Risk factors and outcomes for steroid-refractory immune-related hepatotoxicity in locally advanced and metastatic cancer.

Introduction
Treatment with immune checkpoint inhibitors (ICIs), including cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) and programmed cell death protein-1 (PD-1)/programmed cell death ligand-1 (PD-L1) inhibitors, has achieved extraordinary response rates and has become a standard systematic treatment for patients with various malignancies in early or advanced stages [1]. However, T-cell dysregulation by ICIs can lead to the development of autoimmune manifestations, referred to as immune-related adverse events (irAEs). With the widespread application of ICIs, irAEs can occur in any organ, including the skin, endocrine system, liver, lung, and kidney, with an incidence of 54–76% [2, 3].
Immune-related hepatotoxicity (IRH) represents one of the common irAEs and varies in severity from mild to life-threatening. IRH typically presents with liver or bile duct injury along with elevated serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) with or without concurrent elevated total bilirubin (TBIL) [4, 5]. The clinical management of IRH is based on expert opinion according to local and international guidelines, including the ASCO, SITC, and NCCN guidelines [6–8]. Most grade ≥ 2 IRHs respond well to treatment with high-dose systemic corticosteroids with various time to resolution. However, some patients with IRH can present with a phenotype of steroid-refractory IRH (Ref-IRH) which does not improve after 5–7 days of corticosteroids or reoccurs under corticosteroid treatment [6–8]. Use of increased dose of corticosteroids provides no additional benefit in terms of the time to resolution [9]. These patients are recommended to discontinue ICI treatment and receive additional immunosuppressants including mycophenolate mofetil (MMF), tacrolimus (TAC), intravenous immunoglobulin (IVIG), or azathioprine, on the basis of on published guidelines [6–8]. Retrospective investigations revealed that 23–48% of IRH patients were treated with additional immunosuppressive agents [9, 10]. A recent systematic review and meta-analysis showed that a summary estimate of the proportion of patients with steroid-refractory hepatitis in a random effects model was 16% [11]. However, the incidence of Ref-IRH is unknown. Although successful treatments with various immunosuppressive agents for Ref-IRH have been widely reported in case series, the efficacy of immunosuppressive agent monotherapy or combination therapy for Ref-IRH is not clear [12–18]. Thus, this retrospective cohort study was designed to determine the risk factors and treatment outcomes for Ref-IRH with immunosuppressive agents in a variety of cancer types.

Materials and methods

Patient cohort
We retrospectively reviewed a cohort of patients with locally advanced and metastatic cancer who presented to the First Affiliated Hospital of Shandong First Medical University and were treated with ICIs from December 2019 to September 2024. The inclusion criteria were as follows: age older than 18 years; histologically or cytologically confirmed locally advanced or metastatic solid cancer according to the American Joint Committee on Cancer (AJCC) staging system, 8th version; receiving PD-1/PD-L1 inhibitor monotherapy (nivolumab, pembrolizumab, camrelizumab, sintilimab, tislelizumab, durvalumab, or atezolizumab) or a combination with CTLA-4 inhibitor, chemotherapy, or targeted therapy, as a first-, second- or later-line treatment for at least one tumor evaluation by imaging; and having an Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0–2. Patients whose response to therapy was not evaluated after the initiation of immunotherapy were excluded.

Evaluation of efficacy and safety
The objective response by radiographic findings was assessed as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD) according to RECIST 1.1. The objective response rate (ORR) was defined as the proportion of patients who achieved a CR or PR to immunotherapy. The disease control rate (DCR) was defined as the percentage of patients who achieved a CR, PR, or SD after immunotherapy. The diagnosis of IRH was conducted by the treating medical oncologist and confirmed by a local multidisciplinary team. Briefly, IRH diagnosis was based on a combination analysis of patients’ history, examination, symptoms, radiographic findings, blood parameters, and biological tests and a lack of evidence of tumor progression in the liver, virus infection, and pre-existing autoimmune hepatitis. We evaluated the clinical severity of all irAEs and IRH according to the Common Terminology Criteria for Adverse Events, version 5.0. AST/ALT grades 1 to 4 correspond to < 3 × upper limit of normal (ULN), 3–5 × ULN, 5–20 × ULN, and > 20 × ULN in order, respectively, whereas TBIL corresponds to < 1.5 × ULN, 1.5–3 × ULN, 3–10 × ULN, and > 10 × ULN, respectively. To identify liver injury patterns, the R-value is calculated by comparing ALT and alkaline phosphatase (ALP) levels to their upper normal limits. Based on the R-value, liver injury patterns can be classified as hepatocellular, cholestatic, or mixed. The pattern of injury was based on the R factor to differentiate a hepatocellular, cholestatic, or mixed pattern of injury. The treatment threshold for starting an escalation of IRH varied by clinical context but was in line with international and local guidelines. IRH that did not improve after 5–7 days of corticosteroid treatment or reoccurred under steroid treatment was considered Ref-IRH. Improvement with corticosteroids or immunosuppressive agents was defined as ALT and AST levels falling below 5 × ULN by day 7, and TBIL levels falling below 1.5 × ULN by day 5. In contrast, worsening was defined as the failure to achieve these biochemical thresholds.

Clinical annotation
Data were gathered through electronic patient records and oncology databases. The baseline clinical characteristics of these patients included age, sex, ECOG PS, smoking status, histology, tumor node metastasis (TNM) stages, complete blood counts, proteins, and cytokines at the start of treatment. Liver function, protein, and cytokine levels were also analyzed at the time of IRH diagnosis. The neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), eosinophil-to-lymphocyte ratio (ELR), monocyte-to-lymphocyte ratio (MLR), and systemic immunity-inflammation index (SII) were calculated on the basis of the results of the baseline complete blood test. The NLR, PLR, ELR, and MLR were calculated by dividing the absolute neutrophil, platelet, eosinophil, and monocyte counts, respectively, by the lymphocyte counts. The calculation of the SII is based on the following formula: SII = platelet count × neutrophil count/lymphocyte count. The levels of cytokines including interleukin (IL)-2, IL-4, IL-6, IL-10, interferon-γ (IFN-γ), and tumor necrosis factor-α (TNF-α) were determined via the enzyme-linked immunosorbent assay (ELISA) using Human Cytokine Standard Assays Panel (ET Healthcare, Inc., Shanghai, China) according to the manufacturer’s instructions. Progression-free survival (PFS) was defined as the date of immunotherapy initiation to the date of disease progression or death from any cause, whichever occurred first. Patients who were alive without disease progression were censored on the date of their last disease assessment. Overall survival (OS) was defined as the time from immunotherapy initiation to death from any cause. Patients who were still alive were censored at the date of last contact. This study was approved by the independent research ethics committee of the First Affiliated Hospital of Shandong First Medical University (NO: YXLL-KY-2022–059) and conformed to the principles of the Declaration of Helsinki.

Statistical analysis
Categorical variables are reported as frequencies and percentages. Quantitative patient variables are presented as medians and ranges. For categorical variables, we used Fisher’s exact test to compare patient characteristics among different groups. For continuous variables, independent sample t tests or Mann‒Whitney U tests were used to compare patient groups. Univariable logistic regression was used to analyze the relationships between patient variables and the development of Ref-IRH. Survival outcomes were estimated via Kaplan–Meier curves and the log-rank test. The hazard ratio (HR) was estimated via Cox proportional hazards regression models. To confirm that the differences in Kaplan‒Meier survival curves between patients with and without IRH were not attributable to time bias, we adopted the 3-month landmark analysis method according to our previous report [3]. In this analysis, we excluded patients who experienced events within 3 months of the initiation of ICI therapy, as well as those who developed IRH after 3 months of ICI therapy initiation. Statistical analyses were performed via GraphPad Prism version 9 (GraphPad Software, San Diego, CA, USA), and SPSS 26.0 software (SPSS Inc., Chicago, IL, USA) was used for all the statistical tests. All the statistical tests were two-tailed, and p < 0.05 was considered statistically significant.

Results

The incidence of Res-IRH and Ref-IRH
In this pan-tumor cohort, which included 530 locally advanced and metastatic cancer patients who received immunotherapy, 50 patients were excluded because they were lost to follow-up. Finally, 480 patients were included in this study. A total of 160 patients (33.3%) were developed non-IRH irAEs. A total of 35 patients (7.3%) had IRH, including 10 patients with G1 IRH who did not require steroid treatment and 25 with G2-4 IRH who required steroid treatment (Fig. 1). Among patients with G2-4 IRH, 12 developed Res-IRH, and 13 developed Ref-IRH. The median age was 64.0 years for patients with IRH and 61.0 years for patients without IRH. The median age was 65.5 years for patients with Res-IRH and 64.0 years for patients with Ref-IRH (Table 1).

Comparison of clinical features between patients with IRH and non-IRH
The patients’ clinical characteristics are summarized in (Table 1). We first compared the clinical features of patients with and without IRH. In both groups, most patients were male, had metastatic disease, and received anti-PD-1-containing treatments. Patients with IRH were more likely to have liver metastasis (p = 0.019), and receive anti-PD-1 plus anti-CTLA-4 (p = 0.008), ICI and targeted therapy (p = 0.012), or first-line immunotherapy (p = 0.012). We also analyzed whether peripheral blood biomarkers were associated with the development of IRH. Patients with IRH had lower ALC (p = 0.001), lower MLR (p < 0.001), and lower platelet counts (p = 0.007), and higher procalcitonin (p < 0.001), LDH (p < 0.001), IL-2 (p = 0.035), IL-6 (p = 0.003), and IL-10 levels (p < 0.001) at diagnosis of irAEs compared with those with non-IRH. There was no significant difference in cancer types, absolute blood cell count, protein levels, lymphocyte count, other cytokines (IFN-γ, IL-12, and IL-17), or blood indicators between the two groups (Fig. 2A–D).

Comparison of clinical characteristics between patients with Ref-IRH and Res-IRH
Among 13 patients with Ref-IRH, 7 underwent liver biopsy. All patients with Ref-IRH had mild to moderate bile duct injury, and 28.6% (2/7) had ductopenia. All patients had varying degrees of lobular inflammation, ranging from mild to moderate degrees, with a mixture of infiltrated inflammatory cells. Representative liver histological findings of Ref-IRH were observed in an HCC patient (Fig. S1A), a gastric patient (Fig. S1B), and non-small cell lung cancer patient (Fig. S1C). We next compared the clinical characteristics of patients with Ref-IRH and Res-IRH. Patients with Ref-IRH were more likely to receive ICI and targeted therapy combinations (61.5 vs. 25.0%, p = 0.046). Hepatocellular carcinoma (HCC) patients had a greater risk of Ref-IRH than patients with non-HCC malignancies did (100% vs. 40.0%, p = 0.035). There was no difference between the Ref-IRH and Res-IRH groups regarding smoking status, ICI type, previous liver disease, previous irAEs, or concurrent irAEs. The onset times of Ref-IRH and Res-IRH were similar (101.0 days vs. 68.5 days, p = 0.508) (Table 1). Patients with Ref-IRH had a longer duration of steroid use (61.5 days vs. 37.0 days, p = 0.037) compared with patients with Res-IRH, and the mean dose of steroids was comparable (2 mg/kg vs. 1.0 mg/kg, p = 0.206). There was no difference between the Ref-IRH and Res-IRH groups in terms of hepatic injury (p = 0.612) (Fig. 3A). In addition, Ref-IRH patients were more likely to have higher CTCAE grades (p = 0.044) than Res-IRH were (Fig. 3B). Patients with Ref-IRH had higher peak levels of TBIL (p = 0.003), DBIL (p = 0.001), and IBIL (p = 0.051) but not ALT (p = 0.225), AST (p = 0.469), or ALP (p = 0.348) (Fig. 3C). We also analyzed whether peripheral blood biomarkers were associated with the development of Ref-IRH. The results revealed that patients with Ref-IRH had lower platelet counts (p = 0.006), higher procalcitonin levels (p = 0.012), and higher IL-6 levels (p = 0.038) at diagnosis of IRH than did those with Res-IRH. There was no significant difference in absolute blood cells, proteins, lymphocytes, cytokines, or blood indicators between the two groups (Fig. 3D–G). Univariate logistic regression analysis indicated that high IL-6 levels (p = 0.019) and grade 4 disease (p = 0.031) can predict the development of Ref-IRH (Table 2).

Outcomes of patients with Ref-IRH
We next compared the clinical outcomes between patients with and without IRH. The median PFS times for patients with IRH and non-IRH were 14.2 and 5.8 months, respectively (HR: 0.57, 95% CI: 0.41–0.78, p < 0.001). The median OS times for patients with IRH and non-IRH were 27.2 and 18.5 months, respectively (HR: 0.69, 95% CI: 0.45–1.06, p = 0.087). The ORR with patients for Ref-IRH was comparable to that for patients with Res-IRH (38.2% vs. 20.7%, p = 0.029) (Fig. 4A–C). In addition, 10 patients with grade 1 IRH did not receive corticosteroid treatment, including 4 with the hepatocellular type, 2 with the cholestatic type, and 4 with the mixed type. The ORR was 30%. These patients received continued ICI treatment, with a median PFS of 13.8 months and median OS of 19.5 months.
We next compared the clinical outcomes and efficacy between patients with Ref-IRH and those with Res-IRH. There was no significant difference in PFS or OS between patients with Ref-IRH and those with Res-IRH. The median PFS for patients with Ref-IRH and Res-IRH was 18.3 and 21.0 months, respectively (HR: 0.82, 95% CI: 0.30–2.19; p = 0.685). The median OS times for patients with Ref-IRH and Res-IRH were 28.3 and 23.3 months, respectively (HR: 1.01, 95% CI: 0.32–3.14; p = 0.989). The ORR for patients with Ref-IRH was comparable to that for patients with Res-IRH (50.0 vs. 30.8%, p = 0.428) (Fig. 4D–F). To alleviate the impact of immortal time bias, we used a 3-month landmark analysis in which patients who died prior to 3 months of ICI initiation or developed IRH after 3 months of ICI initiation were excluded. There was no significant in regarding PFS or OS between patients with IRH and those without IRH (Fig. S2A and B), or between patients with Ref-IRH and those with Res-IRH (Fig. S2C and D).
Furthermore, we analyzed the efficacy of corticosteroids and immunosuppressive agents in patients with Res-IRH and Ref-IRH. All patients with Res-IRH improved with corticosteroid treatment, but 76.9% of patients with Ref-IRH improved with immunosuppressive agents. Salvage therapies using immunosuppressive agents for Ref-IRH were administered based on individual clinical conditions, including MMF monotherapy (n = 5) and MMF in combination with intravenous immunoglobulin (IVIG) and/or tacrolimus (TAC) (n = 8). Two patients with MMF combination treatment did not improve and received dual-plasma molecular absorption system (DPMAS) and/or anti-IL-6 receptor tocilizumab treatment, leading to a rapid resolution of transaminase elevation within 10 days (Fig. 4G and H; Table S1). Patients with grade 2 Ref-IRH had the similar improvement rate as did those with grade 3–4 Ref-IRH (p = 0.569) (Fig. 4I). Patients with grade 3 Ref-IRH had a similar improvement rate as did those with grade 4 Ref-IRH (p = 0.735) (Fig. 4J).
However, we also compared the recovery process between the two groups. The results revealed that patients with Ref-IRH had a longer time from the initial diagnosis of IRH to resolution to grade 1 (p = 0.002), from peak ALT to resolution to grade 1 (p = 0.007), from peak AST to resolution to grade 1 (p = 0.011), from peak TBIL to resolution to grade 1 (p = 0.048), and from the initial diagnosis of IRH to use of prednisone ≤ 20 mg/day (p = 0.025). The length of hospital stay was longer in the Ref-IRH group than in the Res-IRH group (14 days vs. 10 days, p = 0.017) (Table 3).

Discussion
This study reported the clinical features, efficacy, and outcomes of IRH. Patients with Ref-IRH were more likely to be hepatocellular carcinoma, have higher grades of liver injury, lower platelet counts, higher procalcitonin levels, and higher IL-6 levels at diagnosis than were those with Res-IRH. High IL-6 levels act as a potential predictive factor for the development of Ref-IRH. Although the treatment efficacy and survival outcomes of patients with Ref-IRH were comparable to those with Res-IRH, patients with Ref-IRH were unlikely to quickly recover with a longer time from initial diagnosis of IRH to resolution to grade 1, from peak ALT and AST to resolution to grade 1, from diagnosis of IRH to use of prednisone ≤ 20 mg/day and had prolonged hospital length of stay.
The frequency of Ref-IRH is unknown. Previous studies reported that the incidence of IRH was 2–25% with ICI monotherapy and 20% with dual therapy, and 0.8–2% of IRH patients in somatic tumor required immunosuppressive agents [9, 19, 20]. Here, we reported that the incidence of Ref-IRH was 2.7%, which was higher than that reported in previous studies. This may be associated with more attention given to be pathological diagnosis and treatment of irAEs in our study where 7/13 Ref-IRH patients underwent liver biopsy.
Specific biomarkers for ICI-associated irAEs and liver injury have not yet been identified. Previous studies have indicated that liver injury is more common in HCC patients; and that HCC patients tend to have a greater risk of immune-related hepatitis than patients with non-HCC malignancies after ICI therapy, but this finding was not statistically significant [21, 22]. In our study, no difference was observed in IRH (17.1 vs. 82.9%) between HCC patients and non-HCC patients, but HCC patients had a higher risk of Ref-IRH than patients with non-HCC malignancies did (100 vs. 40.0%). It is not clear whether the “refractory” nature of liver injury is truly immune-mediated or related to hepatic decompensation in a compromised liver. Further investigations on the frequency and potential mechanisms associated with Ref-IRH are needed in HCC cohorts with large sample sizes. Furthermore, we found that patients with Ref-IRH had significantly higher PCT levels. Although high PCT is a strong marker for bacterial infection, we excluded cholangitis or sepsis in all patients including those with liver metastases or HCC. In our cohort, no patients with higher PCT levels were identified as having infection. High baseline levels of inflammatory markers and PCT are strongly predictive of poor outcomes in patients with metastatic NSCLC receiving PD-1/PD-L1 immune checkpoint blockade [23]. High procalcitonin levels are even found in NSCLC patients with steroid-refractory checkpoint inhibitor pneumonitis (ALT) and aspartate aminotransferase (AST) falling below 5× the upper limit of normal (ULN) by day 7 [24]. Immunotherapy may be associated with noninfectious PCT elevation.
There is growing evidence that immunotoxicity profiles can be tied to specific cytokines that can amplify both pro- and anti-inflammatory responses [25]. Baseline peripheral blood biomarkers [26], cytokines [27], C-reactive protein [28], and early increases in CXCL9/CXCL10/CXCL11 and IFN-γ levels 1 to 2 weeks after the start of therapy are likely predictors of an increased risk of developing irAEs [29]. Increased expression of IL1B, CXCL8, and CXCL2 in monocytes and TNF in macrophages was closely associated with severe irAEs according to single-cell RNA sequencing of peripheral blood mononuclear cells before treatment [30]. A single toxicity score CYTOX, which integrated the expression of 11 cytokines, can be used for more objective prediction of irAEs [31].
IL-6 is an important cytokine for the differentiation of naïve CD4+ T cells into TH17 cells and plays a pivotal role in the pathogenesis of several autoimmune diseases such as rheumatoid arthritis, systemic sclerosis-associated interstitial lung diseases, and cytokine release syndrome, where the use of anti-IL-6 receptor antibodies is currently approved [32]. An early increase in Th17 cytokine signatures and congruent on-treatment expansions of Th17 and Th2 effector memory T-cell populations in peripheral blood were positively associated with the development of grade ≥ 2 irAEs [33]. Specifically, elevated circulating or local IL-6 was shown to be associated with the occurrence of de novo immune-related arthritis, pneumonitis, psoriasis, and other irAEs involving different organ systems [34–40]. In a murine model of live injury secondary to combined anti-CTLA-4 and anti-PDL-1 therapy, hepatocyte apoptosis, caused by the activation of the NLRP3 inflammasome and the production of proinflammatory cytokines such as IL-6, IL-1β, IL-18, and TNF-α were observed [41]. In atezolizumab- and bevacizumab-treated unresectable HCC patients, the presence of fever was a significant risk factor for liver injury, and patients with preceding fever had significantly higher IL-6 levels at 1 and 3 weeks of treatment [42]. Here, we focused on Ref-IRH and reported that high IL-6 levels at diagnosis were significantly associated with the development of irAEs, although this conclusion warrants prospective validation. This is the first study indicating a relationship between Ref-IRH development and elevated serum levels of IL-6. Future studies will focus on whether Ref-IRH is associated with the dynamics of IL-6 levels after treatment with ICIs and the occurrence of Ref-IRH.
Several retrospective studies have indicated that patients who develop IRH have improved survival compared with those that do not develop IRH [10, 43–46]; however, some studies have also suggested negative or opposite results in HCC, melanoma, NSCLC, and other solid cancers [20, 47–51]. Our reports revealed a positive association between prolonged PFS and IRH. However, fatal toxic effects associated with anti-PD-1/PD-L1 therapy often include pneumonitis (35%), hepatitis (22%), and neurotoxic effects (15%) according to the Vigilyze database [52]. We also reported that delayed hepatic and respiratory irAEs tended to occur more commonly in the fatal group in the FAERS database [53]. Here, we did not observe significant difference in ORR, PFS, or OS between patients with Ref-IRH and those with Res-IRH, but patients with Ref-IRH had a longer time from diagnosis of IRH to resolution to low grade and longer exposure to high-dose steroids although all Ref-IRH patients were treated with immune suppressive agents. Thus, high-grade Ref-IRH is not always fatal, but a long time to resolution is needed. Furthermore, additional immunosuppressants alone or in combination are needed but not are always effective for the management of Ref-IRH [54–57]. Recently, neutrophil accumulation and systemic IL-6 release were identified as mechanisms associated with intestinal toxicity in response to CTLA-4 blockade, and IL-6 blockade combined with antibiotic treatment decreased intestinal damage and improved anti-CTLA-4 therapeutic efficacy in inflammation-prone mice [58]. The anti-IL-6 receptor tocilizumab was effective in treating steroid-refractory high-grade immune-related cytokine release syndrome and other steroid-refractory irAEs without hindering antitumor immunity [59–61]. Furthermore, anti-IL-6 receptors have been successfully used in Ref-IRH and are even effective for both the treatment of irAEs and the prevention of flares of pre-existing autoimmune disorders [62–65]. Rechallenge with ICIs can be considered in selective Ref-IRH or checkpoint inhibitor pneumonitis patients whose prophylactic use of the anti-IL-6 receptor is under close monitoring [66, 67]. The antitumor TNF-α antibody infliximab is widely recommended as salvage therapy for immune-related colitis but never for IRH, does not result in increased hepatotoxicity and leads to a long-lasting positive response in patients with Ref-IRH [13, 68].
Our study has several limitations. This is a retrospective design with a small number of IRH cases. Given such a low number of events with Ref-IRH, the multivariate logistic regression model is prone to overfitting; here, we presented only the univariable association between IL-6 levels and Ref-IRH. Further studies are needed to identify whether IL-6 is an independent risk factor for the development of Ref-IRH in cohorts with large sample sizes via multivariate logistic regression analysis. In addition, this study focused on IRH regardless of cancer type, disease stage, and treatment line. Included patients were highly heterogeneous. Survival analyses should be performed in a subgroup of patients with specific cancer types, such as HCC, lung cancer, and melanoma who experienced IRH or Ref-IRH. Live biopsy was not performed in all IRH patients, which means that other reasons associated with liver injury cannot be completely ruled out. In our study, even though immunosuppressive agents were administrated to all patients with Ref-IRH, only one patient was treated with an anti-IL-6 receptor. No histological data for IL-6 is presented to confirm local upregulation of this cytokine in the liver tissue. Investigations on the possible IL-6-mediated molecular mechanisms responsible for Ref-IRH are also needed.
In summary, the present study indicated that high IL-6 at diagnosis was a potential risk factor for the development of Ref-IRH. There was no significant difference in efficacy with immunosuppressive agents and survival outcomes between patients with Ref-IRH and Res-IRH, but much more time from the time of initial diagnosis of IRH to resolution to grade 1 and the use of immunosuppressive agents are needed for Ref-IRH.

Supplementary Information
Below is the link to the electronic supplementary material.