ICANS vs CRS: Mechanistic Origins and Clinical Insights

Cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) are the two most significant toxicities of CAR T-cell therapy, and while they often travel together, they arise through partly distinct biological routes. CRS is fundamentally a systemic inflammatory event driven by macrophages flooding the bloodstream with cytokines, whereas ICANS involves a more targeted assault on the brain’s protective barrier and the tissue behind it. The overlap is real but incomplete, and the clinical differences between the two have major implications for how each is treated.

How CRS Gets Started

CRS begins almost immediately after the engineered T cells engage their target. When CAR T cells latch onto tumor cells and kill them, the dying cells burst open and release molecular debris known as damage-associated molecular patterns. At the same time, the activated CAR T cells themselves secrete pro-inflammatory signaling molecules, including TNF-α, IFN-γ, and GM-CSF. That combination of tumor debris plus T-cell signals converges on bystander immune cells, especially macrophages, which then become the primary amplifiers of the inflammatory cascade.¹

This is the critical point: the CAR T cells light the match, but the macrophages are the ones who set the house on fire. Once activated, macrophages pour out their own cytokines, particularly IL-6, IL-1, and a host of chemokines. The result is a self-reinforcing loop where inflammation begets more inflammation. Research has shown that GM-CSF produced by the CAR T cells is a key driver of this macrophage activation. Blocking GM-CSF in laboratory experiments drastically reduced the macrophage-dependent release of CRS biomarkers like IL-6, IL-8, and MCP-1.²

CRS also has a vascular component. As the cytokine storm escalates, the lining of blood vessels becomes activated and damaged. In patients with acute lymphoblastic leukemia receiving CAR T cells, researchers found that endothelial cell activation tracked closely with CRS severity, with rising levels of markers like von Willebrand Factor and angiopoietin-2 and falling levels of angiopoietin-1.³ This vascular injury matters because it bridges the gap between CRS and ICANS: damaged blood vessels throughout the body set the stage for what can go wrong in the brain.

How ICANS Differs at the Biological Level

If CRS is a whole-body inflammatory wildfire, ICANS is what happens when that fire reaches the brain’s defenses. But it is not simply “CRS that moved upstairs.” ICANS has its own distinct mechanisms, and some of them have nothing to do with the systemic cytokine surge.

The brain is normally shielded by the blood-brain barrier, a tightly sealed layer of specialized cells that prevents most large molecules and immune cells from entering brain tissue. During ICANS, this barrier breaks down. Part of the explanation is straightforward: the same pro-inflammatory cytokines circulating during CRS activate the endothelial cells of the blood-brain barrier, causing them to loosen their connections and let immune cells and toxic molecules pass through.⁴ That much is a direct downstream consequence of CRS.

But there is another mechanism that is specific to ICANS and does not depend on CRS at all. Researchers discovered that pericytes, the cells that wrap around the tiny blood vessels in the brain and help maintain the blood-brain barrier, unexpectedly express CD19 on their surface. CD19 is the same protein that most commercial CAR T-cell products are designed to target on B-cell cancers. When CAR T cells recognize CD19 on brain pericytes, they attack those cells directly, punching holes in the barrier. In mouse experiments, infusion of CD19-directed CAR T cells disrupted the blood-brain barrier and increased its permeability. Human brain pericytes appear to be more vulnerable to this attack than mouse pericytes because they express higher levels of CD19.⁵

Once the barrier is breached, the resulting picture is not subtle. Cerebrospinal fluid from patients with ICANS shows elevated levels of a broad array of inflammatory cytokines, including IL-1, IL-6, IL-10, IFN-γ, and TNF-α, along with evidence that CAR T cells and myeloid cells have infiltrated the brain tissue itself. Markers of astrocyte injury and activation (GFAP and S100b) are significantly elevated in these patients compared to those who do not develop ICANS, confirming that an active inflammatory process is occurring within the central nervous system.⁵

Why the Two Syndromes Overlap but Are Not Identical

Clinically, CRS and ICANS are deeply intertwined. In a study of patients with B-cell lymphoma receiving CD19 CAR T cells, about 57% of patients who developed CRS also experienced ICANS, and nearly all patients with ICANS (97.5%) had CRS, suggesting a potential continuum between the two toxicities.⁶ CRS almost always comes first, typically within the first few days after infusion, while ICANS tends to follow, sometimes overlapping and sometimes emerging as CRS is resolving.

Yet the relationship is not one of simple cause and effect. Some patients develop severe ICANS despite only mild CRS, and the off-target attack on brain pericytes described above proceeds independently of how intense the systemic cytokine storm is. This means that controlling CRS does not automatically prevent ICANS, and it partly explains why the two conditions respond to different treatments.

What Each Syndrome Looks and Feels Like

CRS presents as a systemic illness. Fever is the hallmark, and it can range from low-grade to dangerously high. As severity escalates, patients may develop low blood pressure, rapid heart rate, difficulty breathing, and in severe cases, organ dysfunction. The grading system runs from grade 1 (fever alone) through grade 4 (life-threatening organ failure requiring intensive care). The symptoms are the body’s response to the cytokine flood and are conceptually similar to what happens during a severe infection, though the trigger is immunological rather than microbial.

ICANS shows up very differently. The earliest signs are often subtle cognitive problems. One particularly sensitive early marker is difficulty writing: patients develop handwriting abnormalities including repetitive scribbling of the same word, spatial disorganization on the page, and signs of impaired executive function. These writing problems frequently appear within days of infusion and often during the earliest phase of ICANS, sometimes before other neurological symptoms are obvious.⁷ As ICANS progresses, patients may become confused, develop word-finding difficulty, experience tremors, or in severe cases become unresponsive or develop seizures and cerebral edema.

The standard grading tool for ICANS relies on a bedside assessment called the Immune Effector Cell-Associated Encephalopathy (ICE) score, which tests orientation, ability to name objects, ability to write a sentence, counting, and ability to follow commands. The ASTCT consensus grading system, published in 2019, was created specifically to harmonize how both CRS and ICANS are graded across clinical trials and real-world practice, replacing several earlier scales that made it difficult to compare results between institutions.⁸

Treating CRS and ICANS Requires Different Strategies

The divergent mechanisms behind CRS and ICANS mean that clinicians cannot rely on a single drug for both. For CRS, the first-line treatment has long been tocilizumab, an antibody that blocks the IL-6 receptor. Tocilizumab works well for systemic inflammation, but it has a quirk: by blocking the receptor, it actually causes circulating IL-6 levels to rise temporarily, because the cytokine has nowhere to bind and accumulates in the bloodstream. That transiently elevated IL-6 can cross a compromised blood-brain barrier, which is why tocilizumab may not help with ICANS and could theoretically make it worse.

This pharmacological limitation has driven interest in siltuximab, an antibody that binds IL-6 directly rather than blocking its receptor. By neutralizing the cytokine itself, siltuximab avoids the paradoxical IL-6 spike. In a multicenter retrospective study, about 75% of patients who received siltuximab for CRS saw improvement in their CRS grade, with a median time to resolution of just one day.⁹ Comparative data suggest that siltuximab resolved CRS nearly five times faster than tocilizumab, though the difference has not yet reached statistical significance in available studies, likely because of small sample sizes.¹⁰ In pediatric and young adult patients, siltuximab used as a first-line therapy showed a tendency toward shorter CRS duration, fewer doses of tocilizumab needed, and later onset of ICANS.¹¹

ICANS management is a tougher problem. Corticosteroids remain the backbone, used to suppress the brain inflammation directly. But steroids are a blunt tool, and there is concern they might also suppress the anti-tumor activity of the CAR T cells. Anakinra, which blocks the IL-1 receptor, has emerged as a promising option for ICANS specifically, given the role of IL-1 in driving neuroinflammation. Data on anakinra show that among patients who achieved significant neurologic improvement (defined as at least a two-grade improvement in ICANS) within 72 hours of starting the drug, the median time to full ICANS resolution was just 3 days compared to 9 days for those who did not respond as quickly. Early responders also had substantially lower treatment-related mortality, with none dying within 28 days compared to about 21% in the slower-responding group.¹²

The GM-CSF Angle and Prophylactic Strategies

Rather than waiting for CRS or ICANS to develop and then treating them, some researchers are looking at preventing both upfront by targeting the shared upstream driver: GM-CSF. In an animal model designed to mimic human CRS and neuroinflammation, neutralizing GM-CSF reduced the infiltration of both myeloid cells and T cells into the central nervous system, significantly reduced neuroinflammation, and prevented CRS altogether. Intriguingly, this approach did not impair, and may have actually enhanced, CAR T-cell function against the cancer.¹³ Related work has shown that genetically knocking out GM-CSF production in the CAR T cells themselves abolishes the macrophage-dependent secretion of CRS mediators.²

This idea of engineering the toxicity out at the source, rather than managing it after the fact, is driving a broader movement in the field. Researchers are developing synthetic genetic switches that would give clinicians spatiotemporal control over CAR T-cell activity: the ability to turn cells on or off, dial their intensity up or down, or restrict their function to certain tissues.¹⁴ These are still largely preclinical concepts, but they represent the most ambitious approach to the CRS/ICANS problem: redesigning the therapy so that the toxicities never develop in the first place.

Predicting Who Will Get Hit and How Hard

Not every patient who receives CAR T cells develops severe CRS or ICANS, and figuring out in advance who is at highest risk is an active area of research. A systematic framework for predictive biomarkers has been proposed to help organize the growing list of candidates.¹⁵ Pre-infusion factors like high tumor burden, elevated baseline inflammatory markers, and pre-existing endothelial dysfunction have all been associated with greater toxicity risk. The endothelial activation markers mentioned earlier, like angiopoietin-2 levels, are among the more promising early warning signs.

The target antigen matters too. A real-world analysis of the FDA’s adverse event reporting database found that anti-CD19 CAR T products generated a higher neurotoxicity signal than anti-BCMA products, which target a different protein found on myeloma cells.¹⁶ This difference likely reflects the CD19-on-pericytes mechanism unique to CD19-directed therapy. If your CAR T cells are built to chase a protein that happens to sit on brain blood vessel cells, you are inherently at greater risk for neurotoxicity than someone whose treatment targets a protein found only on tumor cells.

Long-Term Cognitive Fallout

Most CRS resolves within days and leaves no lasting trace. ICANS, however, tells a more complicated story. A systematic review of neurocognitive outcomes found that while cognition was often stable through the first 90 days after treatment, small but persistent declines in memory, language, organization, and divided attention appeared over the following months, extending up to a year after infusion. These lingering deficits were particularly pronounced in patients who had experienced grade 2 or higher ICANS. Quality of life generally improved by day 90 for most patients, but those with more severe ICANS showed persistent impairments in physical and emotional functioning that could last 12 months or longer.¹⁷

A separate study looking specifically at patients with diffuse large B-cell lymphoma who received CAR T cells identified a trend toward long-term neurocognitive dysfunction in those who experienced ICANS, though the association did not reach statistical significance.¹⁸ The evidence here is still maturing, but it points in a worrying direction: the brain inflammation of ICANS may not always fully resolve, and some patients are left with subtle cognitive costs even after the acute syndrome has passed. This is an area where longer follow-up studies and more sensitive neuropsychological testing will be needed to define the true scope of the problem.

When Different Cell Platforms Change the Equation

Almost everything discussed so far applies to CAR T cells specifically. But CAR therapy is expanding beyond T cells. CAR-NK (natural killer) cells are being tested in various cancers, and their cytokine profile differs from T cells in ways that may shift the CRS and ICANS landscape. NK cells produce a different cocktail of inflammatory mediators than T cells, which could in principle lead to less macrophage activation and a milder systemic inflammatory response.¹⁹ Early clinical experience with CAR-NK cells has generally shown lower rates of severe CRS, though the data are still limited and the field is young.

The possibility that swapping the effector cell type could reduce toxicity without sacrificing anti-tumor activity is one of the more intriguing threads in the field. It also underscores a broader point: CRS and ICANS are not inevitable consequences of redirecting immune cells against cancer. They are consequences of the specific way T cells interact with their targets and with bystander cells, and changing any link in that chain can change the toxicity profile.

The Challenge of Faithful Animal Models

One reason the mechanistic understanding of CRS and ICANS has been slow to develop is that both syndromes are difficult to reproduce in laboratory animals. Standard mouse models of cancer immunotherapy do not recapitulate the human immune response well enough to produce the full spectrum of CRS and ICANS symptoms. Researchers have turned to non-human primates as a more faithful platform, creating the first primate model that reproduces both CRS and neurologic toxicity following B-cell-directed CAR T-cell therapy.²⁰ These primate models are valuable because they allow researchers to test interventions and study the biology in a system that closely mimics human physiology, though they come with obvious ethical considerations and high costs that limit their use. As the field moves toward engineered solutions like genetic switches and GM-CSF knockouts, having animal models that reliably produce both toxicities becomes increasingly important for testing whether a modification that looks good in a dish actually works in a living immune system.