CRS, ICANS, Hospitalization and Monitoring Requirements for CAR-T and T-Cell Engager Trials in Autoimmune Disease

CRS, ICANS, hospitalization and monitoring in CAR-T and T-cell engager trials for autoimmune disease

CAR-T cells and CD3-redirecting T-cell engagers (TCEs) are moving from oncology into lupus, systemic sclerosis, myositis, vasculitis and neuroimmunology. Neither FDA nor EMA has yet published CAR-T or TCE guidance written specifically for autoimmune indications. Sponsors therefore work from three sources: oncology-derived CAR-T and first-in-human frameworks, class-wide safety actions on marketed CAR-T products, and the protocol conventions forming in the autoimmune trials themselves. This article sets out what each source requires or recommends on cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity syndrome (ICANS), hospitalization and follow-up.

Key points

• FDA's final CAR-T guidance (January 2024) covers oncology only. For non-oncology uses, including autoimmune disease, FDA tells sponsors to consult CBER's Office of Therapeutic Products before IND submission, for example at a pre-IND meeting ²⁵⁹⁴⁷.
• FDA and EMA both expect a prespecified toxicity grading system, CRS and neurotoxicity management algorithms, DLT definitions, stopping rules and staggered dosing in first-in-human cell therapy trials ^278220172. Neither agency sets a fixed inpatient period for trials. EMA says hospitalization duration should be planned in relation to expected serious adverse events ¹⁷⁴.
• For integrating-vector CAR-T products, both agencies expect 15 years of follow-up ⁴²¹⁷³. After the T-cell malignancy signal, FDA and EMA/PRAC added lifelong monitoring for secondary malignancies. FDA's requirement explicitly includes clinical trial participants ²¹⁰²¹⁹.
• Autoimmune CAR-T protocols mostly use ASTCT consensus grading for CRS and ICANS, 28-day DLT windows and a period of inpatient observation after infusion (about 10 days in at least one systemic sclerosis trial) ^{22923024510363232}.
• TCE trials in autoimmune disease are earlier-stage. Their registry records rarely state hospitalization or CRS monitoring rules. Marketed oncology TCE labels, with REMS programs and 24 to 48 hour hospitalization after step-up doses, remain the main regulatory benchmark ²³³⁷²⁴.

FDA expectations for CAR-T clinical protocols

Scope and population selection

The final guidance Considerations for the Development of Chimeric Antigen Receptor (CAR) T Cell Products (January 31, 2024) addresses hematologic malignancies and solid tumors. It does not give CAR-T recommendations for non-oncology indications and refers those sponsors to CBER's Office of Therapeutic Products for pre-IND discussion ²⁵⁹⁴⁷. On population selection, FDA says the potentially life-threatening or fatal CRS and neurologic toxicities should be weighed against potential benefit, disease stage and available therapies. The guidance cautions that the unknown benefits of a first-in-human CAR-T product may not justify its risks in early-stage disease that has available therapies ⁵⁰. This reasoning helps explain why current autoimmune CAR-T trials enroll refractory patients.

Toxicity grading, management algorithms and DLTs

FDA recommends that the protocol include:

• a toxicity grading system to guide dose escalation and patient management ²⁷⁸
• detailed management algorithms for CRS and neurologic toxicities, including a sign- and symptom-based plan for anti-cytokine therapy such as tocilizumab ²⁷⁸
• well-defined DLTs, stopping rules and attribution criteria ²⁷⁸

FDA recommends NCI CTCAE for general toxicity grading. Sponsors should consider standardized consensus criteria for CRS and neurotoxicity and justify their choice. The guidance cites ASTCT consensus grading ²⁷⁸²⁸⁷. FDA also says psychiatric toxicities should be assessed, reported and managed ²⁷⁸. Pre- and post-dose C-reactive protein and cytokines (TNF-α, IL-2, IL-6, IL-10, IFN-γ) are named as potentially informative CRS markers. They can also help distinguish CRS from anaphylaxis ²⁷⁷²⁷⁹.

The guidance gives these example DLTs:

FDA recommends defining CAR-T DLTs regardless of attribution to the product unless another cause is clearly identified, because disease manifestations and concomitant treatment can confound attribution ³⁴⁰⁴⁹. Example stopping rules include more than two grade 4 CRS events in a first-in-human CAR-T trial, or any death within 30 days after administration ⁴⁹.

Staggering and dose escalation

When there is no prior human experience with the product or a related product, FDA recommends staggered treatment. Staggering may apply within the first cohort and between cohorts. The interval should be long enough to detect acute and subacute adverse events before the next subject is dosed or the dose is escalated ^220329335246. Half-log (about three-fold) increments are common for CAR-T. FDA considers single-patient cohorts and intra-patient escalation typically unsuitable for first-in-human CAR-T studies ³⁴⁰³³⁵.

Hospitalization and proximity

None of the FDA CAR-T guidance text reviewed specifies an inpatient duration or a proximity-to-site requirement for trial subjects. FDA asks instead for a detailed monitoring plan based on nonclinical findings and prior clinical experience with the product or related products ²⁴⁶.

Long-term follow-up

For CAR-T products with an integrated transgene, FDA recommends following subjects for 15 years after treatment. The plan should be funded to continue even if the sponsor ceases operations or transfers or withdraws the IND ⁴². For non-integrating vectors, long-term follow-up generally is not needed ⁴⁴. Protocols should describe how CAR-T persistence will be measured ⁴⁹. FDA's gene therapy long-term follow-up recommendations apply to integrating vectors. They call for PCR testing for vector sequences at least every 6 months for the first 5 years, then at least annually for 10 years. Integration-site analysis is recommended if oligoclonality or monoclonality emerges ³⁵⁰. For retroviral-vector products, FDA's RCR guidance recommends testing patient samples for replication-competent retrovirus at 3, 6 and 12 months, then annually for up to 15 years. If all of a patient's post-treatment samples are negative during the first year, yearly sample collection may stop for that patient, and a yearly review of medical history will generally be sufficient ⁴⁸.

FDA class actions on marketed CAR-T products

The following actions apply to oncology labels, but they shape what IRBs, investigators and FDA reviewers expect in autoimmune programs.

• T-cell malignancy boxed warning (2024). On April 18, 2024, FDA announced class-wide labeling changes, including a boxed warning, for all approved BCMA- and CD19-directed genetically modified autologous CAR-T products. The changes followed reports of T-cell malignancies, including CAR-positive tumors, that can arise within weeks of infusion ²¹⁰. FDA said patients and clinical trial participants should be monitored lifelong for secondary malignancies. If a new malignancy occurs, the manufacturer should be contacted for instructions on collecting samples for CAR transgene testing ²¹⁰. FDA started the safety labeling change process in January 2024 ²¹¹²¹⁷.
• REMS elimination and shorter monitoring (2025). On June 26, 2025, FDA eliminated the REMS for approved BCMA- and CD19-directed autologous CAR-T products. Treatment sites no longer needed special certification or on-site immediate access to tocilizumab ²²¹²²². Labeling now calls for monitoring for at least 2 weeks after infusion, including daily monitoring for at least 1 week. The required period of staying near a healthcare facility fell from 4 weeks to at least 2 weeks. The driving restriction fell from 8 weeks to 2 weeks ²²¹²²⁶.

The 2025 changes cover approved products. They do not by themselves reduce what a sponsor must justify in an IND for a new CAR-T construct in a new population.

EMA expectations

EMA's CAR-T guidance names CRS and neurotoxicity as the principal acute adverse reactions of CD19-directed CAR-T cells and expects a broader range of events with other targets ¹⁷⁴. Trial protocols should:

• define expected and unexpected adverse events using product-specific nonclinical data and experience with other CAR-T products ¹⁷⁴
• include an algorithm to detect and treat potentially life-threatening toxicities ¹⁷⁴
• plan the duration of hospitalization in relation to expected serious adverse events ¹⁷⁴
• provide a follow-up period long enough to identify late toxicities ¹⁷⁴

Advanced therapies fall outside the formal scope of the first-in-human guideline, but EMA says its risk-mitigation principles still apply to genetically modified cells. These include adequate waiting periods between the first and subsequent patients, and stopping rules ¹⁷². The first-in-human framework flags cytokine release as needing special attention ¹⁷⁷. It expects controlled conditions, possibly including inpatient care, with immediate capacity to manage acute emergencies ¹⁸⁷. For genetically modified cells, EMA recommends 15-year follow-up covering malignancy, infection, immunogenicity or immunosuppression, loss of efficacy and vector-related risks, with these issues reflected in the risk management plan ¹⁷³¹⁸⁰.

On secondary malignancies, PRAC began a signal review in January 2024, when secondary malignancy was already an important potential risk in the RMPs of all six EU-authorised CAR-T products ²¹⁸. PRAC reviewed 38 T-cell malignancy cases among about 42,500 treated patients. Tissue was tested in half of the cases, and seven showed the CAR construct ²¹⁹. The outcomes were updated product information and RMPs, lifelong monitoring for secondary malignancies, and a proposed DHPC ²¹⁹.

What autoimmune CAR-T protocols specify

Registry records vary widely in detail. Many lupus and basket-trial entries give no CRS, ICANS, DLT or hospitalization provisions at all ^110114119. The table below lists trials whose ClinicalTrials.gov records do give specific safety-monitoring terms.

Other patterns in the records:

• Lymphodepletion. Most protocols use fludarabine/cyclophosphamide before a single infusion ⁹¹¹²³. Some investigators are testing regimens without lymphodepletion. One cites the risk of severe infection with fludarabine/cyclophosphamide as its reason ¹²⁵. Another runs lymphodepleted and non-lymphodepleted arms side by side ²⁰⁴. The CRC01 protocol requires infusion to be delayed if lymphodepletion-related adverse events exceed grade 1, including new oxygen requirement, hypotension or active infection within 72 hours ¹²².
• Long-term follow-up. Two records state gene-therapy follow-up of up to 15 years after 52 or 104 weeks of primary follow-up ¹¹⁶¹²⁴. Several others stop at 1 to 2 years in the registry record ¹²²⁷⁷.
• Safety-driven exclusions. Common exclusions include severe or active CNS lupus, recent stroke or seizures, and significant cardiac dysfunction (for example, LVEF below 40% or 45%, or QTcF of 470 ms or more). Some protocols also exclude patients with a contraindication to tocilizumab or other standard CRS/ICANS rescue medications ^{275231239237250}.
• Proximity requirements. None of the autoimmune CAR-T or TCE registry records cited here specifies a residence or travel radius after discharge.

Observed CRS and ICANS in published autoimmune CAR-T data

Published autoimmune data show a milder toxicity profile than oncology. In a Phase I/II basket study of 24 patients with SLE, systemic sclerosis or idiopathic inflammatory myopathy, 18 had CRS: 17 at grade 1 and 1 at grade 2. None had ICANS ²⁵¹. A pooled analysis of 47 SLE patients across 10 studies reported CRS in 87%, all grade 1 to 2, with 9 patients receiving tocilizumab. One patient (about 2%) had grade 1 ICANS on day 8, which resolved with steroids. No ICU admissions were reported ²⁵³.

A newer autoimmune-specific event, local immune effector cell-associated toxicity syndrome (LICATS), occurs in organs previously affected by the autoimmune disease. It has been reported in up to 88% of patients, mostly at grade 1 to 2 ²⁵¹²⁵⁴. In the grading used by one cohort, prolonged or new hospitalization defines grade 3 LICATS and ICU treatment defines grade 4. The same observational study kept patients inpatient for at least 10 days after infusion ²⁵⁴.

Expert guidance for rheumatology CAR-T programs recommends delivery within accredited cellular therapy programs. It calls for confirmed inpatient, telemetry and ICU capacity before activation, and a formal escalation plan covering CRS, ICANS, hematotoxicity and LICATS ²⁵².

T-cell engagers in autoimmune disease

Trial landscape

Registered TCE studies in autoimmune disease are mostly Phase 1 or investigator-initiated:

Across these records, none gives a CRS grading schedule, ICANS assessment schedule, mandatory observation period or premedication regimen, apart from the CC312 priming dose and ASTCT grading ^341012161241. Published summaries describe mainly low-grade, infusion-related cytokine symptoms in early autoimmune TCE experience. No consensus or regulatory TCE-specific recommendations for autoimmune use were identified ²⁵¹.

Regulatory benchmarks from oncology TCE labels and guidance

FDA's bispecific antibody guidance says that for bispecifics with agonistic properties, sponsors should consider a minimally anticipated biological effect level (MABEL) approach to first-in-human dose selection. It cites a retrospective analysis supporting a starting dose at 10% to 30% pharmacological activity for CD3 bispecifics. It also notes that healthy volunteers may be unsuitable for some bispecific trials ¹⁰⁸.

Marketed oncology TCE labels show what FDA has required once CRS and ICANS risk is characterized:

Both Tecvayli and Elrexfio carry boxed warnings for CRS and for neurologic toxicity including ICANS ²⁶³⁹. Autoimmune TCE programs that use lower doses or priming strategies will need to justify any departure from these inpatient conventions with their own exposure-response and safety data.

Implications for regulatory strategy

• Engage early. FDA's CAR-T guidance explicitly routes non-oncology programs to a pre-IND discussion with the Office of Therapeutic Products ⁴⁷. That meeting is the place to agree on grading criteria, DLT windows, staggering intervals and the inpatient period.
• Justify the inpatient period with data. Neither agency prescribes a duration. EMA ties it to expected serious adverse events ¹⁷⁴. Where a duration is stated, inpatient observation is about 10 days in SCLEROCAR and at least 10 days in one observational cohort; other registry records do not state a duration ^63232254. The shorter 2-week proximity requirement for marketed oncology CAR-T products ²²¹ may support proposals for shorter post-discharge monitoring. It is not an IND standard.
• Adopt ASTCT grading and add LICATS. ASTCT grading matches FDA's cited consensus reference ²⁷⁸²⁸⁷ and is now the norm in autoimmune protocols ^229274236. Protocols should also define LICATS so it is distinguished from CRS and from disease flare ²⁵⁴.
• Plan 15-year and lifelong malignancy surveillance. Integrating-vector products need 15-year follow-up under both FDA and EMA recommendations ⁴²¹⁷³. The T-cell malignancy actions add lifelong monitoring and CAR transgene testing of any new malignancy ²¹⁰²¹⁹. A younger autoimmune population makes this surveillance commitment larger.
• For TCEs, borrow from oncology deliberately. Step-up or priming dosing ¹⁶¹, MABEL-based starting doses ¹⁰⁸ and a defined observation period after the first step-up doses ²³³⁷²⁴ are the reference points FDA reviewers will recognize. The autoimmune registry records so far give little detail on these points.

Open questions for further research include how FDA and EMA have handled specific autoimmune CAR-T INDs and CTAs in advice or public assessments, whether outpatient administration has been accepted for any autoimmune CAR-T protocol, and how allogeneic and non-lymphodepleted approaches change the expected monitoring burden.