Immune checkpoint inhibitors (ICIs) have revolutionized the therapeutic landscape for advanced cancer patients, including melanoma, non-small cell lung cancer, esophageal cancer, gastric cancer, and hepatocellular carcinoma (1–6). ICIs can enhance the activity of immune cells and promote the immune system’s attack on tumor cells by inhibiting the immune checkpoint molecules such as cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), programmed death 1 (PD-1), and programmed cell death ligand 1 (PD-L1). The activation of the immune system upon ICI treatment may initiate the inflammatory side effects, which are defined as immune-related adverse events (irAEs) (7) with a higher incidence in patients receiving CTLA-4 (53.8%) treatment when compared with those receiving ICIs targeting PD-1 (26.5%) or PD-L1 (17.1%) (8). IrAEs have the potential to target any organ system, with the most frequent toxicities for the integumentary, gastrointestinal, and endocrine systems. The manageable lower-grade irAEs with minimal treatment intervention of ICI treatment are associated with a favorable prognosis (9, 10), while higher-grade irAEs with potentially life-threatening risk may need permanent treatment discontinuation (11). The underlying mechanisms of irAEs are still being investigated. The potential mechanisms include increased T-cell activity, elevated levels of pre-existing autoantibodies, heightened inflammatory cytokine levels, and enhanced complement-mediated inflammation (7).

Liver-related irAEs are referred to as immune-mediated hepatotoxicity (IMH), with severe IMH defined as those meeting Common Terminology Criteria for Adverse Events (CTCAE) grade 3 or higher (12, 13). Notably, severe IMH occurs in approximately 1%–2% of patients undergoing anti-PD-1 or anti-PD-L1 therapy, whereas the incidence can increase to as high as 10% in patients receiving combination therapy with anti-CTLA-4 and anti-PD-1 agents (14). The patients with pre-existing liver injury, such as viral hepatitis B, metabolic disorders, and liver metastases, are more prone to IMH incidence (15). The mechanisms by which hepatocytes are targeted in IMH remain incompletely understood. However, recent studies have shown that liver tissue from patients with IMH exhibits a pattern of immune cell infiltration similar to that seen in peripheral tissues of patients treated with ICIs. This infiltration is characterized by the accumulation of CD8+ T cells and CCR2+ macrophages (16). Therefore, the extensive infiltration and excessive activation of lymphocytes upon ICI treatment may represent a critical underlying mechanism contributing to the development of IMH (17).

The multiple guidelines for managing IMH from the American Society of Clinical Oncology (ASCO) (14) and the National Comprehensive Cancer Network (NCCN) (18) were recommended, but all current recommendations are mainly based on expert consensus rather than evidence-based medicine (19). There remains a notable lack of real-world studies to evaluate the risk factors, clinical progression, and particularly the treatment experience [the optimal timing of steroid administration, the assessment of steroid resistance, the timing and practical value of liver biopsy, and the appropriate time point for plasma exchange (PE)] (20). While several small-sample studies have investigated the risk factors for severe IMH, the results are not consistent regarding clinical application (17–19). To identify the risk factors associated with the development of severe IMH and to evaluate the treatment decisions for this condition, we conducted a retrospective analysis of cancer patients who developed severe IMH following ICI treatment.

To the best of our knowledge, the present study is the largest single-center cohort investigation to date that examined the clinical characteristics of patients who developed severe cholestatic IMH, including their clinical course, treatment methods, and outcomes.

The reported incidence of IMH varies significantly, ranging from 0.7% to 16%. This variation depends on factors such as the specific class of ICIs, dosage regimen, and whether it is monotherapy or combination therapy with ICIs (22). Approximately 1%–2% of patients receiving anti-PD-1/anti-PD-L1 therapy will develop grade 3 or higher IMH (23). Among these 32 patients, only one case developed severe IMH following immunotherapy provided by our center. The rest of the patients were transferred to our center after being diagnosed with IMH elsewhere. Therefore, the incidence in our cohort (1/348) was comparable to that reported in other studies. In the univariate analysis of risk factors for severe IMH, cancer type, specifically lung, esophageal, and gastric cancers, was found to be associated with the incidence of severe IMH. However, given that the majority of severe IMH patients in our center were referred from the thoracic surgery department, a significant selection bias was introduced regarding the types of cancer. Consequently, these variables were not included in the multivariate analysis of severe IMH risk factors.

The pathophysiology of IMH is not fully understood. The liver injury observed in IMH seems to be primarily related to the inherent mechanism of action of ICIs. Specifically, it is due to the increased autoimmunity to hepatocytes resulting from ICI-induced T-cell activation, rather than their intrinsic hepatotoxicity or immunogenicity, which is typically seen in other types of drug-induced liver injury (DILI) (24). In the present study, 56.3% of patients with severe IMH received oral TCM such as cinobufacin, Xihuang pill, and M. tenacissima as a part of their cancer treatment regimen. TCM is a major contributor to DILI among Asian populations; however, its underlying mechanisms remain incompletely understood (25). The pathogenesis of herb-induced liver injury may involve oxidative stress, inflammatory responses, mitochondrial dysfunction, cell apoptosis, lipid metabolism, and the release of fibrosis-promoting factors (26). We employed the RUCAM scoring system to diagnose IMH in order to exclude potential interference form TCM. However, due to the complexity of TCM components and the variability in administration methods, it remains challenging to fully eliminate the influence of DILI. Our findings suggest that TCM use may represent an independent risk factor for severe IMH. Although the inclusion of transferred patients from other departments or hospitals, as well as the potential hepatotoxic effects of TCM, may introduce some statistical bias in the analysis of the overall patient population, these findings still strongly remind us to exercise greater caution in the application of TCM during ICI treatment. Furthermore, a lower ECOG PS score ≤1 was identified as a potential risk factor for cholestatic IMH, possibly attributed to the inclination of patients with better functional status to receive higher doses of combination therapy, including chemotherapy and targeted therapy. In the present study, liver function showed signs of improvement at a median of 4 days. Half of the patients with an initially poor response to steroids displayed improvement in liver function after continued steroid use with increased doses at day 11. Beyond this time point, continued steroid therapy does not lead to improvement of liver function; therefore, we defined steroid resistance for IMH treatment as no improvement in liver function after 11 consecutive days of methylprednisolone treatment with the conventional dose of 1–2 mg·kg⁻¹·day⁻¹. The NCCN guidelines for managing IMH recommend adding mycophenolate mofetil (MMF) after 3–5 days of steroid therapy if the response is inadequate (18), but based on our clinical experience, it occupies only a very subordinate auxiliary position in managing liver function. In our study, seven patients who showed poor responses to steroid therapy were treated with MMF; however, none of them exhibited a significant decrease in bilirubin levels. Among these seven patients, four only experienced a notable decline in bilirubin levels following PE therapy. Based on this, the potentially most feasible solution may be PE. The necessity of liver biopsy in severe IMH patients is controversial (27), but we found an incurable bile duct deficiency patient among the steroid-refractory patients. Therefore, we recommend liver biopsy after confirmed steroid resistance but before PE to avoid unnecessary expenses for the bile duct deficiency patient.

PE has been utilized in the management of severe cholestatic liver injury (28, 29), including liver irAEs (30). It can clear the pathological mechanisms of irAEs, including pathological antibodies, pathological complement, irAE-associated cytokines (7), and the accumulated bilirubin and pseudo-neurotransmitters. However, to date, no established guidelines have explicitly recommended the specific application of PE and the time point for its implementation. In the present study, five of eight steroid-refractory patients (62.5%) responded to the PE treatment. The reported cases of successful PE in IMH patients mostly involved five sessions (31), but whether increased PE sessions can enhance the response rate remains uncertain. We observed a patient who initially received PE and subsequently experienced prolonged steroid treatment as well as subsequent infections; we also reported a case of an irAE-related pulmonary arterial hypertension patient whose pulmonary artery pressure elevated following PE (32). We believe that PE may also be a double-edged sword that clears both pathological factors related to side effects and factors that counteract these side effects in irAE progression. Early PE therapy may inadvertently remove protective antibodies or cytokines that help counteract irAEs. Based on this observation, we recommend performing a liver biopsy after confirming steroid resistance but prior to initiating PE therapy in order to avoid unnecessary interventions in patients with bile duct deficiency syndrome. In the future, we will try biological agents such as tocilizumab and infliximab, which have been proven to work in IMH treatment.

The infection rate among severe IMH patients receiving steroid treatment was 18.8%, including cytomegalovirus, novel coronavirus, and P.c. ( Supplementary Table S1 ). Among the first 20 severe IMH patients, three cases of P.c. pneumonia were observed, resulting in one death. Under the condition of TBIL reduction upon steroid treatment, a prophylactic treatment of trimethoprim–sulfamethoxazole was performed for later severe IMH patients whose total steroid dosage reached the cutoff value of 1,656 mg. It did not induce excessive hepatotoxicity when patients’ TBIL was reduced to a median level of 166.9 μmol/L.

Based on the previous reports and our data, we provided a management algorithm for severe IMH ( Figure 2 ). Methylprednisolone was administered at a dosage of 1 mg·kg⁻¹·day⁻¹ for 4 days, continuing with an increased dosage of 2 mg·kg⁻¹·day⁻¹ with or without immunosuppressive drugs (ISDs) for the non-responders to 11 days, and then a liver biopsy was performed for the remaining non-responders to exclude vanishing bile duct syndrome at that time point, followed by PE. During this process, when the steroid dosage reaches near the infection threshold and TBIL has dropped to 166.9 μmol/L or lower, cotrimoxazole should be used to prevent P.c. According to the proposed therapeutic algorithm for severe IMH, three out of 32 (9.4%) patients at our center experienced treatment failure resulting in death, which is lower than the reported mortality rate of 22% for severe IMH patients in other studies (8).

The present study had several limitations. First, it was a retrospective single-center study with a small sample size. Second, due to the limited sample size, we were unable to analyze potential correlations between factors associated with the occurrence of irAEs—such as family history of autoimmune diseases, prior episodes of other irAEs, use of proton pump inhibitors or metformin, and the development as well as treatment outcomes of IMH. These associations will be further explored in subsequent studies with larger sample sizes. Third, the overrepresentation of certain cancer types (e.g., lung, esophageal, and gastric cancers) in the cohort introduces a notable selection bias, although this variable was not included in the multivariate analysis. Given the limitation of sample size, this study inevitably carries the risk of confounding biases, particularly concerning the two clinical characteristics of TCM usage and ECOG PS scores. Fourth, due to the validity of the sample size, we were unable to conduct an internal validation of the ROC curve’s AUC of 0.923 to assess its stability. Fifth, there was an absence of histological analysis for all IMH patients to guide diagnosis and treatment. The necessity of liver biopsy in severe IMH cases remains controversial. According to the current NCCN guidelines for irAE management, the diagnosis of IMH is primarily one of exclusion. Liver biopsy should be conducted only when deemed absolutely necessary (on rare occasions) and not as a routine diagnostic measure (18). Furthermore, limited data exist regarding the histopathologic findings specific to IMH, and no pathognomonic features have been identified, thereby limiting the diagnostic utility of liver biopsy (33–35). Finally, given the invasive nature and associated risks of liver biopsy, it was selectively performed in this study only on patients with steroid resistance to guide further therapeutic strategies.

We investigated the potential risk factors for severe IMH and provided a feasible treatment roadmap for its management. We hope that our findings can provide some assistance to clinical doctors in dealing with severe IMH.