Steroid-Refractory Immune-Related Adverse Events Induced by Checkpoint Inhibitors

Simple Summary With the common use of immune checkpoint inhibitors (ICIs) as a standard therapy for many tumor entities, the number of immune-related adverse events (irAEs) has increased. Side effect management for irAEs includes the administration of corticosteroids which are mostly very effective. However, a subgroup of patients with checkpoint inhibitor-induced side effects do not adequately respond to steroids with so-called steroid-refractory side effects (sr-irAEs) or cannot be tapered off steroids without the recurrence of side effects, the so-called steroid-dependent side effects (sd-irAEs). Since little is known about the incidence and management of sr/sd-irAEs, we investigated the occurrence and management of these difficult-to-treat side effects. This is the first study to report on the incidence rate of steroid-refractory or steroid-dependent irAEs in patients with skin cancer. Abstract The occurrence, second-line management and outcome of sr/sd-irAEs was investigated in patients with skin cancer. All skin cancer patients treated with immune checkpoint inhibitors (ICIs) between 2013 and 2021 at a tertiary care center were analyzed retrospectively. Adverse events were coded by CTCAE version 5.0. The course and frequency of irAEs were summarized using descriptive statistics. A total of 406 patients were included in the study. In 44.6% (n = 181) of patients, 229 irAEs were documented. Out of those, 146 irAEs (63.8%) were treated with systemic steroids. Sr-irAEs and sd-irAEs (n = 25) were detected in 10.9% of all irAEs, and in 6.2% of ICI-treated patients. In this cohort, infliximab (48%) and mycophenolate mofetil (28%) were most often administered as second-line immunosuppressants. The type of irAE was the most important factor associated with the choice of second-line immunosuppression. The Sd/sr-irAEs resolved in 60% of cases, had permanent sequelae in 28% of cases, and required third-line therapy in 12%. None of the irAEs were fatal. Although these side effects manifest in only 6.2% of patients under ICI therapy, they impose difficult therapy decisions, especially since there are few data to determine the optimal second-line immunosuppression.


Introduction
Immune checkpoint inhibitors (ICIs) have become standard therapy for many tumor entities. They target the PD1-/PD-L1 pathway, the CTLA-4 pathway, or the LAG-3 pathway and are either used as monotherapy or in combination, e.g., with different ICI or chemotherapy [1]. Along with the improved clinical outcome, severe toxicities or socalled immune-related adverse events (irAEs) are induced. Grade 3 and 4 side effects were observed in 6-42% of patients treated with anti-PD1 or anti-PDL1-antibodies [2][3][4][5], in 28% of patients treated with ipilimumab [6], in 59% of patient treated with combined In this study, the occurrence of sr/sd-irAE side effects in patients with skin cancer was investigated in a real-world setting. Additionally, second-line management was compared with treatment algorithms recommended in the management guidelines.

Materials and Methods
Between January 2013 and December 2021, all patients who were treated with an ICI (avelumab, cemiplimab, ipilimumab, nivolumab, pembrolizumab, and/or combined ipilimumab and nivolumab) at a tertiary care university skin cancer center in Munich, Germany, were retrieved from Side Effect Registry Immuno-Oncology (SERIO), an international registry based at the Ludwig-Maximilian University Hospital in Munich that collects cases of irAEs in cooperation with the Paul Ehrlich Institute. The registry contains anonymous data on all patients who were treated with ICI from our department, therefore, informed consent was not mandatory. Once they received systemic tumor therapy, the patients were followed-up at our center until death or long term.
Patients were treated for advanced/metastatic non-melanoma skin cancer and melanoma (including melanoma of unknown primary, uveal melanoma and mucosal melanoma) in an adjuvant setting or as a neoadjuvant therapy. The individual electronic patient files of these patients were screened for the occurrence of irAE, type of irAE, severity, and outcome. At our center, patients are surveyed for potential adverse events (including fatigue for hypophysitis, diarrhea, dyspnea for pneumonitis), physically assessed at each visit (neurological side effect), and their laboratory values are checked (e.g., transaminases for hepatitis) before each infusion. Additionally, the patients received an emergency card with a telephone contact available 24/7 to call in case of side effects. Side effects were attributed to the medication by physicians and grading performed according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0.
IrAE data (type of irAE, severity of irAE, onset of irAE after initiation of ICI therapy) the first-line and second-line management of irAE, time to first response, and outcome of irAE were collected ( Figure 1). Second-line treatments were compared with management recommendations for the treatment of irAEs from ASCO, NCCN, and ESMO guidelines. Descriptive statistics were used to summarize the frequencies and courses of irAEs.
At our center, the decision of systemic tumor therapy is based on the BRAF/c-kit mutation status (for targeted therapy), performance status, and co-morbidities. If there are several treatment alternatives, patients present the pros and cons in terms of efficacy and toxicity to determine the patient preferences. Furthermore, potential study options (e.g., cellular therapies) are presented to the patient.
Continuous data are presented as a median or ranges and categorical data are presented as percentages. Overall survival (OS) was compared between patients with regard to the development, severity and management of irAEs. Log-rank tests were performed to compare OS between two groups (patients with irAEs versus patients without irAE, patients with grade 1/2 irAEs versus patients with grade 3/4 irAEs, and patients with grade 3/4 irAEs who were treated with corticosteroids only versus patients who were treated with second-line immunosuppressants). p values < 0.05 were considered statistically significant. Statistical analyses were conducted with SPSS Version 27.
The institutional review board of the medical faculty of the Munich University Hospital (dating 17 February 2021; 20-1122) approved the data analysis from the SERIO registry. This study was conducted in accordance with the principles of the Helsinki Declaration.

Patients' Characteristics
In total, 406 patients (234 males and 172 females) who were treated with ICIs were identified from our registry. The majority of patients were treated for melanoma (n = 389, 95.8%; including cutaneous melanoma, mucosal melanoma, conjunctival melanoma, and uveal melanoma). Other patients had non-melanoma skin cancer (squamous cell carcinoma, basal cell carcinoma, Merkel cell carcinoma, and eccrine porocarcinoma). In 76.4% of cases, the ICI treatment was performed in patients with advanced/metastatic disease, while 21.2% received ICI as an adjuvant therapy after the complete resection of the tumor or were treated in a neoadjuvant setting (0.2%). Most patients (n = 217, 53.4%) were treated with an anti-PD1/PDL1 antibody (cemiplimab, nivolumab, pembrolizumab, or avelumab) or combined anti-CTLA4-antibody and anti-PD1 antibody (n = 173, 42.6%; ipilimumab and nivolumab). In a total of 181 (44.6%) patients, at least one irAE was documented in the patient chart ( Figure 1, Table 1).

Patients' Characteristics
In total, 406 patients (234 males and 172 females) who were treated with ICIs were identified from our registry. The majority of patients were treated for melanoma (n = 389, 95.8%; including cutaneous melanoma, mucosal melanoma, conjunctival melanoma, and uveal melanoma). Other patients had non-melanoma skin cancer (squamous cell carcinoma, basal cell carcinoma, Merkel cell carcinoma, and eccrine porocarcinoma). In 76.4% of cases, the ICI treatment was performed in patients with advanced/metastatic disease, while 21.2% received ICI as an adjuvant therapy after the complete resection of the tumor or were treated in a neoadjuvant setting (0.2%). Most patients (n = 217, 53.4%) were treated with an anti-PD1/PDL1 antibody (cemiplimab, nivolumab, pembrolizumab, or avelumab) or combined anti-CTLA4-antibody and anti-PD1 antibody (n = 173, 42.6%; ipilimumab and nivolumab). In a total of 181 (44.6%) patients, at least one irAE was documented in the patient chart ( Figure 1, Table 1).

First-Line Management and Outcome of irAEs
First-line treatment of irAEs was performed using systemic corticosteroids in 63.8% (n = 146) of cases and ICI-treatment was continued despite the onset of irAEs without any specific treatment in 7.8% (n = 18). For endocrine toxicities, systemic corticosteroids were usually not administered, but the defective hormone production was addressed by the substitution of the corresponding hormone (n = 38, 16.6%). For the treatment of irAEs affecting easily accessible organs, local treatment was performed. This was the case for topical corticosteroids used for skin toxicities and conjunctivitis (n = 18, 7.9%), intravitreal corticosteroids used for ophthalmic toxicities (n = 1, 0.4%), inhalative corticosteroids used for pulmonary irAEs (n = 1, 0.4%), and intralesional/intraarticular corticosteroids combined with systemic corticosteroids used for neurologic toxicities or arthritis (n = 3, 1.3%). In eight cases (3.5%), ICI therapy was stopped due to the occurrence of an irAE without the initiation of a specific treatment.
Complete resolution after first-line treatment was documented in 57.2% (n = 131), whereas no resolution but the improvement of symptoms was seen in 13.1% (n = 30), and no improvement at all occurred in 2.6% (n = 6). Permanent hormone substitution after the insufficiency of endocrine glands was required in 16.2% (n = 37). Second-line immunosuppressants were required in 10.9% (n = 25) of all side effects.
In responsive adverse events, a significant improvement was first detected between a period of one and 56 days. In irAEs, which were treated by hormone substitution, time to first improvement was shortest with a mean of 1.0 day for irDiabetes, 1.0 day for irAdrenalitis, 1.9 days for irHypophysitis, and 2.4 days for irThyroiditis. The longest treatment response was observed in ophthalmic toxicities that were treated with local corticosteroids with a mean of 8.0 days, and in cutaneous toxicities which were treated with topical corticosteroids with a mean of 9.5 days ( Table 2).

Incidence of sr-/sd-irAEs
Among the patients that were treated with systemic steroids as a first-line therapy for irAEs, the highest rate of sr-/sd-irAEs of 77.8% (n = 7) was documented in patients with irArthritis, followed by 75.0% in patients with dermatologic toxicities (n = 3), which comprised a lichenoid rash and bullous drug reaction, whereas maculopapular exanthema was well managed with topical steroids. The most frequent irAEs which were treated with systemic steroids in our cohort were steroid-refractory or -dependent irPneumonitis in 40.0% (n = 4), irColitis in 34.9% (n = 15), and irHepatitis in 26.8% (n = 11, Table 3).

Second-Line Management
In 25 patients (10.9%), second-line immunosuppression was initiated after the insufficient efficacy of steroid therapy. Infliximab was used most frequently (48% of patients, n = 12), and the indication was for irColitis. Patients with irHepatitis were treated with mycophenolate mofetil, representing 28% of second-line therapies (n = 7). Two patients (8%) with irArthritis were treated with methotrexate. In cutaneous toxicities, a lichenoid rash was treated with acitretin (n = 1) and extracorporeal photopheresis (n = 1), and bullous drug reactions or immune-related bullous pemphigoid (irBP, n = 1) was treated with intravenous immunoglobulins. One patient with irMyocarditis was treated with intravenous immunoglobulins in addition to high-dose corticosteroids.
The time to response was shortest in patients treated with infliximab for irColitis with a mean time of 1.3 days, and in patients with irHepatitis, who were treated with mycophenolate mofetil with a mean time to improvement of 2.8 days. Other second linetherapies took longer to induce an improvement in symptoms, with latencies ranging from 2 weeks to 3 months (Table 4).
In three patients, a subsequent therapy after non-responsiveness to second-line management was initiated: (1) a patient with grade 4 irColitis, who did not respond to two doses of infliximab at all, was treated with 3 doses of the α 4 β 7 integrin blocker vedolizumab and high doses glucocorticoids until a first reduction in stool frequencies and pain was achieved.
(2) A second patient with grade 4 irColitis, who also did not respond to two doses of infliximab, was treated with a combination of extracorporeal photopheresis and two doses of vedolizumab. Here, a first significant improvement was documented after 6 weeks.
(3) A patient who suffered from biopsy-proven grade 4 irHepatitis and who did not show any improvement after the initiation of mycophenolate mofetil 2 mg/kg was switched to the calcineurin inhibitor tacrolimus (10 ng/mL whole blood trough concentration) and eventually to sirolimus (10 ng/mL whole blood trough concentration) when the first significant reduction in liver enzymes occurred after 6 months. Even 5 years after the onset of irHepatitis, it was still not possible to completely taper sirolimus and prednisolone. Complete resolution (n = 9) Improvement, but not resolution (n = 1) Third-line therapy (n = 2) irHepatitis (n = 7) Mycophenolate-mofetil 1g/day (n = 2) 2g/day (n = 5)

Survival
To investigate the influence of irAE occurrence and side effect management on survival, we compared the OSs of patients who developed irAEs (n = 125) with those of patients without irAEs (n = 117). For better comparability, only patients with metastatic cutaneous melanoma (n = 242) were taken into account.
Although not statistically significant, patients without irAEs had a longer mean OS of 90.0 months (95% CI 76.2 to 103.8) compared to 69.3 months (95% CI 59.5 to 79.1) in patients who developed irAEs (p = 0.780).
Regarding the management of severe irAEs (Grade 3 and 4), patients who were treated with a second-line immunosuppressant (n = 16) had a longer mean OS of 56.2 months (95% CI 40.5 to 71.9) compared to 45.6 months (95% CI 34.4-56.8) in patients who were only treated with corticosteroids (n = 43, n = 0.321, Figure 3).

Comparison with Guidelines
Second-line treatment of irColitis with infliximab, irHepatitis with mycophenolate mofetil, and irArthritis with methotrexate was in accordance with all three guidelines. Intravenous immunoglobulins for the therapy of myocarditis are covered along with antithymocyte globulin, infliximab, and mycophenolate mofetil by NCCN guidelines, whereas ASCO guidelines recommend therapy with mycophenolate mofetil, infliximab or antithymocyte globulin, and ESMO guidelines recommend tocilizumab, mycophenolate mofetil, antithymocyte globulin, alemtuzumab, or abatacept. Recommendations for steroid-refractory or steroid-dependent skin toxicities are given by ASCO and ESMO guidelines: rituximab in bullous dermatoses and intravenous immunoglobulins or cyclosporine for severe cutaneous dermatoses (ASCO guidelines), and infliximab or tocilizumab for a maculopapular rash (ESMO guidelines). Guidance on the lichenoid rash is not included. Acitretin and extracorporeal photopheresis for the treatment of lichenoid rash and intravenous immunoglobulins for the treatment of bullous dermatoses were not contained in any guideline. For second-line management in our patient cohort, there was 84% (n = 21) accordance with ASCO guidelines, 88% (n = 22) accordance with NCCN guidelines, and 84% (n = 21) accordance with ESMO guidelines (Table 5).

Discussion
Our study shows a remarkable incidence of 6.2% of patients with steroid-refractory and steroid-dependent side effects under ICI therapy in a large cohort of skin cancer patients. In a total of 406 patients, 10.9% of all documented immune-related side effects did not sufficiently respond to steroids or became exacerbated when trying to taper the steroid therapy.
Compared to the data of Luo et al., who detected 51 patients (1.9%) with sd/sr irAEs among 2750 patients with lung cancer, the frequency in our patient cohort was more than three-fold higher [25]. Whereas the incidence of irColitis was comparable in both groups (48% vs. 53%), the frequency for hepatitis was lower in the cohort of Luo et al. (28% versus 12%), which could be due to the differences in monitoring transaminases or the lower percentage of patients treated with combined ipilimumab and nivolumab. Our cohort had 0% sd/sr-pneumonitis compared to 20% in the lung cancer group. The severity of pneumonitis is known to be higher in lung cancer patients compared to other cancer types, as risk factors such as smoking, previous lung disease, and prior radiotherapy or tyrosine kinase inhibitor therapy are more frequent in this population [26]. Importantly, up to 52% of checkpoint inhibitor-induced irAEs or their sequelae persist after the cessation of ICI therapy [27,28]. Thus, this special toxicity profile has to be taken into account when deciding between immunotherapy or targeted therapy in which side effects are reversible. Currently, the benefit of ICI outweighs toxicity. However, if severe toxicity could be predicted, this would be extremely valuable, especially when treating patients in early disease stages where the absolute risk reduction is smaller [29]. This approach is not yet successful, however. Thus, current efforts are aimed at continuing to reduce exposure to ICIs by combining them with the mRNA vaccine [30].
The higher rate of sr/sd-irAEs in this study can be explained by the high frequency of combined immunotherapy with ipilimumab and nivolumab (43% compared to 37%), known to induce >grade 3 toxicity in 59% [6]. Regarding the outcome of first-line therapies for irAEs, overall, 13.1% of irAE symptoms improved without the complete resolution of the irAEs, and in 2.6%, no improvement was induced. Retrospectively, we estimate that a second-line immunosuppressive treatment should have been started earlier in those patients and more frequently since prompt treatment has been shown to improve the response [31].
In our cohort, patients with metastatic cutaneous melanoma treated with corticosteroids plus second-line immunosuppressants showed a trend towards a longer OS compared to patients treated with corticosteroids only for grade 3/4 irAEs. This is in accordance with data from another retrospective multicenter cohort study which included 771 patients with metastatic melanoma [32].
Nowadays, second-line immunosuppression is indicated early in steroid-refractory irAEs. For instance, according to the NCCN guidelines, if no improvement is noted within 2-3 days of intravenous methylprednisolone at a dose of 1-2 mg/kg per day, the addition of infliximab or the α 4 β 7 integrin blocker vedolizumab is strongly recommended. Similarly, additional treatment with mycophenolate mofetil is recommended for steroid-refractory hepatitis after 3 days of initial treatment [8]. Patients in our study responded within a mean time of 1.3 days after the initiation of infliximab for irColitis, and 2.8 days after the initiation of mycophenolate mofetil for irHepatitis. On the contrary, the time to first response ranged from 2 weeks to 3 months for other steroid-refractory irAEs.
The limitations of our study include its retrospective nature that potentially underestimates the frequency of irAEs. Furthermore, monocentric data may not be applicable to multicentric cohorts of patients.
Treatment recommendations have been established by the ASCO, NCCN, and ESMO. However, these guidelines are not based on studies comparing different treatment approaches. In our patients, accordance with the recommendations in the published guidelines ranged between 76% and 88%. Management data after the failure of first-line treatment for some irAEs, such as lichenoid rash and bullous drug reactions in our population, are lacking.

Conclusions
This study assessed the frequency of steroid-refractory or steroid-dependent irAEs in patients with cutaneous malignancies who were treated with ICIs. Prospective trials with larger patient populations or analyses of combined clinical study data are needed to gain a better understanding of rare steroid-refractory or steroid-dependent toxicities.  Informed Consent Statement: Patient consent was waived due to the approval of the Ethics Committee at the Medical Faculty of LMU Munich as the anonymized collection and analysis of patient data. Data Availability Statement: All data generated or analyzed during this study are included in this article. Further enquiries can be directed to the corresponding author upon reasonable request.