Clinical Features and Outcomes Based on Liver Injury Patterns in Liver Injury Caused by Immune Checkpoint Inhibitors

Background The clinical course of liver injury induced by immune checkpoint inhibitors (ICIs) varies among individuals, and there were few reports on the therapeutic effects of corticosteroids based on the patterns of liver injury. Methods We evaluated the characteristics and clinical course of immune-related liver injury in 1087 patients treated with ICIs for advanced malignancies between August 2014 and December 2020. Results The the marked by an increase in biliary (i.e., ALP and γ-glutamyl [GGT]) was poor. ICIs are reported to rarely immune-related are only increased. recent case utility of is different


Introduction
The number of patients exposed to immune checkpoint inhibitors (ICIs), including anti-programmed cell death 1 (PD-1), programmed cell death ligand 1 (PD-L1), and anti-cytotoxic T-lymphocyte antigen 4 (CTLA-4) antibodies, for advanced malignancies has increased over the past few years [1][2][3] . Although ICIs are generally well tolerated, they have the ability to disrupt the immune system, which can generate unique side effects that mimic autoimmune conditions, termed "immune-related adverse events" (irAEs) 4 . IrAEs can affect various organs, but are most often found in the skin, gastrointestinal tract, and endocrine system 5 . Immune-related liver injury caused by ICIs is a relatively rare irAE; however, it can lead to patient mortality 6 . The appearance of severe immune-related liver injury should be followed by prompt withdrawal of ICIs. Systemic administration of corticosteroids is used as the rst-line therapy for immunerelated liver injury, while second-line immunosuppressants are necessary in steroid-refractory cases 7 .
Recently, we reported that severe immune-related liver injury (≥ Grade 3) occurred in 29 patients (5.3%) in an analysis of 546 patients treated with ICIs, and the median period between the initial administration of ICIs and the incidence of irAEs was 52 days 8 . In addition, more than half of the patients with immunerelated liver injury showed a cholestatic and mixed pattern, not hepatocellular. Some patients were resistant to steroid therapy, but the detailed clinical courses of immune-related liver injury are not well known.
The aim of this retrospective study was to examine the clinical features and response to immunosuppressive treatment based on liver injury patterns in patients with liver injury caused by ICIs.

Study population
We retrospectively collected clinical data from 1087 patients with advanced malignancies who were treated between August 2014 and December 2020 with PD-1, PD-L1, or anti-CTLA-4 agents as monotherapy or in combination with an anti-CTLA-4 agent at Nagoya University Hospital and Ogaki Municipal Hospital (Nagoya University Hospital, n = 660; Ogaki Municipal Hospital, n = 427). We reviewed the detailed clinical course of patients with immune-related liver injury, which was de ned as liver injury induced by ICIs (liver-irAE) in this article. The study protocol was carried out in accordance with the Declaration of Helsinki and was approved by the Institutional Review Boards of both Nagoya University Hospital and Ogaki Municipal Hospital (no. 2018 − 0438, 15006).

Diagnosis of immune-related liver injury
We evaluated the patients' general condition carefully with blood tests at intervals of least 3 weeks after the start of ICI administration to assess the incidence of side effects, including immune-related liver injury. The severity of irAEs was assessed using the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. We checked no evidence of hepatitis A, B, or C virus infection, and no other potential causes of liver disease (i.e., massive alcohol consumption [> 80 g/day], other hepatotoxic drug use, autoimmune hepatitis, primary biliary cirrhosis, or hemochromatosis). In addition, we performed ultrasonography, contrast-enhanced computed tomography, or magnetic resonance imaging to exclude liver injury induced by liver metastasis and bile duct obstruction. The patterns of liver injuries were de ned as follows on the basis of previous reports 9,10 : (i) hepatocellular pattern, alanine aminotransferase (ALT) level alone is elevated ≥ 5-fold above the upper limit of normal (ULN) or the ratio of serum activities (expressed as a multiple of ULN) of ALT and alkaline phosphatase (ALP) is ≥ 5; (ii) cholestatic pattern, ALP level alone is elevated ≥ 2-fold above the ULN or the ratio of serum activities of ALT and ALP is ≤ 2; (iii) mixed pattern, the ratio of the serum activities of ALT and ALP was > 2 and < 5.
Treatment for immune-related liver injury We generally followed the Society' s guidelines for the management of immune-related liver injury 5 7 . These guidelines suggest that the immediate start of corticosteroids should be considered for severe liver-irAEs (Grades 3 and 4). However, some cases showed spontaneous improvement without the introduction of treatment. A recent review also suggested corticosteroid treatment if there was no improvement after a few days of follow-up after the withdrawal of ICIs 11 . Therefore, for patients with Grade 3 liver injury, we performed careful follow-up for one week after the appearance of the injury and started oral prednisone 0.5-1.0 mg/kg/day if there was no improvement. For Grade 4 liver injury, steroid pulse with methylprednisolone was started immediately, followed by treatment with prednisone 1.0-2.0 mg/kg/day. The use of ursodeoxycholic acid (UDCA) was considered if obstructive jaundice could be ruled out in cases of cholestatic or mixed-type disease. We de ned "improvement" as recovery to within normal levels or baseline levels before the start of ICI treatment. A part of patients who could not achieve "improvement" included best supportive care (BSC) for original disease progression after the occurrence of immune-related liver injury.

Assessment of pathological ndings
Percutaneous liver biopsy was performed with a 16-to 17-gauge needle under ultrasonographic guidance, and liver biopsy specimens were immediately xed in 10% formalin and embedded in para n. An adequate liver biopsy sample was de ned as a specimen with a length of > 1.5 cm. Pathological slides were reviewed by a single pathologist without knowledge of the clinical data.

Statistical analysis
Values are expressed as median ( rst-third interquartile) and number (%). We used the chi-square test to compare categorical variables, and the Wilcoxon rank-sum test or Kruskal-Wallis test to compare continuous variables between two or three groups (hepatocellular, mixed, and cholestatic liver injury groups). For all tests, statistical signi cance was de ned as p < 0.05. We used the lower or upper limits of the reference values at Nagoya University Hospital and Ogaki Municipal Hospital as the cut-off values for laboratory data. Statistical analyses were performed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria) 12 .

Results
Clinical characteristics of patients with severe immune-related liver injury based on the type of liver injury Table 1 shows the clinical characteristics of patients with immune-related liver injury (≥ Grade 3), which was noted in 56 patients during the follow-up period (median, 270 days). The median age was 64 years, and the study population had a predominance of men (55.4%). The patterns of liver injuries were hepatocellular (n = 25, 44.6%), mixed (n = 10, 17.9%), and cholestatic (n = 21, 37.5%). Age, sex, tumor type, ICI type, previous history of chemotherapy, and incidence of irAEs other than liver injury did not vary signi cantly among the three groups. The median number of doses from the rst ICI administration to liver injury was 2, 4, and 3 (P = 0.045), and the median time to onset of liver injury was 36, 85, and 53 days (P = 0.036) in the hepatocellular, mixed, and cholestatic types, respectively. Thus, the hepatocellular pattern occurred earlier than the other types. Notably, baseline liver enzyme levels did not affect the incidence of each type of liver injury, and almost all patients showed normal immunoglobulin G (IgG) levels and negative results for anti-nuclear antibodies (86.0%) and anti-mitochondrial M2 antibodies (95.3%).  The number of dose from rst ICI administration to liver injury ** (times) ALT (baseline) * (U/L) 15 [11,22] 15 [13,23] 16 [11,21] 15 [9,22] 0.837 GGT (baseline) * (U/L) 34 [22,49] 28 [19,45] 28 [20,47]

Sub-analysis of the incidence of immune-related liver injury
A recent report showed that irAEs involving multiple organ systems, called multisystem irAEs, differ from single irAEs in terms of the pattern of incidence and prognosis 13 . Therefore, we evaluated the characteristics of immune-related liver injury coexisting with other irAEs. During follow-up, 20 patients (35.7%) developed multisystem irAEs, and 36 (64.3%) developed only immune-related liver injury (Table 1). The most common irAEs other than liver irAEs in patients with multisystem irAEs were endocrine irAEs (n = 8: thyroid, n = 4; type 1 diabetes, n = 2; pituitary, n = 1; and adrenal gland; n = 1), gastrointestinal (GI) irAEs (n = 7: intestine, n = 5; pancreas, n = 2) ( Table 2). The median number of days until the incidence of immune-related liver injury in patients with multisystem irAEs was signi cantly longer than that in those with only liver irAEs. However, the presence of multisystem irAEs did not affect the degree of liver injury (Supplementary Table 1). Next, we investigated the differences in characteristics in relation to the use of the anti-CTLA-4 antibody because administration of this antibody is known to be a risk factor for immune-related liver injury 8,14 (Supplementary Table 2). This analysis showed a signi cant difference in tumor types between the two groups because anti-CTLA-4 antibody treatment for malignant melanoma and renal cell carcinoma is covered by insurance in Japan. The incidence rate of multisystem irAEs in the group treated with the anti-CTLA-4 antibody was signi cantly higher than that in the group that was not treated with this antibody (P = 0.022). Anti-CTLA-4 antibody use was also not associated with the degree of liver injury as well as the presence of multisystem irAEs.

Clinical course in patients with immune-related liver injury
We treated 29 patients (51.8%) with immune-related liver injury with corticosteroids (Supplemental Figure); we mainly treated the cases with hepatocellular-and mixed-type injuries, and only 30% of the cases with cholestatic-type injuries were treated with steroids (Table 1). Figure 1 shows the clinical course of patients with immune-related liver injury based on injury patterns. Importantly, almost all patients with hepatocellular-type injuries who received corticosteroids improved (n = 13/14, 92.9%); however, less than half of the patients with the non-hepatocellular patterns (mixed, n = 4/8 [50.0%]; cholestatic, n = 3/7 [28.6%] ) who received corticosteroid treatment improved regardless of UDCA use. Three patients with mixed patterns needed secondary immunosuppression with mycophenolate mofetil (MMF) because of steroid resistance. They all once improved with initial corticosteroid, but underwent MMF therapy since they deteriorated during steroid tapering. Among them, only one patient subsequently showed improvement using MMF. Additionally, 18 patients (32.1%; hepatocellular, n = 9; mixed, n = 2; and cholestatic, n = 7) improved without treatment, including corticosteroids, and 3 patients with cholestatic type improved with UDCA alone ( Fig. 1 and Supplementary Figure).

Liver pathological ndings in patients with immune-related liver injury
In the present study, we performed a liver biopsy in 13 patients to con rm the diagnosis of immunerelated liver injury induced by ICIs (hepatocellular, n = 11; mixed, n = 1; and cholestatic, n = 1). Table 3 shows a summary of the histologic features, and the detailed information is listed in Supplemental Table 3. The common histologic ndings were lobular injury (n = 11, 84.6%), endothelialitis (n = 10, 76.9%), and spotty necrosis (n = 13, 100%) with in ltration of T cells positive for cluster of differentiation (CD) 3 and CD8, but not CD4 or CD20. Some cases showed the ndings of granulomas, as well as the characteristics reported in previous studies 15,16 (Fig. 2). On the other hand, in ltration of plasma cells, which is observed in autoimmune hepatitis (AIH), was rare (n = 2, 15.4%). Vascular endothelial damage and spotty necrosis with CD3-and CD8-positive T-cell in ltration were also observed in two cases with non-hepatocellular types, indicating that they may be common ndings in liver-irAE independent of liver injury types (Supplementary Table 4).

Discussion
ICIs are an essential breakthrough for the treatment of advanced malignancies, and the incidence of irAEs will continue to rise with the increasing use of ICIs for various types of cancers 4,17 . Immune-related liver injury is a type of irAE, along with an excessive immune response to one's own normal organs by ICIs.
However, the detailed mechanism underlying immune-related liver injury is unclear, and the development process and severity show substantial heterogeneity. Few reports have focused on the outcomes of immune-related liver injury caused by immunosuppression therapy, including corticosteroids.
In this study, we estimated the characteristics and clinical course of immune-related liver injuries based on the patterns of liver injury. Our study showed that the hepatocellular pattern was the most common (44.6%), and it occurred earlier than the other types (mixed or cholestatic). The hepatocellular pattern showed good response to corticosteroids; however, some cases with non-hepatocellular patterns showed corticosteroid resistance. In particular, 3 cases with mixed patterns required additional immunosuppression with mycophenolate mofetil. These results indicate that the response to corticosteroids in patients with liver injuries marked by an increase in biliary enzyme (i.e., ALP and γglutamyl transpeptidase [GGT]) levels was poor. ICIs are reported to rarely induce immune-related cholangitis 8,18,19 , which is characterized by predominant elevation of biliary enzymes, while liver transaminase levels are only moderately increased. A recent case report showed the utility of UDCA administration in patients with immune-mediated cholangitis induced by anti-PD-1 antibody that showed resistance to high-dose corticosteroids 19 . The pathogenesis of immune-related liver injury is different from that of classical autoimmune liver diseases such as AIH or primary biliary cholangitis (PBC) 20 .
However, the results of our study and previous reports indicate the need to plan the treatment strategy for each type of liver injury, such as the use of steroids for AIH with elevated aspartate aminotransferase (AST) and ALT levels, and UDCA for PBC with elevated ALP and GGT levels.
In the society's guidelines, ICIs should be immediately withdrawn on the appearance of severe immunerelated liver injury (≥ Grade 3), and steroid administration should be initiated 5 7 . However, De Martin et al.
reported that 6 patients (38%) with 16 biopsy-proven immune-related liver injuries did not receive any corticosteroid therapy and experienced spontaneous improvement 16 . Similarly, our study also demonstrated that 21 patients (38%) who did not receive any treatment (n = 18) or received only UDCA (n = 3) improved without requiring corticosteroids. A previous report indicated that corticosteroid use is associated with an increased risk of serious infections in patients receiving ICIs 21 . Thus, recovery from immune-related liver injury without corticosteroid administration can contribute to an improved prognosis. Therefore, we followed up Grade 3 liver injuries with stable general conditions without corticosteroid therapy for approximately one week. On the other hand, since ICIs can rarely lead to fulminant hepatitis 22 , an immediate start of corticosteroid therapy should be considered for Grade 4 or acute liver failure.
In our study, one-third of all cases of immune-related liver injury were accompanied by other irAEs. Few studies have assessed the differences in clinical characteristics of immune-related liver injury between multisystem and single irAEs. We found that immune-related liver injuries in multisystem irAEs occurred later than those in cases with liver irAEs alone, and the use of the anti-CTLA-4 antibody was associated with the incidence of multisystem irAEs, consistent with previous reports 23,24 . Our study demonstrated that the common types of irAEs accompanying liver injury were endocrine and GI injuries. Regarding the mechanism of these irAEs, the expression of CTLA-4 in normal pituitary cells for hypophysitis, preexisting autoantibodies for thyroiditis, and cytokines (e.g., interleukin-17) and speci c gut microbiota for colitis have been reported to be key factors for onset 4,25−27 . However, the detailed pathogenic mechanisms underlying hepatic irAEs remains unclear. In multisystem irAEs, the mechanism of liver irAEs may differ from that of the other irAEs, even though they occur in a consecutive clinical course.
Liver biopsies can assist not only in differential diagnosis but also in assessing the severity of histological liver damage in immune-related liver injury 11,16 . Recent research emphasized that the corticosteroid dose should be determined based on the pathological in ammatory state to avoid unnecessary systemic corticosteroid treatment 11,16 . In addition, it is useful to exclude the possibility of classical AIH or drug-induced liver injury caused by other drugs 20 . The pathological ndings of typical ICI-induced liver injury have been described as lobular hepatitis with numerous histiocytes, sometimes forming loose, well-formed, or brin ring granulomas, endothelialitis, and varying portal in ammation 15,20 . Consistent with these previous reports, the main histologic features in the 13 patients with biopsyproven immune-related liver injury in our study were lobular injury, endothelialitis, and spotty necrosis with in ltration of CD3-and CD8 cells. It is unclear why these ndings appear in immune-related liver injury, but we hypothesize that this may be related to the mechanism underlying the injury. PD-L1, the ligand for PD- Although more than half of all immune-related liver injuries were of the non-hepatocellular type according to laboratory data, 11 of 13 patients who underwent liver biopsies (85%) had hepatocellular-type injuries. This is presumably because liver biopsies were performed in patients scheduled to undergo steroid use.
The pathological cholangitis was observed in both mixed or cholestatic pattern, however, these data were limited as there were only two non-hepatocellular patterns in this study.
We also recognize several limitations in our study. First, since this study was conducted retrospectively, treatment for immune-related liver injuries was not completely consistent. Treatment is performed according to the guidelines, but it depends on the judgment of the attending physician. Second, we could not assess the characteristics and outcomes of patients with grade 1 and 2 injuries. In fact, it is di cult to con rm the diagnosis of immune-related liver injury in these patients, especially Grade 1 injury, because patients with advanced malignancies often have a history of multiple medications. Additionally, disease progression, such as liver metastasis, can induce liver injury. It is not practical to perform liver biopsy in all patients with Grade 1-2 injuries because of the risk of bleeding. However, we strictly excluded the other liver diseases in the included patients with ≥ grade 3 injuries in our study. We targeted only patients with immune-related liver injury (grade ≥ 3), which is the most clinically important because they require therapeutic intervention such as steroids and immunosuppression.
In conclusion, our study found differences in the response to immunosuppression therapy in patients with immune-related liver injury due to liver injury. The clinical course of the hepatocellular pattern is generally favorable; however, the e cacy of steroid therapy is limited in some cases with mixed patterns. Liver biopsy is helpful for diagnosing immune-related liver injury and evaluating its severity. Because patients with immune-related liver injury have advanced malignancies, liver transplantation is generally not a treatment option, even in fulminant hepatitis. Clinicians following malignancies treated by ICIs need to diagnose immune-related liver injury early and select adequate treatment based on the severity and pattern of the injury.  Representative hepatic pathological images in immune-related liver injury (hepatocellular injury pattern) a) portal in ammation with interface hepatitis and spotty necrosis, Scale bar 200μm, b) endothelialitis, Scale bar 100μm, c) Granuloma, Scale bar 50μm, d) in ltrating CD8 positive T cells, Scale bar 100μm.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryinformationHIirAE.pdf