Liver Injury After CAR T-Cell Therapy May Be Driven by Cytokine Storm, Landmark Analysis Finds

Chimeric antigen receptor T-cell therapy has transformed the outlook for patients with certain blood cancers, reprogramming a person’s own immune cells to hunt down and destroy malignant clones. But as the technology has moved from experimental centers into routine clinical practice, clinicians have begun cataloging a growing list of side effects that extend well beyond the therapy’s best-known toxicities. Now, a pharmacovigilance study drawing on nearly five million adverse-event reports has turned its attention to one of the least understood of these: liver injury. The analysis, published in Annals of Hematology, suggests that most hepatic damage reported after CAR T-cell infusion occurs early, clusters with cytokine release syndrome, and may not represent a distinct, therapy-specific liver disease at all.

The research team, led by Saikat Mandal and Guruprasad Aithal of the Nottingham Digestive Diseases Centre at the University of Nottingham, together with colleagues at Hull, Sheffield, Belfast and Imperial College London, mined the US Food and Drug Administration’s Adverse Event Reporting System, known as FAERS, covering the period from the third quarter of 2016 through the first quarter of 2026. Out of 4,700,843 total reports, 19,820 identified a CAR T-cell product as the primary suspect drug, and 734 of those contained at least one hepatic adverse event. To capture liver injury comprehensively, the investigators built a prespecified dictionary of 244 MedDRA-coded terms spanning everything from mild transaminase elevations to frank hepatic failure.

The timing of the reported liver events proved strikingly consistent. The median onset was just five days after infusion, and 73.5 percent of all hepatic reports occurred within the first two weeks. That early window is precisely when cytokine release syndrome, the systemic inflammatory cascade triggered by activated T cells, reaches its peak intensity. When the researchers examined the text of the hepatic reports, they found that CRS was co-mentioned in 77.0 percent of them, whereas terms from the haemophagocytic lymphohistiocytosis spectrum, including the related immune effector cell-associated haemophagocytic syndrome, appeared in only 9.5 percent. This imbalance immediately suggested that the inflammatory storm, rather than a rarer hyperinflammatory syndrome, might be the dominant driver of liver-related signals.

To test that hypothesis formally, the team deployed a battery of classic pharmacovigilance statistics: the reporting odds ratio, the proportional reporting ratio, the information component from a Bayesian confidence propagation neural network, and the empirical Bayes geometric mean. Each metric compares how often a given event is reported with a suspect drug against how often it appears with all other drugs, and requiring agreement across multiple methods helps filter out statistical noise. In the pooled analysis, hepatic events showed a modest but detectable association with CAR T-cell products, with a reporting odds ratio of 1.36. The crucial experiment came next: when the researchers excluded every report that also mentioned cytokine release syndrome, the pooled signal collapsed to 0.54, effectively falling below the threshold expected by chance. Excluding HLH-spectrum reports, by contrast, left the association intact.

That asymmetry is the analytical heart of the study. If liver injury were an independent manifestation of CAR T-cell toxicity, stripping out CRS reports should have left the signal largely undiminished. Instead, the signal vanished, implying that the liver abnormalities captured in spontaneous reporting databases are, in the great majority of cases, a downstream feature of the cytokine storm rather than a separate immune-mediated process. This distinction matters clinically, because the two scenarios demand different responses. CRS is typically managed with the interleukin-6 receptor antibody tocilizumab and supportive care, whereas suspected immune-mediated drug-induced liver injury or HLH may require corticosteroids, etoposide, or other immunosuppressive strategies, and misattributing the cause can delay appropriate treatment.

The product-level analysis added a provocative wrinkle. Among the individual CAR T-cell constructs, idecabtagene vicleucel, the anti-BCMA therapy approved for multiple myeloma, showed the most consistent hepatic reporting pattern across all four statistical methods. More importantly, it retained a supportive signal even after the researchers excluded reports co-mentioning CRS, with a reporting odds ratio of 2.22 and a 95 percent confidence interval of 1.38 to 3.57. In other words, a residue of liver-injury reporting persisted that could not be explained away by the cytokine storm alone. No other product showed comparable persistence, and no product retained four-method positivity beyond the 14-day mark, reinforcing the overall picture of early, CRS-linked injury with one notable exception.

Phenotype analysis deepened the product-specific picture. By clustering the individual MedDRA terms within hepatic reports, the researchers identified mixed hepatocellular and cholestatic reporting patterns that differed across products, a pattern reminiscent of how classical drug-induced liver injury presents in distinct biochemical signatures. Cholestatic patterns, in which bile flow is impaired and alkaline phosphatase rises disproportionately, can point toward different mechanistic pathways than pure hepatocellular injury. The authors suggest these residual, CRS-independent patterns, particularly for idecabtagene vicleucel, warrant prospective clinical evaluation with protocolized liver biochemistry monitoring, rather than reliance on the fragmented and potentially biased data of spontaneous reporting.

Spontaneous reporting databases are powerful but imperfect instruments. FAERS aggregates millions of reports, yet it suffers from underreporting, reporting bias, duplicate submissions, and the well-known Weber effect, in which adverse events are reported more frequently in the period immediately after a drug’s approval. The researchers addressed these vulnerabilities with several safeguards: they deduplicated reports, restricted the analysis to records naming a CAR T-cell product as the primary suspect, and performed sensitivity analyses examining delayed-onset events and specific hepatic phenotypes. They also sought directional confirmation in a second database, Canada Vigilance, where the proportion of hepatic reports co-mentioning CRS was similar at 75.0 percent. However, only four hepatic adverse-event reports existed in the Canadian data, far too few for meaningful product-level comparison, underscoring how sparse real-world evidence remains for some of these newer toxicities.

The study’s framing of the differential diagnosis is one of its most useful contributions to practicing clinicians. Liver dysfunction after CAR T-cell infusion can arise from at least three distinct mechanisms: the inflammatory milieu of cytokine release syndrome, the hyperinflammatory HLH-like syndrome that shares laboratory features with familial hemophagocytic lymphohistiocytosis, and a conventional immune-mediated drug-induced liver injury resembling idiosyncratic hepatotoxicity from small-molecule drugs. Each mechanism has a different temporal signature, a different biomarker profile, and a different treatment algorithm. By quantifying how frequently each appears in real-world reporting, the Nottingham-led team has given clinicians an evidence-based prior for interpreting liver enzyme abnormalities in the days and weeks following infusion.

The authors are careful to emphasize the limits of their findings. Pharmacovigilance signals identify statistical associations, not causation, and the absence of denominator data means incidence rates cannot be calculated from FAERS or Canada Vigilance. The persistent idecabtagene vicleucel signal, while intriguing, is based on a modest number of reports and requires confirmation in prospective cohorts with systematic liver monitoring. Still, the study arrives at a moment when CAR T-cell therapy is expanding rapidly beyond academic centers and into community oncology, meaning more clinicians than ever will confront unexplained liver enzyme elevations in treated patients. For most, the message is reassuring: hepatic events are predominantly early, temporally tied to the cytokine storm, and likely to resolve as the inflammatory phase subsides. For a subset of patients, particularly those receiving BCMA-directed therapy, the possibility of a CRS-independent liver injury remains open, and the authors argue that only carefully designed prospective studies can close it. The work was supported by the NIHR Nottingham Biomedical Research Centre and UK Research and Innovation, and the underlying data are publicly accessible, allowing other groups to interrogate and extend the analysis as the CAR T-cell treated population continues to grow.

Subject of Research: Hepatic adverse-event reporting patterns following CAR T-cell therapy analyzed through pharmacovigilance databases

Article Title: Hepatic adverse-event reporting after CAR T-cell therapy: a pharmacovigilance analysis of FAERS and Canada Vigilance

Article References: Mandal, S., Maji, M., Dhali, A., & Aithal, G. (2026). Hepatic adverse-event reporting after CAR T-cell therapy: a pharmacovigilance analysis of FAERS and Canada Vigilance. Annals of Hematology. https://doi.org/10.1007/s00277-026-07318-0

Image Credits: AI Generated

DOI: 10.1007/s00277-026-07318-0

Keywords: CAR T-cell therapy, hepatic adverse events, cytokine release syndrome, haemophagocytic lymphohistiocytosis, pharmacovigilance, FAERS, Canada Vigilance, drug-induced liver injury, idecabtagene vicleucel, immunotherapy toxicity, liver injury, hematology