Health-related quality of life in patients with advanced melanoma treated with ipilimumab: prognostic implications and changes during treatment

Background We have previously reported that the safety and efficacy of ipilimumab in real-world patients with metastatic melanoma were comparable to clinical trials. Few studies have explored health-related quality of life (HRQL) in real-world populations receiving checkpoint inhibitors. This study reports HRQL in real-world patients receiving ipilimumab and assesses the prognostic value of patient-reported outcome measures. Patients and methods Ipi4 (NCT02068196) was a prospective, multicentre, interventional phase IV trial. Real-world patients (N = 151) with metastatic melanoma were treated with ipilimumab 3 mg/kg intravenously as labelled. HRQL was assessed by the European Organisation of Research and Treatment of Cancer Quality of Life Questionnaire at baseline and after 10-12 weeks. Results The European Organisation of Research and Treatment of Cancer Quality of Life Questionnaire was completed by 93% (141/151 patients) at baseline, and by 82% at 10-12 weeks. Poor performance status and elevated C-reactive protein (CRP) were associated with worse baseline HRQL. Clinically relevant and statistically significant deteriorations in HRQL from baseline to weeks 10-12 were reported (P <0.05). Baseline physical functioning [hazard ratio (HR) 1.96, P = 0.016], role functioning (HR 2.15, P <0.001), fatigue (HR 1.60, P = 0.030), and appetite loss (HR 1.76, P = 0.012) were associated with poorer overall survival independent of performance status, lactate dehydrogenase (LDH), and CRP. We further developed a prognostic model, combining HRQL outcomes with performance status, LDH, and CRP. This model identified three groups with large and statistically significant differences in survival. Conclusions Systemic inflammation is associated with impaired HRQL. During treatment with ipilimumab, HRQL deteriorated significantly. Combining HRQL outcomes with objective risk factors provided additional prognostic information that may aid clinical decision making.


INTRODUCTION
The last decade has marked a therapeutic paradigm shift for patients with metastatic melanoma. Ipilimumab, an antibody targeting the cytotoxic T-lymphocyte-associated protein 4, was the first treatment to show improved overall survival (OS) in metastatic melanoma in a randomised phase III trial, 1 followed by programmed cell death protein 1 (PD-1) inhibitors with or without ipilimumab, and BRAF inhibitors with or without MEK inhibitors. 2 Despite therapeutic advances, life expectancy for patients with metastatic melanoma is still restricted. 2 Thus, other treatment goals, such as preserving quality of life, are of great importance, and may affect benefiterisk assessment.
Health-related quality of life (HRQL) is defined by the Center for Disease Control and Prevention as 'an individual's or group's perceived physical and mental health over time', 3 indicating a subjective, multidimensional evaluation of own well-being as opposed to a physician's external, usually more disease-focused perspective. Factors shown to influence HRQL in cancer include weight loss, performance status, and systemic inflammation. 4 Different tools for evaluating HRQL in patients with cancer exist, including more generic and more disease-specific questionnaires. 5,6 Such patient-reported outcome measures (PROMs) have increasingly become an integrated part of clinical trials, acknowledging patients' voice and recognising the valuable contributions of such measures. 7 Few studies have explored HRQL in real-world populations receiving checkpoint inhibitors. Key HRQL issues in metastatic melanoma comprise pain; insomnia; fatigue; appetite loss; itching; nausea and vomiting; postsurgical symptoms; emotional distress; and restrictions in physical, role, and social functioning. [8][9][10][11] Reports from chemotherapy and interferon trials suggest worsening of HRQL during treatment. 10 The Ipi4 trial was a prospective, phase IV trial providing ipilimumab to real-world patients with advanced melanoma. Treatment-associated high-grade toxicity was observed in 28% of patients, and immune-related adverse events (irAEs) in 56%. 12 The median progression-free survival (PFS) was 2.7 months [95% confidence interval (CI) 2.6-2.8 months] and OS 12.1 months (95% CI 8.3-15.9 months), comparable to phase III trials. Poor performance status, elevated lactate dehydrogenase (LDH), and C-reactive protein (CRP) at baseline were independently associated with short survival. In this study, we aimed to investigate HRQL in patients with metastatic melanoma receiving ipilimumab. Further, we assessed how HRQL relates to laboratory and clinical markers, and if PROMs may be combined with such objective markers in a prognostic score.

Patients and study design
The Ipi4 trial was a prospective, national, multicentre, single-armed, interventional phase IV trial (NCT02068196). Adult patients with histologically confirmed inoperable metastatic melanoma were included and received ipilimumab 3 mg/kg intravenously every 3 weeks for up to four doses. Key eligibility criteria included Eastern Cooperative Oncology Group performance status (ECOG PS) 1, no active brain metastases, autoimmune disease, or immunodeficiency. Any previous treatment was allowed. All patients provided written informed consent. The Ipi4 trial was approved by the Regional Committee for Medical and Health Research Ethics South East Norway and conducted in accordance with the ethical principles of the Declaration of Helsinki (1964).

Study assessments
Toxicity was assessed by the Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. Tumour response was evaluated by computed tomography using RECIST 1.1. Patients self-reported HRQL using the European Organisation of Research and Treatment of Cancer Quality of Life Questionnaire (EORTC-QLQ-C30) version 3.0. 5 Questionnaires were completed before each visit at baseline; treatment weeks 4, 7, 10, and 12; and follow-up before progression. The EORTC-QLQ-C30 constitutes 30 questions that form subscales for global health (1), functioning (5), and symptoms (9, including 6 single measures and 3 scales) (13). A high score for global health or functioning scales represents high HRQL, while high scores for symptoms scales indicates more symptoms. Scores were converted according to the scoring manual. 13 Based on the literature, deteriorations in global health, functional scales, fatigue, nausea and vomiting, pain, insomnia, and appetite loss were expected. Thus we decided to focus on these scales. Diarrhoea, being one of the most frequent and serious side-effects of ipilimumab, was also included in our analyses.

Statistics
Statistical analyses were carried out using IBM SPSS Statistics for Windows, version 26 (IBM Corp., Armonk, NY, USA). Mean baseline HRQL was reported with 95% CIs and standard deviations. Mean scores in population subgroups were compared with one-way analysis of variance. The association between HRQL outcomes and baseline characteristics was tested using Pearson's chi-square test, or the chi-square test for linear trend as appropriate.
Changes in HRQL during treatment were estimated by the mean difference between outcomes at baseline and 10-12 weeks. The week 12 assessment was preferred if both weeks 10 and 12 were available. A difference of 10 points in the scale of 0-100 was considered clinically meaningful, 14 and P <0.05 was considered statistically significant. In addition, a mixed linear model including all available patient-reported outcomes from all patients (N ¼ 144) at baseline to week 12 was used to investigate the robustness of the findings in the complete case analysis, constituted by the 102 patients who completed questionnaires at both baseline and weeks 10-12. Survival was estimated by KaplaneMeier analyses. OS was defined as time from treatment initiation to death, and PFS as time from treatment initiation to objective tumour progression or death. Patients were followed for at least 5 years or until death. Patients without an event were treated as censored 1 June 2020.
The effect of HRQL variables on OS was analysed using univariable and multivariable Cox proportional hazard modelling, reported as hazard ratios (HRs) with 95% CI. A cut-off at 66.7 was used for global health (average score 5/7). The average score of 'a little' (2/4) was used as cut-off for functioning scales (66.7) and for symptoms scales (33.3). A stepwise approach was applied to identify HRQL variables that were independently associated with OS in a Cox model including baseline ECOG PS, LDH, and CRP as covariates, factors previously identified as independent predictors of OS in the Ipi4 trial. 12 HRQL variables with P <0.05 in unadjusted analyses were tested one-by-one in adjusted Cox models. Variables included in the final Cox model were checked for multicollinearity by the variance inflation factor and accepted if <3.

Baseline health-related quality of life
Baseline HRQL is outlined in

ESMO Open
E. Aamdal et al.   Treatment-naïve patients more commonly suffered deteriorations in global health (68% versus 41%) and fatigue (54% versus 30%) than pretreated patients. Changes in HRQL seemed independent of sex, metastatic stage, BRAF V600 mutation status, and CRP. No statistically significant association between subgroups and increase in diarrhoea was observed.

Prognostic model combining biological factors and patientreported outcome measures
As PROMs offer prognostic information, we developed a model combining HRQL outcomes with objective markers independently predicting a worse OS in this population ( Figure 2E). This model comprised ECOG PS, LDH, CRP, physical functioning, role functioning, fatigue, and appetite loss. Three distinctive prognostic groups were identified. The risk of death was significantly increased in patients with five to seven risk factors (group C; HR 5.81, P < 0.001) and two to four risk factors (group B; HR 1.72, P ¼ 0.013), compared with patients with zero or one risk factor (group A). Large differences in median survival was observed between the groups, with an eightfold advantage for group A compared with group C (26.0 months versus 3.4 months). No patients in group C survived >27 months. A prognostic model for PFS identified a group of patients with significantly poorer PFS, and thus, supported our findings for OS (Supplementary Figure S2,

DISCUSSION
This study reported on HRQL in real-world patients with metastatic melanoma receiving ipilimumab. Baseline HRQL was comparable to the general population. 15 The relatively high baseline HRQL in this study may partly be related to the hope of starting a new, promising treatment. 16 Moreover, the psychological phenomenon known as response shift, causing patients' conceptions and expectations of own HRQL to readjust during the disease course, 17 may have contributed.
Baseline HRQL was generally poorer in patients only replying to the first questionnaire. These patients were more frequently reported with ECOG PS 1, LDH >ULN, and CRP 10 mg/l, factors associated with poorer OS in this trial. 12 Previously, poor health has been recognised as a likely cause of patients not replying to questionnaires. 18 This is a common issue in HRQL research and challenges generalisability of the results.
ECOG PS 1 was associated with clinically and statistically significantly worse mean scores for global health, functioning scores, fatigue, pain, and appetite loss. As ECOG PS is the physician's grading of a patient's ability to carry out daily activities, this finding is expected. ECOG PS is widely recognised to determine HRQL. 4 Patients with BRAF V600 mutations reported significantly more pain and poorer role functioning at baseline, but no significant difference in global health. Impaired HRQL may be related to rapid tumour growth driven by the BRAF mutation, and consequently, a poorer prognosis, in line with results from a meta-analysis. 19 A phase III trial, comparing treatment with BRAF inhibitors versus combined treatment with BRAF and MEK inhibitors in patients with BRAFmutated metastatic melanoma, did indeed find that pain was a major HRQL issue in this population. 20 4 months), patients in the group B lived for a median of 9.0 months (95% CI 6.9-11.1 months), and patients in group C lived for a median of 3.4 months (95% CI 0.6-6.2 months). CI, confidence interval; HR, hazard ratio; ULN, upper limit of normal.
been associated with generally decreased HRQL. 22 Increased inflammation, indicated by CRP 10 mg/l, was associated with clinically and statistically significantly poorer HRQL including global health, role functioning, social functioning, fatigue, pain, and appetite loss. The influence of systemic inflammation on HRQL and cancer-related symptoms including pain, anorexia, and fatigue has been previously recognised in advanced cancer. 4,23,24 Moreover, CRP 10 mg/l was associated with a worse OS in this trial. 12 Thus, in this patient population, even a mildly elevated CRP negatively affected the present by impairing HRQL, and the future by shortening OS.
We have previously reported that efficacy and toxicity of ipilimumab in the real-world setting were comparable to phase III trials. 12 This also applies to the observed changes in HRQL. Clinically meaningful worsening in HRQL in patients receiving ipilimumab monotherapy was detected in the MDX010-20 25 and KEYNOTE-006 26 trials, with similar findings reported in an observational study in real-world patients with metastatic melanoma. 27 The CheckMate 067 trial, 28 however, noted no clinically meaningful changes in the ipilimumab arm, but compliance was <70% at treatment completion.
Contradicting previous findings, 25 we did not detect that age >65 years predisposed to deterioration in HRQL. Therefore we did not find support that HRQL justifies withholding ipilimumab based on age. In this dataset, pretreated patients and patients with ECOG PS 1 were less likely to experience impairments in global health during treatment. This may simply be due to their baseline levels being poorer. Alternatively, the patients with a higher disease burden experienced more symptomatic relief from treatment. However, this interpretation is not consistent with findings for LDH. Moreover, these observations may be incidental due to multiple testing and need to be validated in an independent patient cohort. Impaired HRQL has previously been associated with progression, 29 and may well be associated with increasing disease-related symptoms, and/ or the psychological burden of disease progression, and therapeutic failure in this study. Importantly, due to lack of a control group, this study does not address whether HRQL deteriorations were treatment related or related to disease progression or other causes. In the MDX010-20 trial, no clinically significant HRQL differences were observed between ipilimumab and the gp100 vaccine. 25 Previous studies of HRQL in patients receiving ipilimumab have suggested distress peaking 10-12 weeks after treatment initiation, 25,26,30 the timing of the second questionnaire in our study. As indicated by a median PFS of 2.7 months, 12 half of patients had progressed by this point. Further, most toxicities were encountered by 10-12 weeks, supporting the timing of the HRQL assessment.
No significant association between high-grade toxicity or irAEs and deteriorations in global health was observed, in line with a phase III trial randomising adjuvant ipilimumab in stage III melanoma against placebo. 30 Notably, the mean reported change in diarrhoea during treatment was not clinically meaningful. Only 23% of patients reported a worsening at weeks 10-12, but investigators registered diarrhoea and colitis as adverse events in 30% of patients. 12 Possible explanations may be transient symptoms, or selective reporting by patients not believing diarrhoea was related.
We evaluated HRQL using the EORTC-QLQ-C30, the recommended assessment in advanced skin cancers. 31 This questionnaire is nonspecific to melanoma. With diarrhoea being the obvious exception, the EORTC-QLQ-C30 does not specifically address irAEs. irAEs may affect any organ in the body and are often managed by immunosuppression. 32 A qualitative study recognised itching, rashes, muscle stiffness, cramping, fever, and chills, as well as a protracted feeling of having a cold, to be associated with HRQL in patients receiving immune checkpoint inhibitors. 16 In the Ipi4 trial, pruritus was recorded in 11% of patients, rash in 25%, musculoskeletal symptoms in 2%, fever in 4%, and flulike symptoms in 2%. 12 Accordingly, HRQL issues may have been missed due to methodological shortcomings.
As compliance was low during follow-up, this HRQL report is limited to 10-12 weeks. Although most irAEs are reversible, endocrinopathies are usually persistent, potentially affecting HRQL long term. Further, a substantial proportion of patients will require immunosuppression. Yet, a recent report on HRQL in long-term survivors following ipilimumab observed no difference in global health despite lower functioning scores and higher symptoms scores compared with healthy controls. 33 In a benefit-risk assessment, a treatment's impact on HRQL may alter the balance. Although >50% of patients in the current trial reported maintained or less symptoms, >50% of patients experienced clinically meaningful deteriorations in global health, stressing the limitations in evaluating means in this context. Overall, however, we did not find support that ipilimumab improved HRQL during treatment in this population.
HRQL has previously been identified as prognostic in patients with metastatic melanoma. 34 In our study, physical and role functioning, fatigue, and appetite loss were associated with OS. Combining these PROMs with previously identified risk factors was more accurately associated with OS than biological factors, or PROMs alone. To our knowledge, this is the first proposed prognostic model comprising both PROMs and biological factors in melanoma and needs validation in other populations to verify its use. For clinical decision making, it may be particularly useful to identify patients that will not achieve long-term survival, and where antitumour therapy may not be in their best interest. It would therefore be important to establish if an entity resembling group C in our dataset, with a poor median survival and no long-term survivors, can be identified. The strategy of combining PROMs with objective markers may be applicable also to other cancer forms. Importantly, PROMs offer complementary information, as compared with laboratory markers and outcomes reported by the doctor.
Despite a high compliance for the baseline questionnaire, our study has some limitations. As expected, the compliance for the second questionnaire was lower due to some patients experiencing rapid progression. Unfortunately, patients who progressed before evaluation did by protocol not reply to the second questionnaire. Therefore, this study contributes little information on the HRQL in these patients. Multiple testing of patient characteristics and HRQL indicators with respect to clinical outcomes may have provided incidental findings. Further, observations should be validated in independent patient cohorts.
Currently, ipilimumab is most frequently used in combination with nivolumab, and our findings may not be directly applicable to this clinical practice. However, ipilimumab is also administered as monotherapy after progression on PD-1 inhibitors. In our study, one in three patients had received prior therapy, although not PD-1 inhibitors. In reports on ipilimumab after progression on PD-1 inhibitors, other clinical outcomes were largely in line with ipilimumab in the first-line setting. 35,36 Thus, our study may provide useful information on HRQL during treatment with ipilimumab in the real-world setting.

Conclusions
In summary, increased CRP was associated with poorer HRQL. Concurring with progression for many patients, clinically meaningful and statistically significant deteriorations in HRQL were observed during treatment, supporting findings from clinical trials. Baseline HRQL was independently associated with survival and may in combination with biomarkers be valuable in prognostication, emphasising the importance of patientedoctor communication in clinical decision making.