Lost in translation: why cytokine release syndrome reporting fails clinical trials

Abstract

Cytokine release syndrome (CRS) is among the most frequently reported adverse events in trials of immune-cell-engaging therapies, yet its identification, grading, and reporting remain inconsistent across sponsors and investigators, driving overly conservative dose-escalation decisions, premature dose-limiting toxicity calls, unwarranted tocilizumab or corticosteroid use, and confounded cross-trial safety comparisons. Cytokine release symptoms, transient findings reflecting expected immune-cell engagement, are distinguished from CRS as a formal syndrome and from cytokine storm as its most severe manifestation. Current practice is associated with several pitfalls, including systematic exclusion of CRS-mimicking differential diagnoses, misclassification of grade 2 hypotension absent formal vasoplegia testing, the American Society for Transplantation and Cellular Therapy (ASTCT) framework’s failure to incorporate organ dysfunction beyond hypotension and hypoxia, and the frequent misattribution of isolated, self-limited transaminitis to severe CRS. While ASTCT remains the most practical standard available, closing these reporting gaps is essential for investigators, sponsors, and health authorities to ensure safety evaluation and decisions reflect genuine toxicity rather than expected pharmacology.

Introduction

Cytokine release syndrome (CRS) has become one of the most frequently reported adverse events in early-phase trials of T-cell-engaging therapies, including bispecific T-cell engagers, chimeric antigen receptor (CAR) T-cell products, and immuno-cytokines.^1–3 Developed initially to characterize CRS in patients receiving CAR T-cell therapies for hematologic malignancies, the American Society for Transplantation and Cellular Therapy (ASTCT) grading criteria⁴ provided a much-needed framework for standardized toxicity assessment across clinical trials, institutions, and therapeutic platforms. The CRS criteria were largely reaffirmed in a 2026 ASTCT consensus update.⁵ However, the rapid expansion of immune-engaging therapies across solid and hematologic malignancies has highlighted important limitations of the current framework, with substantial variability in the recognition, grading, and reporting of cytokine-mediated toxicities across investigators and sponsors. This has practical consequences: overly conservative dose-escalation strategies, premature dose-limiting toxicity determinations, unwarranted use of tocilizumab or corticosteroids, and confounded cross-trial comparisons of safety profiles—all of which shape how these agents will ultimately be managed once approved. A recent call for more granular, symptom-level reporting of cytokine-related adverse events underscores that this remains an active and unresolved concern for the field, not merely a matter of semantics.⁶ Since then, the ASTCT grading criteria have emerged as the predominant consensus framework for both clinical trial reporting and treatment guidelines. Building on earlier proposals for granular toxicity reporting, this commentary examines pitfalls inherent in the current ASTCT system that warrant clarification (illustrated by a clinical case provided in the online supplemental files 1 and 2) and calls for a more refined consensus framework to improve the grading, reporting, and interpretation of toxicities associated with immune cell-engaging therapies.

Defining the spectrum of cytokine release

A central source of confusion in the literature and in clinical practice arises from imprecise, interchangeable use of terms describing cytokine-related phenomena. Three entities warrant separate definitions. Cytokine release symptoms (CR-Sx) are individual, usually transient, clinical findings resulting from cytokine elevation following administration of an immune cell engager (fever, chills, rash, hypotension, hypoxia, or transaminase elevation). CRS denotes the constellation of these findings meeting the formal ASTCT consensus definition. Cytokine storm (CyS) describes uncontrolled, excessive cytokine release culminating in a systemic hyperinflammatory state with potential multiorgan dysfunction or failure; in practice, this corresponds to ASTCT grade 3–4 CRS (online supplemental figure 1). Recognizing this gradient—from isolated, self-limited symptoms, through a defined syndrome, to a life-threatening storm—is essential to avoid conflating a mild, expected pharmacologic effect with a genuine safety signal. Whether CyS represents progressive worsening along a continuum or an on/off phenomenon triggered by specific host, tumor, or therapeutic factors remains unknown. The 2006 first-in-human trial of the anti-CD28 superagonist TGN1412, in which six healthy volunteers developed multiorgan failure within hours of infusion, remains a landmark illustration of the far end of this spectrum.⁷

The ASTCT consensus grading system

Published by Lee and colleagues in 2019, the ASTCT consensus grading system remains the reference standard for classifying CRS severity after immune effector cell therapy.⁴ ASTCT defines CRS as a supraphysiologic response to any therapy activating or engaging endogenous or infused T cells and/or other immune effector cells; signs can be progressive, must include fever at onset, and may include hypotension, capillary leak with hypoxia, and end-organ dysfunction. Grading is anchored to three parameters—fever, hypotension, and hypoxia—with severity increasing according to need for vasopressor and supplemental oxygen support. In 2026, an ASTCT consensus group of 36 experts updated grading across immune effector cell toxicities but retained the CRS criteria, clarifying only that once antipyretic or anticytokine therapy (such as tocilizumab or corticosteroids) is given, fever is no longer required and grading follows hypotension or hypoxia.⁵ Although broadly adopted for its reproducible, mechanistically grounded structure, the framework was developed for CAR T-cell and bispecific antibody therapy in hematologic malignancies, and several assumptions do not map cleanly onto the more heterogeneous population receiving immune-cell engagers in solid tumor trials.

CR-Sx versus CRS: a critical distinction

The distinction between cytokine release as a pharmacodynamic signal and CRS as a pathologic syndrome carries substantial implications for trial conduct. Lower-grade cytokine release is an expected pharmacologic effect—predictable, mechanistic, a direct consequence of immune-cell activation. It is dose-dependent and exposure-dependent, generally transient and reversible, and in many respects defines the upper boundary of biological activity for a regimen; it is evidence that the intended pharmacology is occurring. Nevertheless, when clinically significant, CR-Sx require appropriate symptomatic management and early intervention may be appropriate to improve patient safety. CyS or grade 3–4 CRS, by contrast, is a pathologic syndrome characterized by multiorgan dysfunction driven by uncontrolled cytokine amplification, manifesting as hemodynamic instability, hypoxia, and end-organ damage—a genuine loss of immune control requiring urgent intervention with corticosteroids, tocilizumab, and intensive care.

This distinction matters because current safety reporting frequently over-attributes cytokine-elevation-related symptoms to CRS, labeling manageable immune activation as unacceptable toxicity. Consequences include overly conservative development strategies, premature dose-limiting toxicity decisions affecting escalation and step-up dosing designs, excessive tocilizumab or steroid use, and misguided cross-program comparisons. Inconsistent reporting at the grade 2 boundary also complicates translational and biomarker analyses. Addressing this requires re-educating investigators, sponsors, and regulators; refining CRS grading specifically for immune-cell engagers; and enabling rational, less conservative outpatient development pathways where appropriate—consistent with the proposed shift toward granular, intervention-based reporting rather than syndromic grouping alone.⁶ The 2026 update likewise cautions against adopting these scales as dose-limiting toxicity criteria without considering disease-specific risk tolerance.

Adjudicating grade 2 hypotension: a common source of misclassification

Hypotension is the most frequent and subjective source of CRS misclassification. The central question is whether observed hypotension is asymptomatic and isolated: under ASTCT criteria,⁴ a Common Terminology Criteria for Adverse Events (CTCAE) V.5.0 grade 1 hypotension⁷ would be classified as grade 2 CRS. Systematic evaluation of vasoplegia could help to adjudicate severity correctly—patients should receive a standardized fluid bolus challenge before grading is finalized. If blood pressure normalizes, vasoplegia is absent and the event should be graded grade 2 CRS. If the patient fails to respond, vasoplegia is present, vasopressor support is required, and the event should be upgraded to grade 3 CRS. Overlap frequently occurs at this boundary between non-symptomatic isolated grade 1 hypotension, hypotension that responds well to a fluid challenge, or hypotension empirically hydrated without formal vasoplegia testing—a scenario that risks overestimating severity toward grade 3. An isolated CTCAE grade 1 hypotension for which the treating physician sees no clinical indication for a fluid challenge should not automatically be counted as grade 2 CRS.

Organ dysfunction beyond hypotension and hypoxia

A frequently underappreciated limitation of the ASTCT framework is its failure to formally incorporate organ dysfunction beyond hypotension and hypoxia.⁴ Consequently, cardiac, hepatic, renal, coagulopathic, or neurological dysfunction does not alter CRS grade, even when severe. This feature was retained rather than revised in the 2026 update, which specifies that organ toxicities are graded separately.⁵ Organ injury remains clinically essential and should be reported separately using standard CTCAE V.6.0 criteria.⁸ Whenever CRS symptoms occur, clinicians should systematically evaluate for associated organ injury; because organ injury evolves dynamically, serial monitoring at defined intervals (eg, 4, 12, and 24 hours) is warranted. The current ASTCT definition also requires fever to qualify as CRS, yet organ dysfunction may occur in its absence and is therefore not captured. A further limitation of the current ASTCT framework is its requirement for fever at CRS onset. While this criterion enhances specificity, it may fail to capture patients who develop clinically meaningful organ dysfunction without fever. The 2026 ASTCT update partially addresses this issue by acknowledging that CRS may persist or present without fever following antipyretic or anticytokine treatment.⁵ However, in an era in which antipyretic premedication is routinely used, the continued centrality of fever in the definition of CRS warrants reconsideration.

One challenging case often observed following treatment with immune-cell engager is isolated transaminitis. Isolated transaminitis refers to an elevation of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) occurring within 24–48 hours of treatment, in the absence of clinical symptoms, hyperbilirubinemia, coagulopathy, or other organ dysfunction.⁹ This phenomenon may reflect a direct or indirect pharmacodynamic effect (on-target/off-tumor, cytokine-mediated, or liver immune cell activation) rather than organ dysfunction due to ischemic or microthrombotic hepatitis. Such elevations are often wrongly considered severe CRS; however, in patients remaining clinically stable with downtrending enzymes, this pattern more likely reflects a CR-Sx than true liver failure. Monitoring enzyme kinetics is important—the elimination half-life of AST is approximately 17 hours and of ALT approximately 47 hours—to anticipate the expected resolution pattern. Conversely, worsening transaminase elevation or emergence of rising bilirubin, coagulopathy, or other organ dysfunction should be considered liver organ failure associated with severe CRS.

Differential diagnosis of cytokine release syndrome

Multiple clinical entities can mimic CRS, including infection, tumor progression or disease-related fever, infusion-related reactions, tumor flare reaction, on-target/off-tumor toxicity, hemophagocytic lymphohistiocytosis or macrophage activation syndrome, and tumor lysis syndrome.^{2 6 10 11} While this principle has not yet been incorporated into the grading framework itself, the 2026 ASTCT update acknowledges that fever should not be automatically attributed to CRS.⁵ Rather, alternative etiologies must be carefully considered and reasonably excluded before a diagnosis is made. In practice, clinicians should systematically evaluate these possible diagnoses and may revise the final syndromic diagnosis a posteriori, illustrating that assessing individual symptoms (CR-Sx) should take precedence over syndromic labeling. Currently, no robust data support using biological markers such as interleukin-6, C-reactive protein, or ferritin to formally establish a definitive diagnosis.

Conclusion

Despite its widespread adoption and reaffirmation in the 2026 update, the ASTCT grading system carries several unresolved limitations that merit consideration as the field moves toward refinement. Some of these challenges have been raised in recent calls for a more granular, T-cell-engager-specific reporting framework.⁶ Key questions include whether CRS should be redefined to encompass only grade 3–4 events, effectively equating CRS with CyS; persistent ambiguity around hypotension and hypoxia assessment at the grade 2 boundary; uncertainty over whether fever should remain obligatory, given that organ dysfunction and other manifestations of cytokine release may occur in its absence; the absence of a mechanism for integrating multiorgan dysfunction into the grade itself; and debate over whether transient, isolated transaminitis belongs within the CRS spectrum (figure 1).

Figure 1. Limitations of the ASTCT consensus grading system despite widespread adoption, the ASTCT system carries five unresolved limitations: redefining CRS to grade 3–4 only, grade 2 hypotension/hypoxia ambiguity, fever as an obligatory criterion, no mechanism for multiorgan dysfunction, and the status of isolated transaminitis. ASTCT, American Society for Transplantation and Cellular Therapy; CRS, cytokine release syndrome.

Addressing these issues is a priority for all clinical research stakeholders, including health authorities, sponsors, and clinical investigators, as reporting inconsistencies can directly affect dose-escalation strategies, dose-limiting toxicity determinations, use of tocilizumab and corticosteroids, and cross-trial safety comparisons, ultimately impacting patient care and safety. This commentary serves as a call to action for stakeholders to work collaboratively toward a consensus approach that closes these reporting gaps while preserving the standardization achieved by the ASTCT criteria. In the absence of a validated alternative, the ASTCT system remains, in practical terms, the least imperfect option available; however, further refinement will be essential to ensure that safety assessments, regulatory decisions, and drug development strategies are informed by a more precise, reproducible, and meaningful evaluation of immune activation-related toxicities.