What Is ICANS? Understanding CAR-T Therapy's Neurological Side Effect

What This Article Covers

Chimeric antigen receptor (CAR) T-cell therapy has produced some of the deepest, most durable remissions ever seen in relapsed blood cancers, but that same engineered immune assault can turn on the brain. Immune effector cell-associated neurotoxicity syndrome, or ICANS, is the second of CAR-T's two hallmark toxicities alongside cytokine release syndrome (CRS), and it ranges from a few days of word-finding trouble and hand tremor to, rarely, seizures, coma, or fatal brain swelling. This article explains what ICANS looks like, how doctors grade and treat it, how often it actually happens across the FDA-approved CAR-T products, and — the genuinely reassuring headline underneath the frightening vocabulary — why the large majority of cases are low-grade and clear up completely with prompt treatment.

What ICANS Actually Looks Like

ICANS is a clinical diagnosis built from bedside neurologic findings, not a single lab test. The Cleveland Clinic describes the presentation in plain terms: confusion, headache, tremor, difficulty concentrating, slurred or halting speech, and vision changes, with loss of consciousness possible in more severe cases. Clinically, the most common individual symptoms reported in FDA prescribing information are encephalopathy (a catch-all for confusion, disorientation, and diminished alertness), headache, tremor, dizziness, aphasia (difficulty producing or understanding language), delirium, and insomnia — each showing up in roughly 10% to 50% of treated patients depending on the product and disease being treated. Less common but more dangerous manifestations include focal motor weakness, dysarthria, lethargy progressing to obtundation, seizures (including nonconvulsive status epilepticus that can be missed without EEG monitoring), and, in the most severe cases, elevated intracranial pressure with diffuse cerebral edema.

Aphasia deserves particular mention because it is one of the earliest and most specific warning signs clinicians watch for. A patient who was talking normally an hour ago and now struggles to name simple objects or produce fluent sentences is displaying a textbook early ICANS finding, which is precisely why the standardized bedside grading tool leans so heavily on language and writing tasks.

The underlying biology is related to, but distinct from, cytokine release syndrome. CRS is driven by systemic cytokine release — chiefly interleukin-6 (IL-6) and interferon-gamma — as activated CAR T-cells proliferate and kill tumor cells, producing fevers, hypotension, and hypoxia. ICANS shares that inflammatory trigger but involves a more specific insult: elevated cytokines and CAR T-cell trafficking appear to disrupt the blood-brain barrier, allowing inflammatory mediators and immune cells into the central nervous system and, in the most catastrophic cases, causing the endothelial breakdown that underlies cerebral edema. A 2018 root-cause review of the discontinued JCAR015 program (a since-shelved CD19 CAR-T candidate, not an approved product) found that patients who died from cerebral edema had unusually rapid T-cell expansion within about a week of infusion, alongside a documented “complete breakdown of the blood-brain barrier” on autopsy — a stark illustration of the mechanism at its most severe, and part of why every currently approved CAR-T product carries the same boxed warning language today.

The ASTCT Consensus Grading System

Before 2019, different CAR-T trials graded neurotoxicity using different scales, making it difficult to compare safety data across products or even across sites. In November 2018 (published in Biology of Blood and Marrow Transplantation in early 2019), an American Society for Transplantation and Cellular Therapy (ASTCT) consensus panel led by Daniel W. Lee standardized both CRS and neurotoxicity grading; the neurotoxicity side of that framework is what coined the term ICANS and is now the reference standard cited in every FDA label and NCCN guideline.

The tool at its center is the ICE score (Immune Effector Cell-Associated Encephalopathy score), a simple 10-point bedside test: orientation to year, month, city, and hospital (4 points), naming three objects such as a clock, pen, and button (3 points), following a simple command (1 point), writing a standard sentence (1 point), and counting backward from 100 by tens (1 point). A perfect score of 10 means no impairment. Grade is then set by the worst of four domains — the ICE score itself, depressed level of consciousness, seizure activity, and elevated intracranial pressure/cerebral edema. Grade 1: ICE score 7–9, and the patient arouses spontaneously, managed with close observation and, per most product labels, consideration of a single dose of dexamethasone. Grade 2: ICE score 3–6; the patient awakens to voice, with standard management being scheduled dexamethasone or an equivalent corticosteroid, reassessed every 6–12 hours. Grade 3: ICE score 0–2, or the patient awakens only to tactile stimulation, with any seizure that resolves quickly counting as Grade 3, as does focal edema seen on brain imaging — this grade typically triggers ICU-level care and higher-dose corticosteroids. Grade 4: the patient is unarousable even to vigorous or repetitive stimuli, or has life-threatening seizures lasting more than five minutes, deep focal motor weakness, or diffuse cerebral edema — a medical emergency requiring ICU care, high-dose methylprednisolone, and often mechanical ventilation.

A separate, adapted version of the ICE score — the Cornell Assessment of Pediatric Delirium (CAPD) — is used for children and patients unable to complete the standard adult tasks, since a five-year-old cannot meaningfully “count backward from 100 by tens.”

How Often Does ICANS Actually Happen?

Incidence varies meaningfully by product, target disease, and CAR-T construct, but current FDA prescribing information gives a consistent picture: neurotoxicity of any grade is common, while severe (Grade 3+) neurotoxicity is a minority outcome for most products. Yescarta (axicabtagene ciloleucel, CD19-directed, for large B-cell lymphoma and other NHL) shows neurologic toxicities of any grade in 78% of patients across combined studies (74%–87% depending on the specific lymphoma population), with Grade 3 or higher in roughly 21%–31%. Kymriah (tisagenlecleucel, CD19-directed) shows any-grade neurotoxicity in 71% of pediatric/young-adult ALL patients (Grade 3+ in 22%), 60% of adult relapsed/refractory DLBCL patients (Grade 3+ in 19%), and 43% of follicular lymphoma patients (Grade 3+ in just 6%) — illustrating how disease burden and patient population shift the risk. Breyanzi (lisocabtagene maraleucel, CD19-directed) shows any-grade neurotoxicity in 32% of patients, Grade 3 or higher in 10%. Abecma (idecabtagene vicleucel, BCMA-directed, for multiple myeloma) shows any-grade neurotoxicity in 40% of patients, but Grade 3 in only 4% and Grade 4 in 0.6% — a notably lower severe-event rate than the CD19-directed lymphoma products. Carvykti (ciltacabtagene autoleucel, BCMA-directed, for multiple myeloma) shows ICANS specifically in 13% of patients, with Grade 3 or higher in just 2%. Carvykti's label carries an additional, distinct boxed warning for delayed-onset movement and neurocognitive disorders resembling parkinsonism, and for Guillain-Barré syndrome — separate diagnostic entities from ICANS itself that can appear weeks to months after infusion and are far less responsive to standard steroid treatment.

Across nearly every product, severe Grade 4 neurotoxicity is uncommon (typically low single digits or less), and death directly attributable to ICANS is rare — but not zero, which is exactly why every currently approved CAR-T product carries a boxed warning for potentially fatal or life-threatening neurologic toxicity, the FDA's strongest safety label category.

Timing, Overlap With CRS, and Reversibility

ICANS usually — though not always — trails or overlaps with cytokine release syndrome rather than appearing in isolation. Across FDA labels, median time to onset clusters between roughly 2 and 8 days after infusion, and the large majority of cases in most trials begin within the first one to two weeks, with a long tail of late-onset cases reported out to several months in rarer instances. Yescarta's label notes that neurologic toxicity can occur “concurrently with CRS or after CRS resolution,” and a 2020 review in Frontiers in Immunology similarly describes ICANS as typically accompanying and correlating with CRS, while occasionally occurring independently.

This is where one of the field's more counterintuitive findings comes in: tocilizumab, the IL-6 receptor–blocking antibody that is first-line therapy for CRS, is not considered an effective ICANS treatment on its own, and current management guidance (including the EBMT/EHA CAR-T Cell Handbook) centers on corticosteroids instead. The likely reason is pharmacologic: tocilizumab is a large monoclonal antibody that does not cross the blood-brain barrier, and by mopping up IL-6 in the bloodstream it can actually raise the amount of IL-6 available to cross into the central nervous system. That asymmetry is the core reason CRS and ICANS, despite sharing an inflammatory root cause, are treated as related but genuinely distinct toxicities with different first-line drugs.

The reassuring part, and it is a real one: most ICANS resolves. FDA labels report resolution in a large majority of affected patients — 88% for both Breyanzi and Abecma, for instance — with median symptom duration typically in the range of one to three weeks. The EBMT/EHA handbook states plainly that “most cases resolve and do not result in residual neurocognitive damage” when managed with prompt corticosteroids and supportive care. That said, “usually reversible” is not “always reversible”: a small proportion of patients experience lasting neurocognitive effects, and the field's sobering historical benchmark remains the 2016 clinical hold on Juno Therapeutics' JCAR015 program, discontinued after five ROCKET-trial patients died of treatment-related cerebral edema — a reminder that although current-generation approved products have a much better-characterized and generally more favorable safety profile, ICANS is not a toxicity to be managed casually.

How ICANS Is Managed in Practice

Because ICANS is graded by bedside neurologic exam rather than lab values, management starts with structured, repeated assessment — typically ICE-score checks at least twice daily during the CAR-T monitoring window (usually the first 7–10 days after infusion), with a lower threshold for children and cognitively impaired patients using the CAPD tool instead. Grade 1 cases often need only close observation and reversal of any sedating medications. Grade 2 typically calls for scheduled corticosteroids with reassessment and tapering as the patient improves. Grade 3 usually moves care to an ICU setting with higher corticosteroid doses, and Grade 4 calls for high-dose methylprednisolone, consideration of mechanical ventilation, and aggressive management of seizures or elevated intracranial pressure. Current guidance generally does not recommend routine prophylactic anti-epileptic drugs for every patient, reserving antiseizure medication for those who actually seize or who have Grade 3+ disease, with benzodiazepines used for status epilepticus. For corticosteroid-refractory cases, anakinra (an IL-1 receptor antagonist) has emerging use as a steroid-sparing option.

Bottom Line

ICANS is a real, sometimes serious, and mechanistically distinct companion to cytokine release syndrome in CAR-T therapy — a bedside-diagnosed encephalopathy that, in its most severe form, can progress to seizures, cerebral edema, and death, which is why it carries a boxed warning on every FDA-approved CAR-T label. But the fuller picture, backed by the same prescribing information, is more reassuring than the warning label alone suggests: across approved products, most neurotoxicity is low-grade, Grade 4 events are uncommon, the ASTCT consensus grading system gives clinicians a validated, standardized way to catch and stage it early, and the large majority of cases resolve fully within one to three weeks on corticosteroids and supportive care alone. Patients and caregivers considering CAR-T therapy should understand both halves of that picture — the real risk of severe, occasionally fatal neurotoxicity, and the equally real likelihood, borne out across thousands of treated patients, of a temporary and fully reversible course.