Clinical Course and Treatment Implications of Combination Immune Checkpoint Inhibitor-Mediated Hepatitis: A Multicentre Cohort

Abstract Background Immune-related adverse events can occur after treatment with immune checkpoint inhibitors (ICI), limiting treatment persistence. We aimed to evaluate the clinical course of ICI-mediated hepatitis (IMH) associated with combination ipilimumab and nivolumab treatment. Methods A retrospective cohort study including consecutive patients with metastatic melanoma treated with ipilimumab and nivolumab between 2013 and 2018 was conducted at two tertiary care centres. IMH was defined by the Common Terminology Criteria for Adverse Events (CTCAE). We determined the proportion of patients developing IMH, and compared the duration, treatment patterns and outcomes, stratified by hepatitis severity. Kaplan–Meier survival analysis was used to evaluate time to hepatitis resolution, and a linear mixed-effects model was used to compare longitudinal outcomes by treatment. Results A total of 63 patients were included. Thirty-two patients (51%) developed IMH (34% Grade 1–2, 66% Grade 3–4), at a median of 34 days (IQR 20 to 43.5 days) after the first dose. Baseline FIB4 index ≥1.45 was associated with IMH (OR 3.71 [95% CI: 1.03 to 13.38], P = 0.04). Ninety-four per cent (30/32) of patients had liver enzyme normalization after a median duration of 43 days (IQR 26 to 70 days). Corticosteroid use was not associated with faster IMH resolution or less ICI discontinuation. A total of 24 patients died during the study; no deaths were attributable to hepatitis-related complications. Fifty-three per cent (17/32) of patients resumed anti-PD-1 monotherapy and three patients developed IMH recurrence. Conclusions Approximately half of the patients treated with combination ipilimumab and nivolumab developed IMH in this cohort. However, most patients experienced uncomplicated IMH resolution.

Combination ICI regimens may have more potent antineoplastic effects in some patients, but are also associated with a greater degree of treatment-related morbidity from immunerelated adverse events (irAEs) (10). The immune checkpoint pathways are important regulators of self-tolerance. Therefore, treatment with ICIs can result in irAEs affecting nearly all organ systems (11). In a meta-analysis of randomized controlled trials, Da et al. demonstrated that approximately 1 in 10 patients treated with ICI monotherapy developed immune-mediated hepatitis (IMH) and combination ICI treatment was associated with a nearly threefold increased risk of severe disease (12). Other retrospective cohort studies have demonstrated incidences of IMH as high as 20% (13,14).
The development of irAEs carries therapeutic implications as severe or recurrent irAEs can be life threatening and may limit continuation of ICI treatment. While the combination of ICIs is known to increase the risk of IMH, specific patient-or diseaserelated factors that are predictive of IMH have not yet been identified. Similarly, there are no known predictors of IMH severity or disease course. Current guidelines from the Society for Immunotherapy of Cancer Toxicity Management Working Group and American Society of Clinical Oncology recommend treatment of IMH with immunosuppressive therapy such as corticosteroids alongside withholding immunotherapy in patients with grade 2 or more severe disease (15,16). However, it is unclear if all patients require these interventions.
Therefore, we aimed to characterize the clinical course, treatment, and outcomes of patients developing IMH in a retrospective, multicentre cohort of patients treated with combination nivolumab and ipilimumab for advanced or metastatic melanoma.

Study Design, Setting, and Data Source
This retrospective cohort study was conducted using data collected from consecutive patients with unresectable stage 3 or 4 melanoma treated with combination ipilimumab and nivolumab at the Tom Baker Cancer Centre, Calgary, Canada and Cross Cancer Institute, Edmonton, Canada, between 2013 and 2018. Patient data from both centres were collected from a shared, comprehensive provincial electronic health record, which contains information on outpatient and inpatient visits, laboratory data, diagnostic imaging and pharmaceutical prescriptions.

Patient Population
Our study cohort was identified from a clinical database of melanoma patients from the Tom Baker Cancer Centre, Calgary and Cross Cancer Institute, Edmonton. Demographics, disease staging and treatment information was registered by the patient's primary oncologist and corroborated using the provincial cancer electronic medical record, which contains data on all prescription medication use including anti-neoplastic therapies. Eligible patients with advanced melanoma received induction nivolumab 1 mg/kg and ipilimumab 3 mg/kg administered intravenously (IV) once every 3 weeks for up to four cycles followed by nivolumab 1 mg/kg IV every 2 weeks maintenance therapy. All patients who received at least one dose of combination treatment were included. Patients with cirrhosis, chronic infection with hepatitis B or C, or known co-existing liver diseases were excluded. Monitoring of bloodwork, including alanine transferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP) and total bilirubin were performed prior to each treatment per protocol. Measurements of the international normalized ratio (INR) and albumin were performed at the treating physician's discretion. All patients had cross-sectional imaging for disease staging prior to the initiation of treatment.

Outcomes and Definitions
The primary outcome of interest was the development of IMH, as per the treating physician's judgment, while receiving the first four cycles of induction combination nivolumab with ipilimumab treatment, graded using the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. Subsequently, detailed medical chart review for all ICI-related adverse events was conducted (including IMH) and verified. The CTCAE grade was calculated for each patient using their corresponding laboratory values. Grade 1-2 hepatitis was defined by an asymptomatic rise in the AST or ALT ≤5× the upper limit of normal (ULN), with a Roussel Uclaf Causality Assessment Method (RUCAM) score ≥6, indicating probable drug-induced liver injury (17). Grade 3 hepatitis was defined by symptomatic liver dysfunction, AST or ALT 5-20× ULN, or bilirubin 3-10× ULN. Grade 4 hepatitis was defined by decompensated liver function, AST or ALT > 20× ULN, or total bilirubin >10× ULN. The ULN was defined using sex-appropriate cut-offs (males: ALT 33 U/L, AST 40 U/L, ALP 115 U/L; females: ALT 25 U/L, AST 32 U/L, ALP 100 U/L). Baseline liver enzymes were defined as the liver enzyme profile of the patient within one month prior to initiating combination ICI treatment. For patients with abnormal baseline liver enzymes, these definitions were adapted to reflect change relative to the baseline rather than the ULN. During the study period, liver biopsies were not routinely performed in our centres after the initiation of combination ICI therapy to assess for disease-specific histopathology. Secondary outcomes of interest included time to development of IMH and time to peak liver enzyme elevation (from the first dose of combination ICI treatment), total hepatitis duration (defined from the time of onset to time of liver chemistry normalization), post-hepatitis ICI treatment and death. Treatment of IMH and non-chemotherapeutic drug use was collected from each patient's electronic medical record: sources searched for medication use included physician notes for treatment, histories and discharge summaries, and the patient's medication profile. IMH treatment was categorized as: 1) no treatment; 2) temporary or permanent combination ICI discontinuation; 3) corticosteroid use and 4) biologic treatment. Corticosteroid use was further stratified by route of administration (intravenous versus oral). Corticosteroid dose was decided by the treating physician. Other variables of interest included patient age, sex, cancer staging, previous anti-neoplastic therapy, concurrent liver disease (including radiographic evidence of hepatic steatosis by radiologist report from computed tomography or ultrasound assessment within 12 months of combination ICI use, or suspected liver metastases), baseline Fibrosis-4 [FIB4] index and development of other irAEs (including ICI-related enterocolitis or diarrhoea) as documented by the treating physician.

Statistical Methods
Descriptive statistics for the cohort are presented as medians with interquartile range for nonparametrically distributed continuous variables, and proportions for categorical variables. Differences in baseline characteristics between patients developing IMH and patients without IMH were evaluated using the Mann-Whitney U test for distribution of nonparametric continuous variables, or Pearson Chi-squared test for proportions. Differences in paired continuous data between baseline, onset, peak and resolution of hepatitis were evaluated using the Wilcoxon signed-rank test. To evaluate time to onset of hepatitis, time to peak liver enzyme elevation and time to hepatitis resolution, lifetables were constructed and analysed using the Kaplan-Meier method. Differences between IMH CTCAE grade groups were evaluated using the log rank test. Changes in the relative value of liver enzymes to the sex appropriate ULN from baseline to onset of hepatitis, peak liver enzyme elevation, and after resolution are depicted graphically in violin plots. Violin plots show the median, interquartile range and probability density of data at different values, smoothed using a kernel density estimator. A univariable logistic regression was used to identify associations between baseline factors and development of IMH, expressed as unadjusted odds ratios (OR) with 95% confidence intervals (CI). Multivariable modelling was not performed due to the small sample size.
All analyses were conducted in STATA 16.1 (StataCorp LLC, College Station, TX). The Health Research Ethics Board of Alberta, Cancer Committee approved the study with patient informed consent waived (CC-17-0435).

Patient Demographic Characteristics
A total of 63 patients with malignant melanoma treated with combination ipilimumab-nivolumab were included in the cohort. Baseline demographic characteristics stratified by development of IMH are summarized in Table 1. All patients were started on therapy for advanced melanoma (92% for stage IV, 8% for unresectable stage III). A total of 33 patients (52%) were BRAF mutation positive, although only 10 patients received dabrafenib and/or trametinib as first-line agents. No patients received either dabrafenib or trametinib within 30 days prior to starting combination ICI. Eight patients (13%) had uveal melanoma, 55 (87%) had cutaneous melanoma. Median age at time of combination ICI therapy was 55 years (IQR 44 to 61 years). A total of 28 patients (44%) underwent four complete cycles of combination ICI therapy, whereas 6 patients (10%) only completed a single cycle of treatment. Viral hepatitis serology was re-assessed in 37 of 63 patients, with no patients identified as being positive. No patients were identified as having significant alcohol use disorder. Radiographic evidence of hepatic steatosis was demonstrated in 13% (8/63) of patients and 41% (26/63) of patients had evidence of hepatic metastases. Concurrent irAEs were present in 34/63 patients (54%), including immune-related dermatitis, hypophysitis, myositis and nephritis. Checkpoint inhibitor associated enterocolitis developed in 26 patients (41%).

Treatment of IMH
Corticosteroid treatment was initiated in 97% (31/32) of patients who developed IMH during the study period. Five patients (16%) were admitted as inpatients for treatment of IMH. There were four patients (13%) who developed IMH while already on corticosteroid therapy. The mean induction corticosteroid dose was 69 mg (± 23 mg) prednisoneequivalent/day (adjusted for weight, mean dose of 0.86 mg/kg (± 0.21 mg/kg). Intravenous corticosteroids were used in five patients (16%) at induction. One patient required a single dose of intravenous 5 mg/kg infliximab for management of IMH that was unresponsive to corticosteroids. Of note, no patients within our cohort received MMF or cyclosporine. Combination ICI treatment was discontinued in 62.5% (20/32) of patients who developed IMH.

IMH Resolution
Of the 32 patients who developed IMH, 30/32 (94%) had eventual resolution and normalization of liver enzymes (Figure 1

Other Outcomes
Among the 32 patients who developed IMH, 17 patients (53%) restarted post-combination therapy with PD1 inhibitor monotherapy. Three patients (18%) restarting a PD1 inhibitor re-developed IMH (two patients with Grade 3 IMH, one patient with Grade 2 IMH). A total of 24 patients (38%) died in the cohort, most commonly from sequalae of their progressive metastatic disease manifesting as multi-organ dysfunction. Cause of death was obtained from the last physician documentation in the death summary from the patient's electronic Figure 1. Violin plots demonstrating the distribution of alanine aminotransferase (A), aspartate transaminase (B), alkaline phosphatase (C) and total bilirubin (D) from baseline (blue) to onset of ICI-mediated hepatitis (red), peak hepatic inflammation (green) and hepatitis resolution (yellow). White circles represent median, boxes represent interquartile range, width of violin plots represent probability density of data at each value relative to the upper limit of normal. medical record; no deaths directly attributable to hepatitis were identified.

Discussion
Understanding the clinical course of ICI-mediated adverse events, such as IMH, is particularly relevant as combination immunotherapy becomes the standard of care, despite a greater risk of immune-related complications. In this multicentre retrospective cohort study of 63 patients with unresectable melanoma treated with combination ipilimumab and nivolumab, we found that approximately half of patients developed significantly elevated liver enzymes likely secondary to IMH, with no patients developing fulminant liver failure. Almost all patients had complete normalization of liver enzymes over time, with no significant differences in hepatitis outcomes demonstrated  between IMH grade or with higher dose corticosteroid exposure.
In a recent systematic review, Peeraphatdit et al. determined that the incidence of IMH in patients undergoing ICI treatment was highly variable, ranging from 0.7% to 16% depending on class of therapy and therapeutic dose (18). In contrast, a higher proportion of patients developed liver injury after ICI exposure in our study, in keeping with data from the French CERTIM group (Immunomodulatory Therapies Multidisciplinary Study group), who showed that 23% of patients exposed to any ICI therapy developed grade 3-4 IMH (19). The authors hypothesized that both selection and reporting bias favouring severe cases of hepatitis was associated with differences in the reported incidence. Furthermore, IMH remains a clinical diagnosis of exclusion, and is dependent on the evaluation for other causes of hepatotoxicity. A diagnosis of exclusion can be particularly challenging in the neoplastic setting if there are concurrent infiltrating liver metastases (which may undergo necrosis with ICI treatment) or other immune-related adverse events such as myositis or myocarditis with secondary transaminitis (20)(21)(22). However, we excluded patients with liver metastases in a sensitivity analysis, and this did not appreciably change our point estimates. While we applied the RUCAM score to identify patients with probable to highly probable IMH, the discriminatory capacity of RUCAM for IMH requires validation because the hypothesized injury mechanism is related to indirect hepatotoxicity from immune activation, rather than conventional drug-induced liver injury from direct or idiosyncratic hepatotoxicity (23).
The lack of a diagnostic gold standard for IMH has limited the development of consensus diagnostic criteria or predictors based on biologic or histopathologic features (24,25). In our study, higher baseline FIB4 index was associated with a greater risk of IMH, although we were unable to adjust for other confounders in multivariable logistic regression analysis due to inadequate power. While FIB4 has only been validated in the study of chronic liver disease, we hypothesize that higher baseline FIB4 index may reflect a degree of underlying liver disease that was not identified pre-treatment, as no patients in our study underwent a liver biopsy as part of routine care pretreatment. Radiographic hepatic steatosis, presence of hepatic metastases at baseline (26), and development of other irAEs or ICI-associated enterocolitis were not significantly associated with IMH development, although we are limited by sample size. Currently, no biomarkers or clinical features reliably predict IMH. Potential candidate biomarkers include interleukin (IL)-6 (27), soluble CD163 and CXCL5 (28), and IL17 (29). However, these require independent validation prior to integration into routine clinical practice.
Almost all patients in our cohort were treated with corticosteroids and most patients had ICIs temporarily or permanently withheld in accordance with current clinical practice guidelines (15,16,30). As previously described, most cases of IMH appear to resolve independent of steroid dose and the efficacy of high dose corticosteroids for other causes of druginduced liver injury remains contentious (31). Corticosteroid exposure is also associated with a sevenfold increased risk of infections in this population and there is a theoretical risk that immunosuppression may decrease the efficacy of ICIs (32,33). In our study, similar to Cheung et al., we did not demonstrate a significant benefit with higher dose corticosteroid exposure with respect to the continuation of combination ICI therapy, resolution or time to improvement of IMH (13). While we recognize the necessity for treatment in patients with CTCAE grade 3 and 4 IMH to avoid fulminant hepatitis and subsequent complications, the current recommendations for corticosteroid dosing in IMH requires evaluation in a prospective trial.
For patients who have IMH resolution after discontinuation or completion of combination ICI treatment, the benefits of restarting ICI monotherapy should be considered with the potential risk of irAE reactivation. In our cohort, 14/17 patients were able to safely restart anti-PD1 monotherapy after IMH, which mirrors the positive re-exposure experience reported by other authors (34). Reassuringly, we did not observe any cases of drug-induced acute liver failure in our cohort, in keeping with other studies demonstrating a 0.1% to 0.2% risk of ICIinduced fulminant hepatic failure.
Our study has several important limitations. First, our study was conducted retrospectively, and the inherent limitations of this study design should be considered, including the risk of recall and selection bias. Second, there are limitations around diagnostic definitions of IMH that have been previously discussed. As this is a diagnosis of exclusion, it is particularly difficult to discern retrospectively. We excluded patients with grade 1-2 hepatitis with low RUCAM scores and performed a sensitivity analysis excluding patients with hepatic metastases. However, some degree of residual diagnostic uncertainty remains. Third, despite combining data at two tertiary care sites, our sample size was relatively small. Fourth, no patients in our cohort underwent a liver biopsy so we are unable to correlate clinical outcomes with histologic patterns of disease, preventing the ability to perform subgroup analyses stratified to histopathologic subtype. Finally, the treatment pattern (withholding ICIs and administration of corticosteroids) was similar for almost all patients, so we were unable to specifically evaluate the efficacy of different management strategies. However, we did apply different methods to evaluate the effect of high versus lowdose corticosteroid exposure, which is an important clinical question in this population.
In summary, we describe a multicentre cohort of patients with advanced melanoma treated with combination ipilimumab and nivolumab, with approximately half of patients developing IMH within 1 to 2 months of the first treatment cycle. We identified that IMH development was associated with a higher baseline FIB4 index, although other demographic and biochemical parameters were not predictive. Almost all patients in our cohort experienced IMH resolution with no persistent hepatic dysfunction, irrespective of corticosteroid dosing. Prospective controlled studies are needed to elucidate the precise role of immunosuppression in patients developing IMH.

SUPPLEMENTARY DATA
Supplementary data are available at Journal of the Canadian Association of Gastroenterology online.
Supplemental Figure 1. Onset of immune checkpoint inhibitor-associated hepatitis by sex, existing liver disease, and concomitant immune-related adverse events.

Funding
None declared.

Authors contribution
M.S.: study design and conception, data collection, manuscript drafting and editing for important intellectual content. Y.C.: study design and conception, data collection, manuscript editing for important intellectual content. A.E.S.: study design and conception, data collection, manuscript editing for important intellectual content. A.A.S.