Outcomes of high-grade immune checkpoint inhibitor hepatitis in hospitalized and non-hospitalized patients

Abstract

Background and Aims:

Guidelines recommend hospitalization for severe immune checkpoint inhibitor (ICI) hepatitis. We compared patient outcomes in the inpatient versus outpatient settings.

Methods:

We conducted a multicenter, retrospective cohort study of 294 ICI-treated patients who developed grade 3-4 ICI hepatitis. The primary outcome was time to ALT normalization (≤40); secondary outcomes included time to ALT≤100 U/L and time to death. To account for confounding by indication, inverse probability of treatment weighting (IPTW) was applied to perform Cox regression. A sensitivity analysis was performed excluding patients with grade 4 hepatitis.

Results:

166 patients (56.5%) were hospitalized for a median of 6 [IQR 3-11] days. On IPTW Cox regression, hospitalization was not associated with time to ALT normalization (HR 1.11, 95% CI 0.86-1.43, p=0.436) or time to ALT≤100 U/L (HR 1.11, 95% CI 0.86-1.43, p=0.420). In the sensitivity analysis limited to patients with grade 3 hepatitis, hospitalization was also not associated with time to ALT normalization (HR 1.11, 95% CI 0.83-1.50, p=0.474) or time to ALT ≤100 U/L (HR 1.19, 95% CI 0.90-1.58, p=0.225). In a subgroup analysis of 152 patients with melanoma, hospitalization was not associated with reduced risk of all-cause death (HR 0.93, 95% CI 0.53-1.64, p=0.798). Notably, despite their CTCAE classification of high-grade hepatitis, 94% of patients had “mild” liver injury based on International DILI criteria.

Conclusion:

Hospitalization of patients with high-grade ICI hepatitis was not associated with faster hepatitis resolution and did not affect mortality. Routine hospitalization may not be necessary in all patients with high-grade ICI hepatitis and CTCAE criteria may overestimate severity of liver injury.

INTRODUCTION

Immune checkpoint inhibitors (ICIs) have revolutionized cancer treatment and have become first line treatment for many advanced malignancies. However, ICI treatment is frequently complicated by immune-related adverse events (irAEs), which mimic autoimmune conditions and can affect almost any organ system. Liver involvement (or ICI hepatitis), is one of the most frequent severe irAEs, with up to 10% of ICI-treated patients developing grade 3 or higher hepatitis (aminotransferase elevations over five times the upper limit of normal).^1-3 There is substantial healthcare utilization related to irAEs. 7-11% of all ICI-treated patients are admitted for workup and management of irAEs,^4,5 and total costs for inpatient irAE admissions rose six-fold from 2011 to 2016.⁶ Patients hospitalized for irAE management have substantial healthcare-related costs; mean cost per patient in a 90-day follow-up period after first hospitalization for irAE management was found to be approximately $30,000, which was six times higher than the cost in the same follow-up period after first outpatient visit for irAE management.⁷ At a major academic medical center, irAEs accounted for 5% of all oncology hospitalizations, and nearly all of these patients received a medicine subspecialist consultation.⁸

Major society guidelines generally advise hospitalization for patients with high-grade ICI hepatitis. The American Society of Clinical Oncology suggests that “inpatient monitoring may be offered” and to “consider transfer to tertiary care facility”,^2,9 while the European Society of Medical Oncology recommends hospitalization for all patients with grade 4 ICI hepatitis while suggesting inpatient care for patients with grade 3 hepatitis.^3,10 An American Gastroenterological Association clinical practice update also advises that all patients with grade 4 ICI hepatitis be hospitalized, while in patients with grade 3 ICI hepatitis, “hospitalization for urgent management should be considered on a case by case basis”.¹¹ However, these recommendations are based on expert opinion and studies examining the impact of hospitalization in patients with ICI hepatitis are lacking. We therefore investigated the effect of hospitalization on the outcomes of patients with high-grade ICI hepatitis.

METHODS

Patients and study design

We conducted a multicenter, retrospective cohort study of cancer patients who developed grade 3 or higher ICI hepatitis after receiving ICI therapy at three different cancer centers: the University of California, San Francisco (UCSF), the Dana-Farber/Brigham and Women’s Cancer Center, and the Mass General Cancer Center. Patients treated with at least one ICI who subsequently developed an ALT >200 U/L were identified through the UCSF Clinical Data Warehouse and the Mass General Brigham Research Patient Data Registry. Patient data between 2010 and 2021 was gathered from the Mass General Brigham system, and between 2015 and 2021 from the UCSF system. Individual chart review was performed to determine the etiology of liver injury for patients meeting the search criteria. As in our previous studies,^12,13 ICI hepatitis was defined as ALT elevation in the absence of other probable causes of liver injury, including viral hepatitis, ischemia, or concurrent medications known to cause drug-induced liver injury. Patients who received a diagnosis of ICI hepatitis by their treating providers, did not have any evidence of these alternative causes of liver injury, and received corticosteroid therapy were considered to be true ICI hepatitis cases. In addition, pathology reports were obtained and reviewed for the 100 (34%) patients who underwent a liver biopsy for workup of their liver injury to confirm that they had histologic findings compatible with ICI hepatitis, such as those previously described by our group.¹² Patients with hepatobiliary cancers or autoimmune liver disease, excluded. Patients who did not receive corticosteroid treatment with at least 0.5 mg/kg/day of prednisone equivalents were also excluded. Per Common Terminology Criteria for Adverse Events, grade 3 and grade 4 ICI hepatitis were defined as an ALT>200 U/L (5 times the upper limit of normal) and >800 U/L (20 times the upper limit of normal), respectively. In addition, severity of liver injury was graded according to the International Drug-Induced Liver Injury (DILI) Expert Working Group criteria, with a total bilirubin level of 1.2 mg/dL considered as the upper limit of normal.^14,15 This was chosen over the similar US DILI Network criteria¹⁶ because hospitalization, the exposure investigated in this study, is part of the US DILI Network criteria but not the International DILI Expert Working Group criteria. Institutional review board approval was obtained from the UCSF, Mass General Brigham, and Dana Farber.

Outcomes and covariates

The primary outcome was time to normalization of ALT (defined as ≤40 U/L). Secondary outcomes were time to ALT improvement to ≤100 U/L (i.e., grade 1 injury and the level at which reinitiation of ICI therapy can be considered per guidelines) and time to all-cause death in the subgroup of patients with melanoma. Baseline data at the time of initiation of ICI therapy (demographics, clinical parameters, and laboratory values) were collected. Additional laboratory data were collected at the time of diagnosis of grade 3 or higher ICI hepatitis and at the time of peak ALT. Details regarding treatment of hepatitis were collected, including whether patients were hospitalized, length of hospital stay, corticosteroid dosing and start date, and additional immunosuppressive therapy for steroid-refractory cases. Patients were stratified into two groups: those who were hospitalized for workup and management of their ICI hepatitis and those who remained in the outpatient setting.

Statistical analysis

Baseline patient demographics and covariates are reported as means and standard deviations for continuous normal data, medians and interquartile ranges for continuous non-normal data, and frequencies and percentages for categorical data. Univariable analyses were performed using two-sided Student’s t-test or Wilcoxon rank sum test for continuous variables, Chi-squared test or Fisher’s exact test for categorical variables, and Kaplan-Meier analysis with log-rank testing for time-to-event data. For time to event analyses, patients who did not reach the outcomes being studied were censored at the time of last follow-up or death. The start date for all time to event analyses was the date of first diagnosis of grade 3 or 4 ICI hepatitis. Multivariable Cox regression was employed to adjust for potential confounders. Statistical significance was defined as p<0.05 for all analyses.

For the outcomes of time to ALT normalization and time to ALT improvement to <100 U/L, to address confounding by indication between the two groups (hospitalized versus not hospitalized), inverse probability of treatment weighting was utilized. First, a propensity score model predicting the risk of being hospitalized was constructed with logistic regression using variables available at the time of diagnosis that were different between the two groups. Covariates included in the model were age, sex, BMI, pre-existing liver disease, liver metastases, history of prior immune-related adverse event, ALT level at time of diagnosis, and total bilirubin at time of diagnosis. Inverse probability weights were calculated using 1/(propensity score) for patients in the hospitalized group and using 1/(1 – propensity score) for patients in the not hospitalized group. These weights were used in inverse probability of treatment weighted Cox regression to calculate propensity score-adjusted hazard ratios for the primary and secondary outcomes associated with hospitalization as compared to outpatient management. Covariate balance following inverse probability of treatment weighting was measured by standardized mean difference, and variables were considered to be balanced with an absolute standardized mean difference of less than 0.1 after weighting.^17,18 A sensitivity analysis was performed in the subset of patients with grade 3 hepatitis given the expected imbalance in hospitalized patients with grade 4 hepatitis. Standard Cox regression was utilized to adjust for potential confounders in the sensitivity analysis. Covariates included in the Cox regression model were those that were imbalanced on univariable analysis or those that were considered a priori to be associated with the primary and secondary outcomes (e.g., alcohol use, pre-existing liver disease, liver metastases, and aminotransferase levels). Statistical analyses were performed using SAS version 9.4 (Cary, NC, USA) and R version 4.3 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

A total of 7046 patients at Mass General Brigham and 3392 patients at UCSF received ICI therapy for non-hepatocellular carcinoma cancers. Across the 3 study centers, 294 patients (2.8%) developed grade 3-4 ICI hepatitis (Supplemental Figure S1). Of these, 166 patients (56.5%) were hospitalized for a median of 6 [3-11] days, while the remaining 128 patients were not admitted to the hospital for workup or management. Notably, despite all patients being classified as having high-grade ICI hepatitis based on CTCAE criteria, only 19 (6.4%) patients had moderate liver injury based on DILI severity classification, with the remaining patients all being classified as having mild liver injury. Baseline characteristics of the two groups are shown in Table 1. Hospitalized patients were younger (58.0 +/− 14.6 vs 61.8 +/− 14.9 years of age, p=0.030), were more likely to be male (53.6% vs 40.6%, RR 1.32, 95% CI 1.02-1.70, p=0.027), and were more likely to have pre-existing liver disease (27.7% vs 17.2%, RR 1.61, 95% CI 1.03-2.54, p=0.034); other baseline characteristics were similar, including laboratory values prior to ICI treatment and type of cancer. Table 2 presents details of the ICI hepatitis episodes stratified by hospitalization status. While there was no difference between the two groups in terms of their ICI regimens, patients who were hospitalized had significantly higher AST (312 [187-494] vs 227 [168-333], p<0.001) and ALT (347 [248-635] vs 284 [233-409], p=0.001) levels at time of diagnosis than patients who were not hospitalized. Hospitalized patients were also significantly more likely to develop grade 4 liver injury (31.3% vs 7.8%, RR 4.01, 95% CI 2.12-7.58, p<0.001), have moderate DILI severity (10.2% vs 1.6%, RR 6.55, 95% CI 1.54-27.86, p=0.003), receive a liver biopsy (41.6% vs 24.2%, RR 1.72, 95% CI 1.20-2.45, p=0.002), develop steroid-refractory hepatitis (39.2% vs 12.5%, RR 3.13, 95% CI 1.91-5.14, p<0.001), and have a prior history of a grade 2 or higher irAE (51.8% vs 30.5%, RR 1.70, 95% CI 1.26-2.30, p<0.001). Regarding treatment, both groups received prompt corticosteroid treatment at a median of 0 days following first grade 3 or higher ALT elevation, although hospitalized patients received significantly higher median corticosteroid doses (2 [1-2] vs 1 [1-1] mg/kg/day, p<0.001).

Table 1 –

Baseline characteristics and labs

Characteristic	Not hospitalized (n=128)	Hospitalized (n=166)	p-value
Age	61.8 +/− 14.9	58.0 +/− 14.6	0.030
Males	52 (40.6%)	89 (53.6%)	0.027
Ethnicity			0.827
White	110 (85.9%)	142 (85.5%)
Black	2 (1.6%)	1 (0.6%)
Hispanic	6 (4.7%)	10 (6.0%)
Other/unknown	10 (7.8%)	13 (7.8%)
Ethnicity, white	110 (85.9%)	142 (85.5%)	0.923
Body Mass Index	25.7 +/− 5.2	26.9 +/− 5.9	0.084
Nonsmoker	96 (75.0%)	119 (71.7%)	0.525
Pre-existing liver disease	22 (17.2%)	46 (27.7%)	0.034
Non-alcoholic fatty liver disease	17 (13.3%)	34 (20.5%)
Alcohol-associated liver disease	5 (3.9%)	9 (5.4%)
Other etiology	0 (0%)	3 (1.8%)
Alcohol use	20 (15.6%)	25 (15.1%)	0.894
Cancer type
Melanoma	64 (50.0%)	88 (53.0%)
Non-small cell lung cancer	12 (9.4%)	17 (10.2%)
Renal cell carcinoma	9 (7.0%)	16 (9.6%)
Breast cancer	10 (7.8%)	5 (3.0%)
Urothelial cancer	6 (4.7%)	6 (3.6%)
Other	27 (21.1%)	34 (20.5%)
Melanoma	64 (50.0%)	88 (53.0%)	0.608
Stage 4 cancer	102 (79.7%)	141 (84.9%)	0.238
Liver metastases	25 (19.5%)	48 (28.9%)	0.065
Baseline labs
AST (U/L)	23 (18-28)	20 (17-26)	0.019
ALT (U/L)	18 (15-26)	19 (14-27)	0.899
Alkaline phosphatase (U/L)	76 (62-97)	76 (61-93)	0.428
Total bilirubin (mg/dL)	0.4 (0.3-0.7)	0.5 (0.3-0.7)	0.409
Albumin (g/dL)	4.1 (3.8-4.4)	4.2 (3.8-4.4)	0.349
Creatinine (mg/dL)	0.8 (0.7-1.0)	0.9 (0.7-1.0)	0.256
Platelet count (x109/L)	221 (194-285)	233 (196-270)	0.825

Data are mean +/− SD, median (IQR), or n (%)

Table 2 –

Details of hepatitis

	Not hospitalized (n=128)	Hospitalized (n=166)	p-value
ICI class at time of hepatitis
Anti-PD-1 monotherapy	54 (42.2%)	63 (38.0%)	0.462
Anti-PD-L1 monotherapy	6 (4.7%)	11 (6.6%)	0.480
Anti-CTLA-4 monotherapy	15 (11.7%)	16 (9.6%)	0.565
Combination anti-CTLA-4 and anti-PD-1/PD-L1 therapy	53 (41.4%)	75 (45.2%)	0.518
Prior irAE	39 (30.5%)	86 (51.8%)	<0.001
Grade 4 liver injury	10 (7.8%)	52 (31.3%)	<0.001
DILI severity classification			0.003
Mild	126 (98.4%)	149 (89.8%)
Moderate	2 (1.6%)	17 (10.2%)
Severe	0 (0%)	0 (0%)
Received liver biopsy	31 (24.2%)	69 (41.6%)	0.002
Treatment of hepatitis
Maximum steroid dose, mg/kg	1 (1-1)	2 (1-2)	<0.001
Days to initiation of steroids from diagnosis	0 (0-3.5)	0 (0-2)	0.835
Developed steroid-refractory hepatitis	16 (12.5%)	65 (39.2%)	<0.001
Treated with mycophenolate mofetil	16 (100%)	57 (87.7%)	0.346
Treated with another agent	0 (0%)	8 (12.3%)	0.346
Hospitalization days	N/A	6 [3-11]
Labs at time of diagnosis
AST (U/L)	227 (168-333)	312 (187-494)	<0.001
ALT (U/L)	284 (233-409)	347 (248-635)	0.001
Alkaline phosphatase (U/L)	159 (108-321)	198 (101-366)	0.457
Total bilirubin (mg/dL)	0.6 (0.4-1.0)	0.8 (0.5-1.3)	0.023
Albumin (g/dL)	3.8 (3.5-4)	3.6 (3.2-4)	0.002
Creatinine (mg/dL)	0.8 (0.7-1.1)	0.8 (0.7-1.1)	0.764
Platelet count (x109/L)	223 (182-276)	215 (158-257)	0.115
Outcomes
ALT normalization	122 (95.3%)	157 (94.6%)	0.777
Time to ALT normalization	37 [21-72]	43 [22-69]	0.763
ALT improvement to ≤100 U/L	128 (100%)	163 (98.2%)	1
Time to ALT≤100 U/L	20 [13-36]	20 [12-37]	0.840

Data are mean +/− SD, median (IQR), or n (%)

Time to event analyses, ALT improvement

In terms of the primary and secondary outcomes related to ALT improvement, 279 patients (94.9%) experienced complete ALT normalization and 291 patients (99.0%) experienced ALT improvement to ≤100 U/L. When comparing the 166 patients who were hospitalized and the 128 patients who were not, there was no difference between the two groups in time to ALT normalization (median 43 vs 37 days, log-rank p=0.915; Figure 1A) or in time to ALT improvement to ≤100 U/L (median 20 vs 20 days, log-rank p=0.792; Figure 1B).

Figure 1. Kaplan-Meier curves, stratified by hospitalization status.

A: Time to ALT normalization (ALT ≤40 U/L)

B: Time to ALT improvement to ≤100 U/L

A propensity score model predicting the risk of hospitalization was constructed using logistic regression. Covariates included in the model were age, sex, BMI, pre-existing liver disease, liver metastases, history of prior immune-related adverse event, ALT level at time of diagnosis, and total bilirubin at time of diagnosis. Following inverse probability of treatment weighting, the propensity score as well as all individual covariates included in the model were balanced with standardized mean differences of less than 0.1. Figure 2 is a Love plot displaying the standardized mean differences of each of the variables pre- and post-adjustment.

Figure 2. Love Plot evaluating covariate balance after inverse probability of treatment weighting.

Following confirmation of covariate balance, inverse probability of treatment weighted Cox regression was performed. Hospitalization was not associated with time to ALT normalization (HR 1.11, 95% CI 0.86-1.43, p=0.0.436) or time to ALT improvement to ≤100 U/L (HR 1.11, 95% CI 0.86-1.43, p=0.420) after using inverse probability of treatment weighting to adjust for all of the covariates included in the propensity score model.

Sensitivity analysis – patients with grade 3 hepatitis

Given the expected imbalance in grade 4 ICI hepatitis patients (who were significantly more likely to be hospitalized), a sensitivity analysis was performed in the subset of study patients with grade 3 hepatitis. Supplemental Table S1 shows baseline characteristics and clinical details in the 232 patients with grade 3 hepatitis, stratified by hospitalization status. Unlike the overall cohort, clinical characteristics and laboratory values were generally similar in the sensitivity analysis regardless of hospitalization status, including pre-existing liver disease, aminotransferase levels at time of peak ALT, and total bilirubin levels. Hospitalized patients in the sensitivity analysis were more likely to have a prior history of a grade 2 or higher irAE (48.2% vs 30.5%, RR 1.58, 95% CI 1.13-2.20, p=0.006) and had higher AST levels at time of diagnosis (255 [162-304] vs 219 [162-304] U/L, p=0.040) than non-hospitalized patients, though they had similar ALT levels at time of diagnosis (307 [234-392] vs 273 [230-362] U/L, p=0.350). Hospitalized patients in the sensitivity analysis also received higher corticosteroid doses (2 [1-2] vs 1 [0.8-1] mg/kg/day, p<0.001) than non-hospitalized patients, though both groups received prompt corticosteroid treatment a median of 0 days after first grade 3 ALT elevation.

There was no difference between hospitalized and non-hospitalized patients in time to ALT normalization (median 35 vs 37 days, log-rank p=0.501; Supplemental Figure S2A) or time to ALT improvement to ≤100 U/L (median 15 vs 20 days, log-rank p=0.519; Supplemental Figure S2B). After multivariable adjustment for age, sex, pre-existing liver disease, alcohol use, liver metastases, prior irAE development, and ALT and AST level at time of diagnosis using standard Cox regression, hospitalization was not associated with time to ALT normalization (HR 1.11, 95% CI 0.83-1.50, p=0.474) or time to ALT improvement to ≤100 U/L (HR 1.19, 95% CI 0.90-1.58, p=0.225) (Supplemental Table S2).

Survival analysis

We restricted the survival analysis to the subgroup of 152 patients with melanoma (Supplemental Table S3) to avoid confounding by including different cancer types. Fifty-seven patients with melanoma died during the study period, with no deaths attributed to liver failure. Hospitalized melanoma patients were more likely to have pre-existing liver disease (27.3% vs 10.9%, RR 2.5, 95% CI 1.1-5.4, p=0.014) and liver metastases (3.75% vs 21.9%, RR 1.7, 95% CI 1.0-2.9, p=0.040) than non-hospitalized melanoma patients; other baseline clinical characteristics and type of ICI regimen received were similar between the two groups. There was no significant difference in time to all-cause death between the 88 melanoma patients who were hospitalized and the 64 melanoma patients who were not (median not reached in the hospitalized group vs median of 4.9 years in the non-hospitalized group, log-rank p=0.884; Figure 3). After adjusting for age, pre-existing liver disease, presence of liver metastases, and combination ipilimumab and nivolumab therapy using standard Cox regression, there remained no difference in the risk of death comparing the hospitalized versus non-hospitalized patients (HR 0.93, 95% CI 0.53-1.64, p=0.798).

Figure 3. Kaplan-Meier curve, time to all-cause death in melanoma subgroup stratified by hospitalization status.

DISCUSSION

Current guidelines suggest that patients with severe ICI hepatitis should be hospitalized for management. In a large, real-world cohort of 294 patients across three tertiary care cancer centers, we found that hospitalization was not associated with improved outcomes, including time to hepatitis resolution and time to all-cause death. As all patients received prompt corticosteroid treatment a median of 0 days following their first grade 3 or higher ALT elevation, it is perhaps unsurprising that hospitalization did not impact clinical outcomes. These data suggest that routine hospitalization for high-grade ICI hepatitis may not be necessary; however, we acknowledge that multiple factors may influence decision-making regarding the need for inpatient workup and management. For example, patients with ICI hepatitis presenting with fevers in conjunction with their elevated liver tests likely warrant hospitalization for infectious workup and biliary imaging to rule out bacterial cholangitis. However, once alternative etiologies have been excluded, our findings suggest that patients need not remain hospitalized for ongoing treatment of their ICI hepatitis, assuming good compliance and ready access to outpatient laboratory monitoring for frequent liver tests to assess for steroid treatment response and, if necessary, management of steroid-refractory disease.

The most important point of consideration when assessing these results is that following current major society recommendations to grade ICI hepatitis using CTCAE criteria may overestimate patient risk. Aminotransferase elevation alone does not convey the true severity of liver injury as liver synthetic function (as measured by total bilirubin and INR) can be unaffected even when ALT values are markedly elevated. This was seen in our patient cohort, in which 94% of patients were graded as “mild” liver injury based on the International DILI Expert Working Group criteria despite having grade 3 or 4 elevations in ALT. Aminotransferase elevations alone do not reflect the severity of liver injury or liver dysfunction, unlike bilirubin and INR levels, and because aminotransferase elevations alone are sufficient per CTCAE criteria for patients to have high-grade injury, a classification of high-grade ICI hepatitis is discordant with liver injury severity in a substantial proportion of patients. While all current guidelines continue to recommend using CTCAE criteria for classification of injury severity, the addition of DILI-specific criteria may more accurately risk stratify patients with ICI hepatitis.

The prompt corticosteroid therapy that study patients received could feasibly have contributed to limiting the severity of liver injury in both groups. Given the small number of patients who experienced more severe liver injury as assessed by the DILI Expert Working Group criteria, we cannot draw conclusions from this study regarding the safety of outpatient management in patients with moderate to severe liver injury based on DILI criteria and suggest that patients with bilirubin and/or INR elevations due to ICI hepatitis in conjunction with aminotransferase abnormalities be managed in the inpatient setting. Our findings also suggest that CTCAE criteria may not adequately risk-stratify patients as marked aminotransferase elevations are considered to be equally “severe” as laboratory derangements indicating true underlying liver synthetic dysfunction.¹⁹

While studies have evaluated risk factors associated with serious irAEs requiring hospitalization and have described hospitalization rates, treatment, and outcomes of these irAEs, to the best of our knowledge, this is the first study to evaluate the influence of hospitalization on outcomes of a serious irAE. Additional strengths of this study include its relatively large cohort size and the inclusion of patients from multiple tertiary care centers. The major limitation of this retrospective study as in any pharmacoepidemiology study is confounding by indication. We did attempt to address this concern using inverse probability of treatment weighted Cox regression, and all covariates included in the propensity score model predicting risk of hospitalization were balanced following weighting. In addition, in the sensitivity analysis limited to patients with grade 3 ICI hepatitis, the hospitalized versus non-hospitalized groups were more balanced (due to the exclusion of patients with grade 4 ICI hepatitis who were predominantly hospitalized), suggesting a lesser degree of confounding by indication. Potential confounders were also adjusted for using Cox regression in the sensitivity analysis which produced similar findings to the overall cohort. However, the possibility of unmeasured confounders remains due to the retrospective design of this study.

While both groups received prompt corticosteroid treatment, hospitalized patients received higher doses than non-hospitalized patients, raising the possibility that the more “aggressive” treatment in hospitalized patients may have been responsible for balancing the outcomes between the two groups. However, we previously have demonstrated that patients with high-grade ICI hepatitis receiving median steroid doses of 1 mg/kg/day have similar outcomes as those receiving 2 mg/kg/day.¹³ Furthermore, including steroid dose in the final weighted Cox regression models for time to ALT normalization and time to ALT improvement to ≤100 U/L did not alter the findings, nor was steroid dose associated with faster hepatitis resolution.

There is risk that retrospective determination of ICI hepatitis cases could potentially lead to misclassification, even though chart review was performed by hepatologists and other probable causes of liver injury were screened for and excluded (e.g., viral hepatitis, ischemia, other DILI precipitants). Against this possibility is that 99% of patients in our study experienced ALT improvement to <100 U/L and 95% of patients experienced complete ALT normalization following ICI discontinuation and corticosteroid treatment. Furthermore, of the 34% of patients in the study who underwent liver biopsy, all demonstrating histologic findings compatible with ICI hepatitis.

Finally, while a relatively small proportion of patients in our cohort had pre-existing liver disease, we specifically excluded patients with hepatocellular carcinoma from this and prior studies. As almost all hepatocellular carcinoma patients have underlying cirrhosis and are therefore a uniquely vulnerable patient population with regards to risk of hepatic decompensation, our findings should not be generalized to liver cancer patients or patients with cirrhosis.

In summary, our data suggest that patients hospitalized for high-grade ICI hepatitis do not experience faster hepatitis resolution compared to those who are managed in the outpatient setting. Additionally, the subset of patients with melanoma had similar overall survival regardless of inpatient versus outpatient management. As current guidelines recommend hospitalization for patients with grade 3 or higher ICI hepatitis, we believe that our data support a shift away from routine hospitalization for all patients with high-grade hepatitis. We propose that patients can undergo outpatient workup for elevated liver tests suspected to be due to ICI hepatitis and treatment-dose corticosteroids can be initiated without the need for hospitalization. We acknowledge that select patients, such as those manifesting fevers or an atypical liver enzyme pattern, may still require hospitalization initially before transitioning to outpatient management, and warn that our data does not apply to those with hepatocellular carcinoma who are likely to have underlying liver cirrhosis.

What You Need to Know:

Background:

While clinical guidelines suggest that patients with high-grade immune checkpoint inhibitor hepatitis should be admitted, the impact of hospitalization on patient outcomes is unknown.

Findings:

After accounting for confounding by indication through inverse probability of treatment weighting, hospitalization was not associated with faster hepatitis improvement or resolution. Notably, almost all patients with high-grade hepatitis still only had “mild” liver injury based on International DILI criteria.

Implications for patient care:

CTCAE criteria may overestimate severity of true liver injury. Patients with typical laboratory findings, no evidence of liver synthetic dysfunction, and without any symptoms concerning for infection may be able to avoid hospitalization and undergo outpatient workup and treatment for high-grade ICI hepatitis.

List of abbreviations:

ICI

immune checkpoint inhibitor

irAE

immune-related adverse event