Real-world experience with pembrolizumab toxicities in advanced melanoma patients: a single-center experience in the UK

Aim: We aimed to characterize the safety profile of pembrolizumab in advanced melanoma patients at our center to better reflect ‘real-world’ data on anti-PD-1 inhibitors. Materials & methods: At our institution, 58 ipilimumab-naive and 30 ipilimumab-treated patients with advanced melanoma who have received pembrolizumab between June 2014 and June 2017 were included for analysis. Results: Incidence of any-grade and grade 3/4 toxicities were 81.8% (n = 72) and 12.5% (n = 11), respectively. The most common side effects were skin-related (61.4%, n = 54) and gastrointestinal-related (51.1%, n = 45) events. In total, 25% of patients required oral steroids to manage immune-related adverse events with a median cumulative prednisolone dose of 683 mg (range: 40–3745 mg). Conclusion: Pembrolizumab is well tolerated in ‘real-world’ patients and severe toxicities can be effectively managed with systemic steroids.

Although the range of irAEs is quite similar across checkpoint inhibitors, there are differences in frequency of toxicities. Grade 3/4 toxicities are less commonly observed with PD-1 inhibitors compared with CTLA-4 inhibitors [8]. Furthermore, PD-1 inhibitors have higher rates of thyroiditis and pneumonitis compared with CTLA-4 inhibitors [8]. Other factors that may influence toxicity profiles include combination therapies, treatment sequencing, tumor type and preexisting autoimmune conditions such as rheumatoid arthritis and inflammatory bowel disease [9,10].
Pembrolizumab is generally well tolerated but severe irAEs can lead to significant patient morbidity. Majority of grade 3/4 irAEs can be managed with oral (and intravenous) steroids. In steroid-refractory cases, other immunomodulatory drugs may be used including infliximab, mycophenolate mofetil and calcineurin inhibitors [11].
It is a common concern among clinicians that the incidence of AEs in clinical trials do not necessarily reflect 'real-world' experiences due to underrepresentation of the patient population [12]. While early identification of AEs is dependent on both the clinician and self-reporting by patients, underrepresented data can hinder effective clinical practice. Moreover, published real-life data on irAEs with pembrolizumab are limited. In the UK, the PD-1 inhibitors, pembrolizumab and nivolumab, are funded by the NHS for use in melanoma. The majority of patients in our institution receive pembrolizumab rather than nivolumab due to its less intense 3-weekly schedule (2-weekly for nivolumab). In this study, we analyzed the incidence and management of pembrolizumab-associated irAEs in our hospital to guide medical decision-making and health information provision for patients.

Patient sample
We retrospectively analyzed the data on all patients with advanced melanoma who received pembrolizumab (2 mg/kg every 3 weeks) at a single UK center between 15th June 2014 and 15th June 2017. Inclusion criteria included adequate documentation of the treatment course and histologically confirmed advanced cutaneous melanoma. Patients who received treatment as part of an expanded patient access program were also included (n = 5). Patient-specific information was collected using our hospital's electronic databases. Data collected included patient demographics, tumor characteristics (Tumor, Node, Metastasis (TNM) staging, brain metastasis, baseline LDH, BRAF V600 status), previous systemic therapies, date of first immunotherapy dose, date of death (or the last followup), baseline WHO performance status (PS), number of treatment doses and deferrals, best overall response and AEs. Data were also collected on the management of irAEs including cumulative steroid dose.

Data analysis
AEs were graded using the Common Terminology Criteria for Adverse Events v4.0. Select AEs of interest were categorized into organ-specific groups for analysis. Of note, 'general' AEs encompass nonspecific symptoms including fatigue/lethargy, asthenia, pyrexia, and decreased appetite. Due to the limitations of retrospective grading of toxicities, diarrhea and colitis were grouped together. AEs were grouped into grades 1/2 and 3/4 due to inherent difficulties in retrospective grading from some clinical notes.
Estimated total steroid dose per patient was calculated based on documented prescriptions and steroid-tapering guidelines. All steroids were converted to the equivalent prednisolone dose for ease of comparison. Any steroids prescribed for hormone-replacement therapy in adrenal insufficiency and for non-irAEs, such as brain metastasis and radiation-induced pneumonitis, were excluded from analysis. Time to first grade 3/4 AE was defined as time from first treatment dose to time of first reported AE. Patients who did not experience grade 3/4 AEs were censored at time of last follow-up. OS was defined as time from first dose of pembrolizumab to death of any cause. Patients who were still alive at data cutoff were censored at time of the last follow-up. Median follow-up time was defined as time from first treatment dose to time of the last follow-up. Patients who have died prior to data cutoff were censored at their date of death. Best overall response was identified based on documented responses, and was not formally assessed retrospectively using the immune-related Response Evaluation Criteria In Solid Tumors (ir-RECIST). The TNM staging system used was version 7 of the American Joint Committee on Cancer staging (AJCC). Kaplan-Meier survival curves were produced using GraphPad Prism 7. 95% CI were calculated using the Anderson's method (1993). We compared the incidence of AEs in the study groups using the Chi-square test, where p = 0.05 was considered statistically significant.

Patient demographics
A total of 88 patients with melanoma were identified as suitable for inclusion and analysis. In total, 58 patients received pembrolizumab as first-line immunotherapy (ipilimumab-naive group) and 30 patients received pembrolizumab following ipilimumab therapy (ipilimumab-treated group). Demographics are shown in

Adverse events
The incidence of any-grade and grade 3/4 AEs were 77.6 and 10.3% in ipilimumab-naive patients, and 90.0 and 16.7% in ipilimumab-treated patients ( Table 2). No treatment-related deaths were observed. The most common any-grade AEs in ipilimumab-naive and ipilimumab-treated patients were fatigue/lethargy (34. 5   from a skin lesion and required oral steroids. All three patients were incidentally diagnosed on follow-up CT to assess for treatment response and were shown to have new bilateral hilar lymphadenopathy and mediastinal lymphadenopathy. All patients had raised ACE levels and endobronchial ultrasound-guided biopsy was performed on all patients to rule out new metastatic lesions and confirmed the histological diagnosis of sarcoidosis. Of the total 88 patients, 15 patients (17.0%) were diagnosed with hypothyroidism during treatment, of which, eight patients (53.3%) were preceded with biochemical hyperthyroidism. Only one of these eight patients was symptomatic, requiring treatment with β-blockers. Furthermore, two ipilimumab-treated patients required longterm steroid replacement due to primary and secondary adrenal insufficiency.
Time to adverse event Median times to any grade AE were 1.5 months (95% CI: 0.7-2.3) and 2.0 months (95% CI: 1.5-2.5) in ipilimumab-naive and ipilimumab-treated patients, respectively ( Figure 1A & C). The majority of grade 3/4 AEs occurred within 4 months of starting pembrolizumab. 43.1 and 36.7% of ipilimumab-naive and ipilimumab-treated patients, respectively, experienced their first any-grade AE after one cycle of treatment. Gastrointestinal-related toxicities were the most common AE postcycle 1 in ipilimumab-naive (28.9%) and ipilimumab-treated (42.9%) patients. The majority of new AEs occurred between cycles 1 and 4 in all cohorts with the frequency of new events gradually decreasing as treatment time progressed. In the ipilimumab-treated group, there was a rise in the number of new skin-related toxicities from cycle 2 onward. It should be noted that a number of side effects did not occur until after ten cycles of treatment.   All patients on systemic treatment for irAEs experienced symptomatic relief. The median total duration and cumulative dose of systemic steroids used to manage irAEs were 31.5 days (range: 1-259 days) and 683 mg (range: 40-3745 mg), respectively. One patient developed diabetes due to prolonged exposure to high-dose steroids (estimated cumulative dose of prednisolone = 630 mg; Table 3). There were two patients who were unable to wean off steroids at data cutoff due to ongoing symptoms from immune-related arthritis.

Treatment response
The median numbers of treatment cycles administered to ipilimumab-naive and ipilimumab-treated patients were six (range: 2-26) and 10 (range: 2-38), respectively; the proportions of patients continuing treatment at the time of data cutoff were 58.6 and 20.0%, respectively. Three patients permanently discontinued their treatment due to immune-related skin rash (n = 1), pneumonitis (n = 1) and sarcoidosis (n = 1). The proportions of ipilimumab-naive and ipilimumab-treated patients requiring treatment deferral were 29.3 and 43.3%, respectively, of which 36.4 and 35.0% were related to treatment toxicity, respectively. Patient choice, such as vacation plans, was also a major contributing factor to treatment deferral.
Having a preexisting autoimmune condition, receiving systemic steroids for irAEs, or having a BRAF-mutant positive status, did not appear to significantly affect response to pembrolizumab. A comparison of median OS in patients who had one of these factors to patients who did not was not possible, as the median OS was not reached in the majority of the subpopulations. Median OS of ipilimumab-treated patients who received systemic steroids for irAEs was lower compared with those who did not receive steroids (5.75 vs 24.0 months, respectively). However, we recommend that these comparisons should be interpreted with caution due to our small sample size.

Discussion
The success of immunotherapies has led to pembrolizumab being established as a successful treatment option for advanced melanoma. However, there are currently limited published data on 'real-world' experiences with pembrolizumab in melanoma patients, particularly the incidence and management of irAEs. The primary objective of this study was to characterize our center's experience of pembrolizumab use. The KEYNOTE-002 and KEYNOTE-006 clinical trials were used as comparators as they demonstrated superiority of pembrolizumab against the previous standard of care in ipilimumab-treated and ipilimumab-naive patients, respectively [3,4]. Patient baseline characteristics in both cohorts were comparable to pivotal studies with regard to age, gender, and PS. There was a higher incidence of brain metastases in our series than the clinical trials, which is to be expected in a real-world population. Interestingly, our cohorts had less M1c disease at baseline and were less heavily pretreated compared with trial participants. ORRs in ipilimumab-naive and ipilimumab-treated patients were 48.3% (95% CI: 40.1-56.5) and 50.0% (95% CI: 37.3-62.7) compared with 32.9% (95% CI: 27.4-38.7) and 21.1% (95% CI: 15-28) in KEYNOTE-006 and KEYNOTE-002, respectively [3,4]. Although ORRs were more favorable in our cohort, this could be a result of not formally assessing treatment response using the (ir)RECIST criteria in clinical practice, thus reflecting one of the limitations with retrospective studies. One-year estimates of survival were similar between our ipilimumab-naive cohort (64.9%; 95% CI: 47-1-78.1) and KEYNOTE-006 (68.4%; 95% CI: 62.5-73.6) [4]. However, median OS was superior in our ipilimumab-treated cohort (23.5 months; 95% CI: 13.9-33.1) compared with KEYNOTE-002 (13.4 months; 95% CI: 11.0-16.4), which could reflect the better prognostic features of the patients in our series [3].
In the Phase II randomized CheckMate-064 trial comparing the safety of sequential administration of nivolumab and ipilimumab, higher incidences of grade 3/4 irAEs were observed in the second induction period compared with the first induction period in either sequences [13]. This suggests that patients who were previously primed to a class of checkpoint inhibitor (e.g., anti-CTLA-4) are at an increased risk of more toxicities with a different class (e.g., anti-PD-1). This may explain the higher incidence of toxicities in our ipilimumab-treated patients compared with ipilimumab-naive patients.
The frequency of oral steroids administered for any-grade irAEs were 12 (20.7%) and 10 (33.3%) in ipilimumabnaive and ipilimumab-treated patients. This is a significant proportion of our cohort requiring steroids for irAEs and one patient developed steroid-related diabetes. The cumulative dose of steroids received was significant and the long-term outcomes and implications of long-term steroid use in this group of patients have been poorly studied. In the retrospective setting, it was difficult to measure other steroid-related AEs such as proximal weakness and weight gain. The short-term and long-term implications of steroid use in cancer patients have already been extensively reviewed in other papers [14,15]. Some of the severe sequelae of chronic steroid use include an increase in the risk of opportunistic infections, metabolic syndrome, adrenal suppression, osteoporosis and neuropsychiatric disorders. Moreover, prolonged steroid use could theoretically reduce the efficacy of checkpoint inhibitors, although this is still an area of uncertainty [16]. We recommend that there should be more research in this area as patients survive for longer after immunotherapy treatment.
There is little published literature documenting real life experiences with pembrolizumab. There are some small retrospective series that also highlight the increased incidence of irAEs in clinical practice in comparison to clinical trials [17][18][19]. A single-center observational study of 39 ipilimumab-treated patients receiving anti-PD-1 therapy reported any-grade and grade 3/4 irAEs of 48.7 and 15.4%, respectively [20]. Although the incidence of severe irAEs was similar to our results, there was a large difference in any-grade irAEs. One important difference between our results and those in clinical studies is the incidence of thyroid disorders, in particular hypothyroidism. Incidence of hypothyroidism reported in the KEYNOTE-002 and KEYNOTE-006 studies was 5.1 and 8.7%, respectively. In comparison, the incidence of hypothyroidism in our ipilimumab-treated and ipilimumab-naive patients was 13.3 and 19.0%, respectively. This has been similarly reported in a small observational study of ipilimumabtreated patients with a reported incidence of approximately 15% [20]. It is important that thyroid function tests are monitored regularly on pembrolizumab treatment and is important to recognize that this is a permanent side effect.
Early recognition of irAEs is key to management and can potentially reverse it, thus avoiding long-term sequelae [21]. Early identification of irAEs (and non-irAEs) is often dependent on self-reporting by patients. Patient self-reporting of systemic anticancer treatment (SACT) toxicities can occur either in clinician/nurse-led consultations or through a nurse-led 24 h helpline. The UK National Chemotherapy Board has proposed two approaches to promote early identification of SACT toxicities: Empowerment and Proactive monitoring [22]. It is recommended that patients starting on SACT receive information about the treatment and provide appropriate consent. Studies such as ours highlight the need for specific preimmunotherapy information to be given to patients, carers and health professionals involved in the patient's care.
There are inherent limitations in a retrospective study that have to be acknowledged. In particular, accuracy of grading and reporting of irAEs in clinical records and documentation of cumulative steroid exposure. Although these limitations can impact the reliability of the data, our series reflects a more realistic 'real-world' experience with pembrolizumab treatment. We were also able to confirm that patients with preexisting autoimmune conditions can be safely treated with PD-1 inhibitors but need to be counseled about the significant risk of exacerbating their underlying autoimmune condition [10]. We recognize that our median follow-up time of 12 months is relatively short. However, the majority of the irAEs appear early on in the treatment course, from within a few weeks to up to 3 months, and therefore a 12-month follow-up is sufficient.

Conclusion
We observed a higher incidence of irAEs with 'real-world' patients with metastatic melanoma. The need for more 'real-world' data from observational studies is required to guide clinical decision-making and information provision to patients. Furthermore, with the recent positive results for checkpoint inhibitors in the adjuvant setting and the approval for combination ipilimumab/nivolumab in advanced melanoma patients in the UK, there will an increasing number of patients with melanoma-receiving immunotherapy. Immunotherapy treatments are increasingly licensed in other tumor types, such as lung and renal cancer, and vigilance is needed in identifying and treating these side effects. Further work to clarify the long-term effects of immunotherapy toxicity and treatment is required.