The Institute of Medicine’s (IOM) Standards for Developing Trustworthy Clinical Practice Guidelines were used as a model to develop the evidence- and consensus-based recommendations in this article. IOM standards dictate that guideline development is led by a multidisciplinary team using a transparent process where both funding sources and conflicts of interest are readily reported. Recommendations are based on literature evidence, where possible, and clinical experience, where appropriate.25 The American Society for Hematology (ASH), the American Society for Transplantation and Cellular Therapy (ASTCT), FACT at the University of Nebraska Medical Center, and the Emily Whitehead Foundation also provided representatives to serve on SITC’s Immune Effector Cell-related Adverse Events Expert Panel. For transparency, a draft of this clinical practice guideline was made publicly available for comment after journal submission. All comments were considered for inclusion into the final article. This clinical practice guideline is intended to provide guidance and is not a substitute for the professional judgment of individual treating physicians.

Panel recommendations were derived from evidence within the published literature, as well as discussions during an in-person consensus meeting and regular communications and collaborative editing throughout the manuscript development process along with responses to a clinical questionnaire that addressed current practices in the use or recommendation for use of immunotherapy agents . Evidence supporting panel recommendations was graded according to the Oxford Center for Evidence-Based Medicine (OCEBM) Levels of Evidence (2016 version).26 A summary of the OCEBM grading scale may be found in table 1. The level of evidence (LE) for a given consensus recommendation is expressed in parentheses following the recommendation (eg, LE: 1). Recommendations without an associated LE were based on expert consensus. Consensus was defined as ≥75% agreement among expert panel voting members.

As outlined by IOM standards, all financial relationships of expert panel member that might result in actual, potential, or perceived conflicts of interest were individually reported. Disclosures were made prior to the onset of manuscript development and updated on an annual basis. In addition, panel members were asked to articulate any actual or potential conflicts at all key decision points during guideline development, so that participants would understand all possible influences, biases, and/or the diversity of perspectives on the panel. Although some degree of relationships without side interests among panel members are to be expected, those with any significant financial connections that may compromise their ability to fairly weigh evidence (either actual or perceived) were not eligible to participate.

Recognizing that guideline panel members are among the leading experts on the subject matter under consideration and guideline recommendations should have the benefit of their expertise, any identified potential conflicts of interests were managed as outlined in SITC’s disclosure and conflict of interest resolution policies. As noted in these policies, panel members disclosing a real or perceived potential conflict of interest may be permitted to participate in consideration and decision-making of a matter related to that conflict, but only if deemed appropriate after discussion and agreement by the expert panel.

The financial support for the development of this guideline was provided solely by SITC. No commercial funding was received.

The onset and severity of therapy-associated toxicities typically correspond with the timing and degree of peak CAR T cell activation and proliferation, as well as underlying disease.19 27–32 Therefore, product and disease-specific characteristics may place some patients at higher risk for adverse events. For the approved CAR T cell therapies, tisagenlecleucel, axicabtagene ciloleucel, and brexucabtagene autoleucel, structural differences in the costimulatory domains used to promote T cell activation have been implicated in the distinct risk profiles reported in trials to date. Both tisagenlecleucel and axicabtagene ciloleucel are based on identical antigen recognition domains (the anti-CD19 short-chain variable fragment FMC63), and both are constructed with identical CD3ζ T cell signaling domains. Tisagenlecleucel deploys a CD8α transmembrane region and a 4-1BB costimulatory domain, while axicabtagene ciloleucel and brexucabtagene autoleucel use a CD28 transmembrane region and a CD28 costimulatory domain.1 2 CD28-based CARs have been shown to induce rapid early expansion of T cells with boosted effector functions, whereas 4-1BB-based CARs cause a more gradual expansion of T cells, shifted toward a central memory-like phenotype, with a potential for longer persistence.33 34 The different signaling and expansion characteristics may explain the earlier timing of onset of adverse events that has been reported with CD28-containing CAR T cell therapies.35

In the pivotal ZUMA-1 trial, which led to the approval of axicabtagene ciloleucel, patients who had undergone prior allogeneic hematopoietic stem cell transplantation (allo-HSCT) were excluded, as were patients who had previously been treated with CAR T cells. Other exclusion criteria included central nervous system (CNS) involvement or a history of or active CNS disorder, such as seizure, cerebrovascular ischemia/hemorrhage, dementia, cerebellar disease, or any autoimmune disease with CNS involvement.24 36 In the ZUMA-2 trial that led to approval for brexucabtagene autoleucel for RR MCL, durable remissions were attained in the majority of patients, and the toxicities were similar to those observed with other CD28 costimulated products.14 37 The JULIET trial, which was pivotal for the approval of tisagenlecleucel for the treatment of diffuse large B cell lymphoma (DLBCL), used similar criteria, with the addition of excluding patients with class III/IV cardiovascular disability according to the New York Heart Association Classification and patients in complete remission—although a history of prior CNS disease was allowed as long as no active involvement was present at the time of treatment.38 Subsequent trials have observed acceptable safety profiles for CAR T cell therapy in patients with active CNS disease—the phase I TRANSCEND NHL (Non-Hodgkin Lymphoma) 001 study, which evaluated lisocabtagene maraleucel, a CD19 CAR T cell product with defined CD4⁺/CD8⁺ composition, reported only one neurological event among the nine treated patients with secondary CNS lymphoma.39 40

Current approvals only permit the use of CAR T cell therapies in patients with more advanced disease. However, relapse after complete remission is relatively common in both leukemia and lymphoma, and outcomes for these patients are generally poor. Based on the known mechanisms of CAR T cell therapies, patients with very low disease burden or in complete remission can likely be treated, potentially with greater safety, given the importance of disease burden as a risk factor for CRS.41–43

The adverse events associated with FDA-approved CAR T cell therapies may develop suddenly and progress rapidly to life-threatening toxicities. As additional targets and indications expand the available repertoire of IEC therapies, the management of the toxicities associated with new products may require further modification. Regardless, establishing baseline values for markers of CRS and neurotoxicity is important so that the caregiving team can quickly recognize and respond to adverse events at the earliest possible onset. Several scoring systems have been developed to monitor baseline mental state, with the Immune Effector Cell-Associated Encephalopathy (ICE) score (a modified version of the CARTOX-10 criteria25) being the most widely used for adult patients.45 For pediatric patients, the Cornell Assessment of Pediatric Delirium (CAPD),46 has been validated for evaluating mental state after CAR T cell therapy.47 Elevated C-reactive protein (CRP) and ferritin levels of >10,000 ng/mL have been observed in almost every patient who developed severe CRS across public trials. It is important to note, however, that elevated levels of these markers are associated with, but not predictive of, toxicity.23 48–52

Immunosuppression secondary to lymphodepleting conditioning is an expected toxicity associated with CAR T cell therapy. Patients’ immune defenses may be further compromised by on-target killing of B cells by CD19-targeting CAR T cell products,7 9 54 55 as well as by emerging therapies such as BCMA-directed CAR T cells for multiple myeloma. Bacteremia, fungal infections, recurrent urinary tract infections, as well as viral infections such as influenza, respiratory syncytial virus, and herpes zoster have all been reported after CAR T therapy8 9 24 37 56–60 and reactivation of cytomegalovirus (CMV), Epstein-Barr virus (EBV), and human herpesvirus 6 are also of concern. In the ELIANA registration trial of tisagenlecleucel for children and young adults with RR ALL, infections occurred in 43% of patients, with 21% experiencing grade 3 and 3% grade 4 infections. In this trial, some patients with prolonged grade 3 or 4 neutropenia before and after tisagenlecleucel infusion experienced severe or fatal infections, including grade 3 human herpesvirus 6 encephalitis.23 A review of 133 patients with RR ALL, chronic lymphocytic leukemia (CLL), and NHL treated with CD19 CAR T cells observed that 26 patients (23%) developed infections within 4 weeks of infusion and 19 (14%) developed infections within 29–90 days after treatment. Bacterial infections were most common within the first 28 days, and viral infections occurred more frequently after day 29.58 In the ZUMA-2 trial, infections of grade 3 or higher occurred in 32% of the patients, with the most common being pneumonia (9%).14

Protocols vary from center to center for the routine administration of prophylactic antibacterial, antiviral, and antifungal agents for patients undergoing CAR T cell therapy. The US Centers for Disease Control and Prevention (CDC), together with the Infectious Diseases Society of America (IDSA) and ASTCT, has published guidelines for preventing infections in patients undergoing allo-HSCT,61–63 which apply to all myeloablative transplants regardless of recipient (adult or child). The published guidelines do not make a definitive statement on antibacterial prophylaxis, although the prospect of septicemia is a significant concern for patients undergoing CAR T cell therapy. A review of 31 centers with experience from approximately 3400 stem cell transplants found substantial heterogeneity in adherence to published guidelines, yet a majority of institutions use some form of antiviral prophylaxis, most often reflecting concern for latent CMV, herpes simplex virus (HSV), and varicella zoster virus (VZV) reactivation.64 Additional considerations for infection prevention in patients undergoing CAR T cell therapy may be found in a 2020 perspective from Hill and Seo.65

Patients with cancer are also at high risk of severe influenza-related complications, with a mortality rate 10 times higher than the general population. A Cochrane review of six studies with a total of 2275 participants found evidence, although weak, that the benefits of influenza vaccination outweigh potential risks in patients with cancer.45 66–68 However, patients with hematological malignancies frequently present with profound immunosuppression and cytopenias, both due to disease and prior therapies, which may limit the efficacy of seasonal influenza vaccines, especially in younger populations.69 70

Reports have emerged of two patients with HIV and RR DLBCL successfully being treated with axicabtagene ciloleucel. One patient was non-adherent to antiretroviral drugs prior to initiating therapy and underwent lymphodepletion with CD4⁺ cell counts of 53 cells/mm³ and detectable viral load. After CAR T cell infusion, the patient developed grade 2 CRS and grade 3 ICANS, both of which resolved after tocilizumab and dexamethasone. The other patient was adherent with antiretroviral therapy (bictegravir, emtricitabine, and tenofovir alafenamide) with an undetectable HIV viral load and a CD4 count of 127 cells/mm³. Both patients achieved complete remissions of their lymphoma after CAR T cell therapy.71 More experience is needed, however, to determine if HIV-positive patients are at increased risk of infections compared with other patients undergoing CAR T cell therapy.

The pandemic caused by the novel coronavirus SARS-CoV-2 has disrupted almost all aspects of the healthcare system, including cancer care. Treatment plans for patients undergoing CAR T cell therapies must take into account potential limitations in hospital resources during the ongoing pandemic, as well as an utmost concern for the safety of everyone involved. Because cellular therapy offers a potentially curative option for patients with otherwise extremely poor prognoses, delaying treatment may not be an option in many cases. However, adequate staff and supportive care resources are an absolute prerequisite for proceeding with CAR T cell therapy, and it is important to ensure that tocilizumab—which, along with other interleukin (IL)-6-modulating agents, was used as an investigational intervention for COVID-19—is available for the management of CRS. To assist in clinical decision making during the pandemic, the panel recommends referring to the interim guidelines published by the CAR T cell consortium investigators, which discuss necessary resources, determinants of cell therapy use, patient selection for B cell non-Hodgkin lymphomas and B cell ALL, supportive care measures during cell therapy administration, availability of tocilizumab doses for each patient during the CRS risk period, and collaborative care with referring physicians.72

Some of the most frequently seen adverse events associated with the FDA-approved CAR T cell therapies are CRS, with rates of occurrence ranging from 37%–93% in patients with lymphoma24 38 71 and 77%–93% in leukemia,23 73 74 and neurological toxicities, now referred to as ICANS,75 which have been reported in every trial involving CD19 targeted CAR T cells.76 Early trials frequently deployed different assessment and grading criteria, making it difficult to compare the incidence and severity of CRS and other adverse events across studies. For example, in the National Cancer Institute’s Common Terminology Criteria for Adverse Events (CTCAE) V.4.03, which was in effect when most CAR T trials began, fever was not included as a prerequisite for CRS and grading largely depended on whether infusion was interrupted,77 which is not applicable for CAR T cells that are normally given as a single dose over less than 30 min. Several grading systems, including the Lee Criteria, Penn Criteria, and CARTOX Criteria, were developed, defining severity by the requirement for supplemental oxygen and the need for intravenous fluids and vasopressors.28 59 78–80 Until recently, neurological events were considered to be a component of CRS. However, although severe neurotoxicity seems to exclusively affect patients who develop CRS, the timing can vary greatly, occurring concomitantly or days after CRS resolves.51 56 In 2019, the ASTCT published consensus recommendations for separate grading systems for CRS and ICANS (tables 2–4).75 Advantages to the ASTCT system are its ease of use, objectivity, solid basis on criteria that may be immediately assessed by all members of a caregiving team, and accuracy in categorizing the severity of toxicities, although disadvantages include dependence on the selected interventions for grading.

Retrospective analysis of CRS grading systems demonstrated some alignment between the ASTCT and Lee criteria, in addition to ASTCT versus Penn criteria, while there were notable differences between Penn and Lee criteria. Similarly, the CTCAE neurotoxicity grading and ASTCT ICANS grading are quite different, with retrospective grading demonstrating key areas of disagreement.81 As such, interpretation of published CRS and ICANS incidence and severity data must always take into account the grading system used.

Patients with CRS frequently present with elevated serum CRP and ferritin; however, many standard markers (including ferritin, CRP, LDH, aspartate aminotransferase, alanine aminotransferase, and creatinine) may only reach abnormal levels after the onset of clinical manifestations of CRS.50 Although not typically reported in clinical charts, the hallmark cytokines of CRS associated with CD19-directed CAR T therapies are IL-10, IL-6, and interferon gamma (IFN-γ).21 50 59 78 85 87 89 90 Other markers of inflammation, including tumor necrosis factor alpha (TNF-α) and granulocyte macrophage colony-stimulating factor (GM-CSF), have also been observed at elevated levels in patient serum.50 85 90 Inhibition of GM-CSF has also been demonstrated to ameliorate CRS pathology in xenograft models of leukemia.82 A recent effort to identify risk-associated biomarkers determined that none of the standard clinical laboratory tests were predictive of CRS severity. That study developed a three-cytokine signature, based on samples drawn in the first 72 hours, using IFN-γ, soluble gp130, and soluble IL-1 receptor antagonist, which enabled accurate prediction of which patients with ALL developed grade 4–5 CRS with 86% sensitivity (95% CI 57% to 98%) and 89% specificity (95% CI 73% to 97%). In pediatric patients, a signature based on IFN-γ, IL-13, and the macrophage inflammatory protein MIP1α in samples drawn within the first 72 hours after treatment predicted grade 4–5 CRS with 100% sensitivity (95% CI 72% to 100%) and 96% specificity (95% CI 81% to 100%).50 Because real-time monitoring of cytokine levels is typically unavailable in most treatment centers, most grading systems (including the ASTCT consensus grading system75) are based on clinical observations.

Per the ASTCT grading system (table 2), fever defines the onset of CRS, with a temperature ≥38°C not attributable to any other cause being the sole symptom required for classification as grade 1. Although fever defines the onset, if patients are treated with antipyretics or anticytokine therapy, elevated temperature is not required for subsequent grading, and hypotension and hypoxia are used to assess severity. At grade 2 CRS, patients present with hypotension not requiring vasopressors and/or hypoxia requiring low-flow nasal cannula (≤6 L/min) or blow-by oxygen. Hypotension requiring a vasopressor (with or without vasopressin) and/or hypoxia necessitating high-flow (>6 L/min) nasal cannula, face mask, non-rebreather or venturi mask defines grade 3 CRS. At grade 4, patients need multiple vasopressors (again, excluding vasopressin) and/or positive pressure ventilation up to and including intubation. Grade 5 CRS is defined as death in which another cause is not the principle factor. Resolution is defined by the normalization of all signs and symptoms that led to the diagnosis of CRS. In patients treated with anticytokine therapy whose temperatures have come down, CRS may be downgraded as hypotension and hypoxia improves.75

Beyond fever, hypotension and hypoxia, symptomology varies widely. CRS symptoms may resemble tumor lysis syndrome (TLS),90 92 or at higher grades, the clinical presentation can mimic infection, with malaise, fatigue, anorexia, myalgias, nausea, vomiting, headache, and mental status changes.20 21 27 59 75 78 80 87 90 Myalgias and headaches are extremely common.93 94 Since CRS and sepsis can coexist, it is very important to presume patients are infected and managed accordingly while investigating underlying infections in patients undergoing IEC therapies.

Typically, CRS occurs within 1–2 weeks of cell administration.28 Peak signs and symptoms have been reported to correlate with maximal CAR T cell proliferation. For tisagenlecleucel and axicabtagene ciloleucel, the median time to onset is 2–3 days but can be as early as within 24 hours, particularly with axicabtagene ciloleucel. Notably, the time to onset of fever may be earlier in patients treated with CD28− (ie, axicabtagene ciloleucel and brexucabtagene autoleucel) compared with 4‐1BB‐costimulated (ie, tisagenlecleucel) CAR T cell products.35 Most cases of CRS resolve within 7–8 days of the onset of symptoms, but cases have been documented that persist more than 30 days.95 The package inserts for Yescarta (axicabtagene ciloleucel) and Tecartus (brexucabtagene autoleucel) mandate 7 days of daily monitoring after infusion, whereas Kymriah (tisagenlecleucel) may be administered in the outpatient setting, provided patients are monitored two to three times during the first week after infusion. Patients or their caregivers need to have a thermometer available to monitor and report any fever. Both products advise patients to remain within a 2-hour drive of their treatment center for 4 weeks,1–3 and the clinical infrastructure and communications need to be clear and able to expeditiously admit and evaluate a patient post-CAR T who reports a fever.

Across several trials, the most important predictor of severe CRS after CAR T therapy has been disease burden.4 31 60 90 96 A retrospective single-center study of 98 children with ALL treated with tisagenlecleucel found that pretreatment blast count of >25% increased the risk for hypotension-requiring inotropic support. In the study, pre-existing cardiac risk factors, systolic or diastolic dysfunction, or pre-existing ECG abnormalities were also risk factors.97 Higher doses of CAR T cells and CARs containing CD28 costimulatory domains are also associated with increased risk.49 57 95 96

Management of CRS often involves direct targeting of elevated cytokines, especially IL-6. The IL-6 receptor-blocking antibody tocilizumab was approved by the FDA in 2017 for the treatment of severe or life-threatening CAR T-induced CRS.98 However, some protocols advise administering tocilizumab at earlier onset. A treatment algorithm developed from experience with the axicabtagene ciloleucel clinical trials advised considering 8 mg/kg intravenous tocilizumab for persistent (lasting >3 days) and refractory fever in grade 1 CRS, and anti-IL-6 therapy is recommended for managing hypotension, hypoxia, and organ toxicity at all grade 2 and higher cases.28 CRS management for brexucabtagene autoleucel in ZUMA-2 was similar to previous experience from anti-CD19 CAR T cells that have CD28 and CD3 ζ costimulatory domains.14 During the tisagenlecleucel clinical trials, early CRS symptoms were managed with tocilizumab, and corticosteroids were only used as second-line therapy, in case of refractory hypotension or other severe CRS symptoms not responsive to the first dose of tocilizumab.80 Methylprednisolone (2 mg/kg/day) and dexamethasone (0.5 mg/kg; maximum, 10 mg/dose) are both commonly used corticosteroids. Some protocols, in particular those with CD19-directed CAR T cell therapies with CD28 signaling domains, recommend using dexamethasone for patients with neurological symptoms due to more efficient penetration of the blood–brain barrier. Many patients, especially those with ALL or multiple myeloma, have received corticosteroids as part of their treatment regimen, and therefore some patients who experience CRS may have a relative corticosteroid deficiency due to suppression of their hypothalamic–pituitary axis.59 High-dose steroids are not recommended as a first steroid therapy, as >100 mg daily of prednisone equivalent may reverse symptoms but risks concurrently ablating 19-28z CAR T cells.79

Consideration has been given to using tocilizumab preemptively or as prophylaxis. The Fred Hutchinson Cancer Research Center, Seattle group reported early administration of tocilizumab with and without dexamethasone after treatment with a 4-1BB second-generation CAR T cell product with encouraging data of a modest decrease in severe CRS.99 A similar clinical trial at Children’s Hospital of Philadelphia prospectively tested a preemptive approach in pediatric ALL patients with high disease burden (>40% marrow blasts), giving tocilizumab at the time of fever. This preemptive trial met the endpoint for grade 4 CRS reduction (by one-third) in these high-risk patients.100

Other cytokine‐directed therapies, such as siltuximab, a chimeric anti‐IL‐6 mAb, have occasionally been used in the management of CRS.24 85 Based on preclinical models and known cytokine profiles, anti-IL-1 therapies, such as anakinra, have been proposed for CRS treatment, but data are lacking regarding their efficacies in humans.9 83 Reports have emerged of patients presenting with atypical, tocilizumab-refractory CRS, characterized by early increases in GM-CSF and IL-2,101 highlighting a need for alternate approaches. In cases of CRS that do not respond to 2 to 3 doses of tocilizumab, benefit with additional doses is unclear. Additionally, a retrospective analysis of 100 patients with RR large B cell lymphoma treated with axicabtagene ciloleucel found a trend for shorter progression-free survival (PFS) among patients receiving corticosteroids for a prolonged time.102

Several of the cytokine profiles and serum markers associated with ICANS overlap with CRS. During the ZUMA-1 study, severe neurotoxicity was associated with elevated levels of IL-1Rα, IL-2Rα, IL-6, IL-8, IL-10, IL-15, IFN-γ, CCL-2, granzyme B, GM-CSF, and ferritin.114 Similarly, elevated levels of IL-1α, IL-2, IL-3, IL-5, IL-6, IL-10, IL-15, IFN-γ, G-CSF, GM-CSF, and MCP1 were observed in the 33 of 53 adult patients with RR B-ALL who developed neurotoxicity during phase I trials of 19-28z CAR T cells.51 A study encompassing 133 patients with relapsed B cell malignancies reported elevated CRP, ferritin, IL-6, IL-8, MCP1, IFN-γ, and TNF-α in the 53 patients who developed ICANS.57 Markers of diffuse intravascular coagulation have also been correlated with high-grade neurotoxicity, with two studies reporting elevated vWF and high ANG2:ANG1 ratios in patients with severe ICANS.51 56

Prior to infusion, patients should be evaluated for baseline neurological and mental status. The ASTCT consensus grading system for ICANS (see tables 2 and 3) deploys a 10-point scoring metric called the ICE tool75 for ICANS-associated encephalopathy in adults, which builds and simplifies the assessment items developed in the CARTOX criteria.28 Points are assigned for orientation to year, month, city, and hospital, ability to name three objects, ability to follow simple commands, ability to write a standard sentence, and ability to count backwards from 100 by 10. For pediatric patients, the CAPD is used.75

Across studies the most consistent factors associated with the development of neurotoxicity have been disease burden and peak CAR T cell expansion.51 56 114 Classification tree modeling from the Fred Hutchinson Cancer Research Center study demonstrated that patients with fever of ≥38.9°C and serum IL-6 of ≥16 pg/mL and MCP-1 of ≥1343.5 pg/mL in the first 36 hours after CAR T cell infusion were at high risk of subsequent grade ≥4 neurotoxicity (sensitivity 100%; specificity 94%).56 Other risk factors include extramedullary disease,51 younger age, pre-existing neurological comorbidities, higher total CAR T cell doses, and cytopenias.56 High-grade CRS is associated with a greater risk of ICANS.23 Although real-time serum cytokine monitoring is typically unavailable outside a few academic centers, several biomarkers have been significantly associated with neurological adverse events, including IL-6, IL-10, IL-15, IL-2Rα, and granzyme B.24 Fludarabine-containing conditioning regimens have been raised as a concern,56 and initially fludarabine was thought to have increased the risk of lethal cerebral edema during the ROCKET trial of JCAR015.115 This was not borne out, because additional deaths occurred with a modified lymphodepletion regimen. Because the neurological symptoms and timing of onset associated with fludarabine are distinct from those seen with CD19 CAR T neurotoxicity, the agent has now come to be broadly recognized as not a primary driver for ICANS.97 More likely, the addition of fludarabine may have contributed indirectly to greater baseline lymphopenia and consequent further expansion of CAR T cells mediating toxicity. Therefore, it should be recognized that the interaction between an individual cell product and the lymphodepleting regimen may be a determinant of toxicities.

The incidence of ICANS seems to be highest for CD19-targeting CAR T cell therapies, potentially due to the robust T cell expansion seen with CD19 CAR T cell products.7 19 22 27 31 35 48 51 52 56 79 Neurological toxicity has been reported in studies of CAR T cells directed against CD22 and BCMA, but initial data suggest that ICANS may occur at lower rates than for CD19 CAR T cells, although the number of published studies is too small to make any definitive conclusions.16 116 No studies have directly assessed differences in risk for ICANS between CAR T cells with CD28 and 4-1BB costimulatory motifs, and comparisons are complicated by the fact that these products were evaluated in trials with different grading systems, patient populations, and disease states. In general, neurological events were reported more frequently for axicabtagene ciloleucel1 than tisagenlecleucel2 and the robust early expansion of CD28-costimulated CD19 CAR T cells could potentially set the stage for CRS and ICANS, but neurotoxicity has been reported in trials of products containing both costimulatory domains.

Common clinical manifestations of ICANS include confusion or delirium, expressive aphasia, weakness, tremor, headache, seizures, and altered level of consciousness.95 Headache alone is not considered a useful diagnostic symptom for ICANS, as it is very common with CRS and frequently co-occurs with fever.75 Seizures have occurred with variable frequency, with some centers observing non-convulsive status epilepticus in roughly 10% of patients treated with CAR T and one study reporting generalized tonic–clonic seizures in roughly 30% of patients.48 51 56

Aphasia, in particular anomia, and word-finding defects may be an important early warning sign, as one study of 53 patients with RR ALL found that 85% of subjects who displayed impaired naming of objects, stuttering, or perseverative speech after treatment with an anti-CD19 CAR T went on to develop severe ICANS.51 Handwriting changes have also been reported as the first sign to appear prior to the onset of high-grade ICANS.23 28 116 Although CRS and ICANS are considered separate pathologies, fever precedes the onset of neurological symptoms in a majority of patients.23 24 51 56

Cerebral edema may appear as areas of low density on unenhanced CT images and as increased T2 and FLAIR signal changes on MRI. Visible edema on neuroimaging is classified as grade 3 or grade 4 ICANS by the ASTCT consensus system, depending on whether it is local or diffuse.75 In the majority of cases, however, patients with ICANS frequently present with unremarkable MRI and CT scans.51 79 Findings on EEG may reveal non-convulsive status epilepticus in CAR T-treated patients, with diffuse slowing and frontal intermittent rhythmic delta activity (FIRDA).48 However, FIRDA and generalized background slowing on EEG are both non-specific signs of diffuse cerebral dysfunction also seen in metabolic encephalopathy, infections, centrally acting medications, or neurodegeneration.120

Across trials, the median time-to-onset of neurological events has been reported as 4–5 days after infusion and resolution has typically been seen within 3–8 weeks.23 24 56 121 Neurological toxicity may occur earlier in patients treated with CD28 costimulated CAR T cells than for those receiving products with a 4-1BB domain. For axicabtagene ciloleucel, the median time to onset is 4 days after infusion, and for tisagenlecleucel, the median time to first event is 6 days.1 2 ICANS has been observed concurrently with CRS, shortly after CRS symptoms subside, as well as a delayed-onset form occurring up to 1 month after CAR T cell infusion.9 28 114 The duration of neurological toxicities may vary depending on the product and disease state. For axicabtagene ciloleucel, the median duration was 17 days, and for tisagenlecleucel, the median duration was 6 days for patients with RR ALL and 14 days for patients with RR DLBCL. In most patients ICANS resolved within 3 weeks, although prolonged encephalopathy lasting up to 173 days was noted.1 2 Long-term follow-up studies of patients treated with CAR T cells are ongoing, and late-onset neurological events have been reported. Among 86 patients treated with CD19 CAR T cells in a phase I/II trial, 9 (10%) were found to have 11 new neurological findings at a median of 28.1 months (range 12.1–62.6) after infusion, including three cerebrovascular accident events and one transient ischemic attack. Additionally, eight patients (9%) experienced psychiatric events requiring intervention: four with newly diagnosed mood disorders and four with exacerbation of previous depression and anxiety.122

Across several trials, tocilizumab has failed to resolve symptoms of ICANS, despite alleviating severe CRS.23 24 51 56 It remains to be determined whether targeting IL-6R in isolation during established CRS is insufficient to prevent subsequent neurotoxicity or if the lack of efficacy is due to tocilizumab’s inability to cross the blood–brain barrier.123 Because tocilizumab may not penetrate the CNS and causes at least a transient rise in serum IL-6, some have postulated that it may worsen neurotoxicity by increasing CSF IL-6 levels.56 123 Siltuximab, an anti-IL-6 antibody that may prevent high IL-6 concentration in the CSF by removing it from circulation in the serum, and the IL-1 antagonist, anakinra, have both been proposed as potential alternatives, but data are lacking on their safety and efficacy.

Corticosteroids have been successfully used for the management of ICANS. The axicabtagene ciloleucel package insert advises 10 mg IV dexamethasone every 6 hours for neurological toxicity of ≤grade 3 and 1000 mg IV methylprednisone daily for 3 days with grade 4 ICANS.1 Although corticosteroids may reduce circulating CAR T cell counts, evidence is emerging that even longer courses of steroids (>7 days) do not alter efficacy of cancer treatment.118 However, a trend towards shorter PFS has been observed in patients with lymphoma treated with axicabtagene ciloleucel who received longer and earlier intervention with steroids.102

Seizure prophylaxis has been implemented in a few studies, but the ideal dose and duration have not yet been determined.28 48 51 Levetiracetam has a better drug–drug interaction profile and lower risk of cardiotoxicity compared with other antiepileptic agents, can be administered safely to patients with hepatic dysfunction, and does not affect cytokine levels.28 In the Memorial Sloan Kettering Cancer Center trial, 14 patients developed seizure despite levetiracetam prophylaxis, but all resolved with standard management with benzodiazepines and antiepileptic agents.51

The high-profile termination of the phase II ROCKET trial (NCT02535364) of the investigational CD19-CAR T JCAR015 in adults with RR ALL raised concerns about the danger of fatal cerebral edema in patients undergoing IEC therapies.115 Fatalities were also reported during the safety expansion phase of the ZUMA-1 study,36 in the standard of care setting,110 as well as the phase I/II trial of JCAR014, a defined composition CD4⁺/CD8⁺ CAR T cell product for the treatment of RR ALL, CLL and NHL (NCT01865617).11 Initially, the use of fludarabine as a lymphodepleting agent was suspected as the etiological agent for cerebral edema, leading to a modified conditioning regimen in the ROCKET trial. However, subsequent fatalities, in addition to the distinct manifestation and timing of symptoms for adoptive T cell transfer-associated neurotoxicity, led to the conclusion that fludarabine may not be the primary cause for cerebral edema and that the agent may continue to be safely used for lymphodepletion.97

All five patients who died in the ROCKET trial experienced rapid, early expansion of the modified CAR-bearing T cells within a week of being infused. Additionally, high levels of CD8⁺ T cells and a concurrent sharp spike in IL-2 and TNF-α were significantly correlated with fatal brain swelling.115 Postmortem analyses of patients from both the ROCKET and ZUMA studies revealed blood–brain barrier disruptions in patients who succumbed to cerebral edema subsequent to CAR T treatment.56 119 In the ZUMA-2 trial, one case of grade 4 cerebral edema occurred; the patient had a full recovery and was in complete remission at 24 months of follow-up with no unresolved neurological sequela.14 Blood–brain barrier pathology was also reported in a patient who developed edema as a feature of CRS after CD19-directed CAR T therapy.126 Disruption of the blood–brain barrier may allow systemic cytokines including IFN-γ to leak into the CSF, thus inducing vascular pericyte stress and the secretion of endothelial cell-activating cytokines. Therefore, there is speculation that patients with evidence of endothelial cell activation before lymphodepletion may be at increased risk of ICANS.56 A report has also surfaced of severe neurotoxicity consistent with posterior reversible encephalopathy syndrome in a phase I trial of CART-BCMA for multiple myeloma (NCT02546167), which resolved after high-dose methylprednisolone and cyclophosphamide.111

In roughly one-third of patients treated with CD19-directed CAR T cells, cytopenias persisting for >1 month have been reported.7 23 57 58 One report has described prolonged cytopenias persisting for 15.2–21.7 months after CD19 CAR T cell treatment122 and another reported that 10% of adult lymphoma patients that remain in remission after axicabtagene ciloleucel may continue to experience grade 3–4 neutropenia 1 year after the therapy.131 Of note, the latter study of 85 patients in the standard of care setting noted that while CD8⁺ counts rapidly recovered, CD4⁺ T cells decreased from baseline and were persistently low with a median CD4 count of 155 cells/µL for those remaining in remission 1 year after axicabtagene ciloluecel.131 For other therapies, hematological recovery following lymphodepletion generally occurs more rapidly. For example, 97% of patients receiving BCMA-targeting bb2121 CAR T cells who experienced grade 3 or higher cytopenias recovered to an absolute neutrophil count (ANC) of at least 1000 cells/µL within 1 month. In that trial, the median time from infusion to recovery of an ANC of at least 1000 cells/µL was 1.3 weeks (95% CI 1.0 to 1.4) and recovery to a platelet count of at least 50,000/µL occurred in a median of 2.0 weeks (95% CI 1.4 to 8.4).16 132

One study identified a biphasic pattern to cytopenias following CD19 CAR T cell therapy. In the trial, 27 of 29 responding patients experienced cytopenias later than 21 days postinfusion. Of the 22 patients who experienced late neutropenia, 15 showed two distinct nadirs in absolute neutrophil counts with one trough shortly after lymphodepletion and a second occurring more than 40 days after infusion. A similar biphasic pattern occurred in the 22 patients who experienced grade 3 thrombocytopenia, with 10 patients displaying an initial recovery of platelet counts before a second nadir between 20 and 40 days after infusion.133 In another analysis, baseline platelet counts of <75,000 and the early onset of CRS on day 0 or day +1 following infusion were associated with a higher likelihood of development of severe and prolonged cytopenias, which in some cases required therapy as for graft failure following allo-HSCT.132

A correlation between late (more than 21 days after infusion) thrombocytopenia, neutropenia and anemia was observed (p=0.018 for thrombocytopenia and neutropenia, p<0.0001 for thrombocytopenia and anemia, and p=0.05 for anemia and neutropenia) in a phase Ib/II trial of CD19 CAR T cells for 38 children and adults with RR B cell cancers. In that study, prior HSCT (p=0.0015, 0.0083, and 0.02 for anemia, thrombocytopenia, and neutropenia) and higher CRS grade (p=0.003, 0.018, and 0.04 for late anemia, thrombocytopenia, and neutropenia) predicted the development of cytopenias.133 Grade 3 or 4 CRS was also linked to delayed hematological recovery in 133 patients with RR B-ALL, CLL, or NHL who underwent lymphodepletion chemotherapy followed by infusion of CD19 CAR T cells. Examination of bone marrow biopsies from patients with grade 4 CRS showed no evidence of increased hemophagocytosis that might contribute to the delayed recovery; however, pretreatment bone marrow disease burden and a high number of prior therapies were associated with prolonged cytopenias in the study population.57

The package insert for tisagenlecleucel recommends against using myeloid growth factors, particularly GM-CSF during the first 3 weeks after cell infusion or until CRS has resolved because GM-CSF may theoretically aggravate CRS.1 6 Some centers recommend filgrastim (recombinant G-CSF) for all patients with ANCs lower than 500/µL.28 31 Without growth factor support, thrombocytopenia and anemia may resolve more slowly than neutropenia. One retrospective analysis of 32 RR DLBCL patients treated with CD19 CAR T cells observed median times to neutrophil, platelet, and hemoglobin recoveries of 11 days (range 5–218 days), 59.5 days (range 4–241 days), and 76 days (range 0–218 days), respectively. Fifteen patients in the cohort were treated with filgrastim.134

A baseline measurement of cardiac function is important. Echocardiography is the principal technique for evaluating cardiac toxicity in patients with cancer and three-dimensional echocardiography-derived LVEF correlates excellently with MRI findings, while two-dimensional speckle tracking echocardiography-derived strain can detect changes in myocardial mechanics before changes in LVEF occur.138–140 Historically, equilibrium radionuclide angiography/MUGA has been a preferred method for serial assessment of LVEF in adult patients undergoing cardiotoxic chemotherapy due to high reproducibility, low interobserver and intraobserver variability, and extensive validation literature. Routine use of MUGA has markedly diminished in recent years because of the relative ease of use and low cost of echocardiogram.141 Additionally, MUGA exposes patients to radiation through the use of 99m-technetium labeled red blood cells, making echocardiogram the preferred modality for monitoring adult and pediatric patients.142

Several cardiac biomarkers have established utility in the detection of cardiac injury with cancer therapies and have also been validated for their prognostic value both in symptomatic and asymptomatic patients free of cancer. For example, the N-terminal pro-brain natriuretic peptide (NT-proBNP) investigation of dyspnea in the emergency department study of 600 patients who presented in the emergency department with dyspnea found that increased serum NT-proBNP was the strongest independent predictor of a final diagnosis of acute congestive heart failure (OR 44; 95% CI 21.0 to 91.0, p<0.0001) and NT-proBNP testing alone was superior to clinical judgment alone (p=0.006).143 Cardiac troponins are the gold standard biomarker for myocardial injury, and high-sensitivity assays allow for the improved detection of heart tissue damage even in the absence of overt signs and symptoms.144 145 However, an elevated troponin may occur independently of acute myocardial infarction, and myriad etiologies have been linked to troponin release, including pulmonary embolism, heart failure, myocarditis, end-stage renal disease, tachycardia, strenuous exercise, septic shock, and treatment with cardiotoxic chemotherapy such as doxorubicin and with immune checkpoint inhibitor therapy.140 146–149

Many patients treated with CAR T cells have undergone several prior lines of therapy, including chemotherapy with cardiotoxic agents. According to some estimates, 57%–70% of elderly lymphoma patients and 50%–60% of childhood cancer survivors are treated with an anthracycline-containing regimen.149 Anthracyclines, especially doxorubicin, can cause irreversible, dose-dependent cardiac injury, and the risk for future complications increases with cumulative doses. The onset of a decreased LVEF may occur while a patient is on treatment with anthracyclines or several years later.149–151 Serial serum troponin measurement in 78 patients with hematological malignancies undergoing 142 treatment cycles, including various anthracyclines, revealed delayed subclinical myocardial damage even after minor anthracycline exposure, with peak levels observed on median day +21.5 (range day +6 to day +35) after initiation of anthracycline therapy. Follow-up echocardiography in 28 patients showed a greater decrease in LVEF in troponin-positive patients compared with the troponin-negative group (10% vs 2%; p=0.017).124

Elevated troponins and BNP/NT-proBNP have been associated with increased all-cause mortality and have been shown to predict heart failure in patients with cancer undergoing chemotherapy.152 153 Increased troponin levels correlated with decreased LVEF in patients with acute myeloid leukemia (AML) undergoing induction chemotherapy with anthracycline-containing regimens.154 Among patients undergoing CAR T cell therapy, elevated serum troponins have been observed prior to cardiac events and post-CAR T cell infusion.53

In addition to prior therapies, the risk factors for cardiac toxicity with treatment are likely similar to those for the development of severe CRS, specifically, disease burden and peak CAR T cell expansion.27 28 137 155 156 The pathophysiology responsible for cardiac toxicity in the context of CAR T cell therapy has not been extensively studied; however, the inflammatory cytokine profile seen with CRS, specifically elevated levels of IL-6, has also been implicated in causing myocardial dysfunction in cases of septic shock.157 One of the hallmarks of CRS is hypotension,28 29 31 85 90 155 156 which may lead to distributive shock, myocardial injury, and arrhythmias. Medications that increase bleeding risk could accelerate the onset of life-threatening shock.

The underlying cardiovascular substrate, beyond cardiovascular function, is also likely a risk factor for the development of cardiac injury and cardiovascular events with CAR T. Specifically, many adult patients receiving CAR T cells may be elderly and have underlying cardiovascular risk factors such as hypertension and diabetes mellitus and may have established cardiovascular diseases such as heart failure and coronary artery disease. There are limited data specific to CAR T, but in a single retrospective study of 137 subjects, the rates of these cardiovascular risks were higher in those that subsequently developed cardiac injury.91

Patients with high preinfusion disease burden are also at risk of developing TLS,7 8 10 which may cause cardiac complications secondary to the flood of metabolites and electrolytes, including uric acid released into circulation.92 In particular, elevated serum uric acid has been implicated as a risk factor for cardiovascular events, including acute myocardial infarction, angina, and heart failure.158–160

Coagulopathies frequently occur in the context of CRS, and disseminated intravascular coagulation has been reported at rates of more than 2% for the approved CAR T cell products.161 Patients with cancer are at higher risk of thromboembolism, and clots are a leading cause of death in patients undergoing chemotherapy.162 163 The goal of anticoagulation therapy is generally to prevent clot formation at the lowest possible dose of anticoagulant medication, and management in patients with cancer may be challenging because some agents are long acting, interact with other drugs and have narrow therapeutic windows. In particular, warfarin (Coumadin) has a serum half-life of, on average, roughly 40 hours and the duration of effect is 2–5 days.164 165 Current major guidelines from the European Society for Medical Oncology, the National Comprehensive Care Network and International Clinical Practice recommend low molecular weight heparin over warfarin for thrombosis prophylaxis in patients with cancer.166–169

The risk for bleeding complications in patients treated with anticoagulants is considerable. A meta-analysis of 33 studies involving 4374 patient-years of oral anticoagulant therapy found that the case-fatality rate of major bleeding was 13.4% (95% CI 9.4% to 17.4%), and the rate of intracranial bleeding was 1.15 per 100 patient-years (95% CI 1.14 to 1.16 per 100 patient-years).170 Multiple studies have demonstrated that the presence of malignant disease significantly increases risk of major bleeding after anticoagulation therapy with warfarin.171 172 Antiplatelet blood thinners such as aspirin also increase bleeding risks. Additionally, aspirin, and other cyclooxygenase (COX) inhibitors, have been demonstrated to suppress T cell activation by inhibiting T cell receptor signaling to p38 Mitogen-activated protein (MAP) kinase,173 174 though no large controlled trials have assessed whether COX-inhibition negatively affects the efficacy of CAR T cell therapy.

Hypogammaglobulinemia has been reported across trials, although the rates have varied, likely due to inconsistent definitions, follow-up times, patient ages and disease states. Further complicating matters, many patients who undergo CAR T cell therapy have undergone prior HSCT without revaccination and are thus beginning treatment with a limited antibody repertoire. Younger patients may be more susceptible to hypogammaglobulinemia, as total serum IgG levels gradually increase at early ages, with adult levels of IgG₁ and IgG₃ subtypes seen between ages 5 and 10 years and IgG₂ not reaching maximum levels until adulthood,176 reflecting increasing plasma cell mass. The direct link between B cell depletion and hypogammaglobulinemia is difficult to establish because long-lived plasma cells are CD19 negative175 177 178 and therefore should not be direct targets for elimination by CD19 CAR T cells. Plasma cells persisting for at least 746 days in the absence of B cells has been reported in one patient treated with tisagenlecleucel, and the same study observed stable levels of pathogen-specific IgG despite prolonged B cell aplasia and decreasing total serum IgG in two adult patients.54

Immunoglobulin replacement therapy is FDA-approved for the treatment of primary immunodeficiencies as well as select few secondary antibody deficiencies, including hypogammaglobulinemia in CLL.179 180 A number of off-label uses of immunoglobulin supplementation have become incorporated into routine clinical practice, with varying levels of supporting evidence. Although long-term B cell aplasia and hypogammaglobulinemia can persist for several years after CAR T cell infusion,55 181 no controlled studies have demonstrated definitive benefits for IgG supplementation. In trials where patients did receive IgG supplementation, no serious infections were reported after the initiation of replacement therapy.9 10 55 One study of 28 patients with persistent B cell aplasia and agammaglobulinemia following CD19 CAR T cell therapy, where patients were transitioned from intravenous to subcutaneous immunoglobulin replacement at a median of 11.5 months (range 4–20), found that increasing serum IgG level was significantly associated with a lower rate of sinopulmonary infection (p=0.0072).182

All IgG preparations available in the USA are made from 10,000 to 50,000 units of plasma pooled from 3000 to 10,000 healthy blood donors. Donors are carefully screened for blood-borne pathogens, including HIV, hepatitis A, B, and C viruses, and parvovirus B19, and at least two distinct viral inactivation steps are used by all manufacturers. The adverse events associated with IgG infusions are typically transient, infusion-related reactions. Intravenous and subcutaneous formulations have been developed, and typical starting doses are in the range of 400–800 mg/kg every 3–4 weeks intravenously or 100–200 mg/kg/week subcutaneously.183 The decision to administer immunoglobulin replacement therapy may be constrained by product availability, as an ongoing national shortage has limited supplies at most centers at the time of writing this article.184

Additional toxicities associated with primary IEC induction have been described, including allergic reaction, sepsis, and TLS. One patient developed anaphylaxis and cardiac arrest within minutes of completing a third infusion of T cells that had been transduced with a CAR derived from a murine antibody to human mesothelin,185 although this appears to be an isolated incident.

More commonly, several studies have reported TLS after CAR T cell treatment, a pathology arising due to dramatic electrolyte and metabolite imbalances after widespread release of cancer cell contents into the bloodstream, which can cause kidney damage due to elevated levels of serum uric acid and LDH.92 To prevent TLS in patients from undergoing CAR T therapy, prophylactic allopurinol is sometimes administered prior to infusions.86 186 However, several trials have reported TLS, despite preinfusion allopurinol. Symptoms typically resolved after fluid resuscitation and administration of the recombinant urate oxidase drug rasburicase.7 8 10 186 In an analysis of 328 pediatric ALL patients undergoing chemotherapy, factors predictive of TLS were male sex (OR 1.8; p=0.041), age ≥10 years (OR 4.5; p<0.0001), splenomegaly (OR 3.3; p<0.0001), mediastinal mass (OR 12.2; p<0.0001), T cell immunophenotype (OR 8.2; p<0.0001), CNS involvement (OR 2.8; p=0.026), serum LDH ≥2000 U/L (OR 7.6; p<0.0001), and elevated white blood cell (WBC) counts (≥20×10⁹/L) (OR 4.7; p<0.0001).117 In two separate studies of TLS in patients with AML, predictive risk factors were male sex, LDH levels above normal values, creatinine>1.4 mg/dL, uric acid>7.5 mg/dL, and WBC count>25×10⁹/L.187 188

Fatal sepsis has been reported after CAR T cell therapy. A 69-year-old man with refractory CLL, despite negative blood cultures at the time of infusion, died of acute renal failure consistent with sepsis after administration of 19-28z CAR T cells and cyclophosphamide.189 An analysis of infection events in 19 patients during the first 30 days after CD19 CAR T cell therapy developed a prediction model based on three cytokines (IL-8, IL-1β, and IFN-γ) that could predict life-threatening infection with high sensitivity (training 100.0%; validation 100.0%) and specificity (training 97.6%; validation 82.8%). During the study, a characteristic pattern of ‘double peaks’ of IL-6, where serum levels of the cytokine spiked, dropped and then rose again, appeared in 9 of 11 patients with grade 4–5 infections.190