases with routinely collected health data: A systematic review Alzheimer’s Dement. 2018 14 1038 1051 10.1016/j.jalz.2018.02.016 29621480 PMC6105076
31. Sterner R.M. Sakemura R. Cox M.J. Yang N. Khadka R.H. Forsman C.L. Hansen M.J. Jin F. Ayasoufi K. Hefazi M. GM-CSF inhibition reduces cytokine release syndrome and neuroinflammation but enhances CAR-T cell function in xenografts Blood 2019 133 697 709 10.1182/blood-2018-10-881722 30463995 PMC6376281