Article Text

Download PDFPDF

Extended report
Rheumatic disorders associated with immune checkpoint inhibitors in patients with cancer—clinical aspects and relationship with tumour response: a single-centre prospective cohort study
  1. Marie Kostine1,
  2. Léa Rouxel1,
  3. Thomas Barnetche1,
  4. Rémi Veillon2,
  5. Florent Martin2,
  6. Caroline Dutriaux3,
  7. Léa Dousset3,
  8. Anne Pham-Ledard3,
  9. Sorilla Prey3,
  10. Marie Beylot-Barry3,
  11. Amaury Daste4,
  12. Marine Gross-Goupil4,
  13. Julie Lallier4,
  14. Alain Ravaud4,
  15. Edouard Forcade5,
  16. Bernard Bannwarth1,
  17. Marie-Elise Truchetet1,
  18. Christophe Richez1,
  19. Nadia Mehsen1,
  20. Thierry Schaeverbeke1
  21. and on behalf of the FHU ACRONIM
  1. 1 Rheumatology Department, Centre Hospitalier Universitaire, Bordeaux, Aquitaine, France
  2. 2 Pulmonology Department, Centre Hospitalier Universitaire, Bordeaux, Aquitaine, France
  3. 3 Dermatology Department, Centre Hospitalier Universitaire, Bordeaux, Aquitaine, France
  4. 4 Oncology Department, Centre Hospitalier Universitaire, Bordeaux, Aquitaine, France
  5. 5 Hematology Department, Centre Hospitalier Universitaire, Bordeaux, Aquitaine, France
  1. Correspondence to Dr Marie Kostine, Rheumatology Department, Centre Hospitalier Universitaire de Bordeaux, Hôpital Pellegrin, Place Amélie Raba Léon, Bordeaux 33000, France; marie.kostine{at}


Objectives To evaluate the prevalence and type of rheumatic immune-related adverse events (irAEs) in patients receiving immune checkpoint inhibitors (ICIs), as well as the correlation with tumour response.

Methods This was a single-centre prospective observational study including all cancer patients receiving ICIs. The occurrence of irAEs and tumour response was assessed on a regular basis. Patients who experienced musculoskeletal symptoms were referred to the department of rheumatology for clinical evaluation and management.

Results From September 2015 to May 2017, 524 patients received ICIs and 35 were referred to the department of rheumatology (6.6%). All but one of the rheumatic irAEs occurred with anti-programmed cell death protein 1(PD-1)/PD-1 ligand 1(PD-L1) antibodies, with a median exposure time of 70 days. There were two distinct clinical presentations: (1) inflammatory arthritis (3.8%) mimicking either rheumatoid arthritis (n=7), polymyalgia rheumatica (n=11) or psoriatic arthritis (n=2) and (2) non-inflammatory musculoskeletal conditions (2.8%; n=15). One patient with rheumatoid arthritis was anti-cyclic citrullinated peptide (anti-CCP) positive. Nineteen patients required glucocorticoids, and methotrexate was started in two patients. Non-inflammatory disorders were managed with non-steroidal anti-inflammatory drugs, analgesics and/or physiotherapy. ICI treatment was pursued in all but one patient. Patients with rheumatic irAEs had a higher tumour response rate compared with patients without irAEs (85.7% vs 35.3%; P<0.0001).

Conclusion Since ICIs are used with increasing frequency, knowledge of rheumatic irAEs and their management is of major interest. All patients were responsive either to low-to-moderate doses of prednisone or symptomatic therapies and did not require ICI discontinuation. Furthermore, tumour response was significantly higher in patients who experienced rheumatic irAEs.

  • arthritis
  • polymyalgia rheumatica
  • psoriatic arthritis
  • inflammation
  • treatment

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Immune checkpoint inhibitors (ICIs) targeting the cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) or programmed cell death protein 1 (PD-1) axes have accelerated the immunotherapy revolution in oncology. Since the approval of ipilimumab (anti-CTLA-4) for the treatment of metastatic melanoma in 2011,1 pembrolizumab and nivolumab (anti-PD-1) have also demonstrated a survival benefit for patients with advanced melanoma, non-small cell lung cancer (NSCLC) and renal cell carcinoma and are now being tested in many other cancers.2–5 Recently, agents targeting PD-1ligand 1 (PD-L1) as well as other immune checkpoint molecules and combination strategies have been developed.6 7

In healthy individuals, the role of the immune checkpoints is to promote tolerance to self-antigens, thereby preventing autoimmunity and damage to normal tissues. Many tumour types undergo these T cell inhibitory pathways as immune escape mechanisms in order to evade activated cytotoxic T cells.8 9 ICIs currently used in the clinical setting target CTLA-4 or PD-1 receptors expressed on T cells, or PD-L1 expressed on both antigen-presenting cells and tumour cells, allowing T cell activation and tumour destruction. By enhancing antitumour T cell activity, these treatments have resulted in some unprecedented long-lasting tumour responses in patients with unresectable or metastatic disease. However, releasing the natural brakes of CTLA-4 and PD-1 signalling on T cells may affect self-tolerance to healthy tissues, leading to immune side effects known as immune-related adverse effects (irAEs).

In clinical trials, reported irAEs mainly involved the gastrointestinal tract, skin, endocrine glands, liver and lung, but case reports suggest that nearly all organs can be affected. Rheumatic and musculoskeletal irAEs observed in patients receiving ICIs have been the focus of a recent systematic review of the literature.10 Arthralgia and myalgia were by far the most commonly reported rheumatic irAEs in clinical trials, with a prevalence ranging from 1% to 43% and from 2% to 21%, respectively. However, prevalence may have been underestimated as only high-grade irAEs were reported in some trials. In addition, case series and case reports provide more details on patients with inflammatory arthritis (IA) or disease patterns suggestive of polymyalgia rheumatica (PMR), vasculitis, polymyositis or sclerodermoid reaction induced by ICIs.11–13 Up to now, the two largest series of patients with rheumatic irAEs included 13 and 15 patients with IA and sicca syndrome, respectively.14 15 Recently, cases of seropositive rheumatoid arthritis (RA) with either rheumatoid factor (RF) and or/anti-cyclic citrullinated peptide (anti-CCP) antibodies were reported in a French retrospective study.16

To evaluate the prevalence and type of rheumatic and musculoskeletal irAEs in patients receiving ICI treatment at a single institution (University Hospital of Bordeaux, France), we implemented systematic referral to the department of rheumatology for evaluation of any rheumatic symptoms occurring with ICIs from September 2015 to May 2017.

Patients and methods

Study design

All patients aged 18 years and older treated with anti-CTLA-4 (ipilimumab and tremelimumab), anti-PD-1 (nivolumab and pembrolizumab) or anti-PD-L1 (atezolizumab, avelumab, durvalumab and MSB0011359C) agents alone or combined, either in clinical trials or in routine clinical practice in the department of oncology, dermatology, pulmonology and haematology from September 2015 to May 2017 were included in this prospective observational study. Safety data were recorded at each follow-up visit, just before administration of the next dose.

To evaluate the prevalence and type of rheumatic events associated to ICIs, a systematic referral of patients with any musculoskeletal symptom occurring during or after ICI treatment to the department of rheumatology was implemented. In order to organise this referral process, rheumatologists contacted the different specialists in charge of ICIs prescription at our institution, first by email, then during a face-to-face meeting. When musculoskeletal manifestations were observed in a treated patient, a rheumatologist (MK or LR) was contacted either by phone or by email, and a rheumatologic evaluation was organised. One or two reminders of our interest in such referral have been sent by email during the first weeks of the study. Patients with a known rheumatic or autoimmune disease were also included if they had a flare of the underlying disease or new musculoskeletal symptoms.

Rheumatological evaluation

For each referred patient, a rheumatologist (MK or LR) performed the history of musculoskeletal symptoms and the clinical evaluation. Demographic data (age and gender), personal or familial history of autoimmune diseases, type of cancer, ICIs received, the starting date, number of cycles and days before the occurrence of rheumatic irAE, previous cancer therapy and other associated irAEs were recorded. In addition, X-ray imaging and ultrasounds of swollen and/or painful joints were also obtained if needed. Patients were tested for serum inflammatory markers as well as serum autoantibodies depending on the clinical findings. Human leukocyte antigen - antigen D related (HLA-DR) phenotyping was also performed for some patients to search for shared epitopes. Finally, a treatment decision was made after discussion with the referring oncologist.

Evaluation of other irAEs and tumour response

Demographic data, type of cancer, the occurrence of other irAEs and tumour response according to ICI agent were also abstracted from medical records for patients without rheumatic manifestations, using local software DxCare and CHIMIO. For each patient, the best tumour response was collected and defined by the Response Evaluation Criteria In Solid Tumors 1.1 criteria as read by a radiologist on serial CT imaging.17


Data analyses were performed using Stata/SE software V.13.1. Associations of the occurrence of rheumatic or non-rheumatic irAEs and tumour response in relation to ICI treatment were analysed using Pearson’s χ2 test or Fisher’s exact test as needed. Subgroup analyses were also completed when appropriate, according to tumour type or ICI type. ORs were estimated with their 95% CIs. P values less than 0.05 were considered as statistically significant.


Patient characteristics

From September 2015 to May 2017, 524 patients received ICIs at our institution and were included in the study. Patients were treated either with anti-CTLA-4 (n=5) or anti-PD-1/anti-PD-L1 (n=407) agents alone or received a combination or sequential administration of both molecules (n=112). Demographic data, type of cancer and ICI molecules are listed in table 1.

Table 1

Demographic data, type of cancer and ICI molecules

Description of musculoskeletal irAEs

Thirty-five patients developed musculoskeletal symptoms after initiation of immunotherapy and were referred to the department of rheumatology (6.6%). Of note, two patients had a pre-existing autoimmune disease—one with axial spondyloarthritis and one with psoriasis. The median exposure time to ICIs was 70 days (range: 1 day–650 days), and for 80% of patients, the exposure time was less than 8 months. The mean age was 64 years (±SD 12 years) and 65.7% of patients were males. Cancer types included melanoma (n=16), Merkel carcinoma (n=1), NSCLC (n=13) and renal cancer (n=5) (table 2). Patients were treated with anti-PD-1 or anti-PD-L1 alone (n=30), or received anti-CTLA-4 and anti-PD-1 in combination or as sequential treatment (n=5), as described in online supplementary table 1.

Supplementary file 1

Table 2

Demographic data, type of cancer, ICI molecules and tumour response according to irAE occurrence

Reported musculoskeletal symptoms were varied, as detailed in online supplementary table 2. Both large and small joints in the upper and lower extremities were involved. The most common symptoms were arthralgia (all patients), with joint swelling being observed in only eight patients and myalgia in five patients. Ten patients complained of back pain. Overall, two distinct clinical presentations were observed: (1) IA mimicking either RA (n=7), PMR (n=11) or psoriatic arthritis (PsA)(n=2) and (2) non-inflammatory musculoskeletal conditions (n=15) (figure 1 and online supplementary figure 1).

Supplementary file 2

Supplementary file 3

Figure 1

Description of rheumatic irAEs occurring with ICI treatment. ANA, antinuclear antibodies; CCP, cyclic citrullinated peptide; CRP, C reactive protein; CTLA-4, cytotoxic T-lymphocyte-associated protein 4; ICI, immune checkpoint inhibitor; irAEs, immune-related adverse effects; NSAIDs, non-steroidal anti-inflammatory drugs; PD-1, programmed cell death protein 1; PD-L1, PD-1 ligand 1; PMR, polymyalgia rheumatica; PsA, psoriatic arthritis; RA, rheumatoid arthritis.

Inflammatory arthritis

IA occurred in 20 patients (3.8%). PMR was diagnosed in 11 patients, with a clinical presentation of acute predominant bilateral shoulder pain and stiffness in all patients. One patient had a pre-existing stable rheumatic disease (axial spondyloarthritis) and developed a PMR-like condition 20 days after ICI initiation. Increased levels of C reactive protein, defined as >5 mg/L, were found in seven patients, and subdeltoid bursitis was detected by ultrasound analysis in five patients. All but one patient fulfilled the 2012 European League Againt Rheumatism/American College of Rheumatology (EULAR/ACR) criteria for PMR. In the patient who did not fulfill the classical criteria of PMR, diagnosis was retained based on the typical clinical presentation of acute bilateral shoulder aching and stiffness, the absence of RF and anti-CCP antibodies and the rapid and complete resolution with 12.5 mg of prednisone. Nine patients were treated with oral glucocorticoids (mean dosage 15 mg/day), and clinical improvement was achieved in all of them while two patients recovered with non-steroidal anti-inflammatory drugs (NSAIDs).

Bilateral and symmetric hand pain and stiffness were observed in seven patients who developed rheumatic irAEs mimicking RA, mainly within the first 3 months after ICI initiation. Two patients developed psoriatic arthritis, including one with pre-existing psoriasis who experienced severe arthritis associated with psoriasis exacerbation. Clinically significant joint swelling and/or ultrasound-confirmed synovitis were observed in five patients. Elevated inflammatory markers were present in four patients. One patient tested positive for anti-CCP antibodies but negative for RF. Unfortunately, there was no available serum before ICI treatment to determine whether the anti-CCP antibodies were present before immunotherapy. Antinuclear autoantibodies (ANA) were positive at a titre of 1:160 and 1:640 with anti-extractable nuclear antigens (anti-ENA) negative in two patients, and a shared HLA-DRB1*01:01 epitope was present in two patients. X-ray evaluation did not detect any structural joint damage. All patients required prednisone or prednisolone at low-to-moderate doses, with a maximum of 30 mg/day, leading to clinical improvement or remission. One anti-CCP antibodies and RF-negative patient with RA and one patient with PsA required initiation of disease-modifying antirheumatic drug therapy (DMARD) with methotrexate. When remission was achieved, a progressively tapered glucocorticoid regimen was initiated. After 6 months of follow-up, two patients were able to discontinue glucocorticoids.

ICI therapy was pursued in all but one patient. This patient was participating in a clinical trial, and ICI treatment was temporarily discontinued in accordance with the study protocol.

Non-inflammatory musculoskeletal conditions

Fifteen patients presented with non-inflammatory rheumatic and musculoskeletal disorders (2.8%), characterised by pain, which worsened with physical activity and relieved with rest, and the absence of joint stiffness. All suffered from arthralgia of proximal or distal joints (mainly shoulders and hands), with associated back pain in five patients. Only one patient presented with swelling of interphalangeal joints together with hand, shoulder and cervical pain, associated with osteoarthritis based on the clinical examination, imaging and blood tests. Increased levels of C reactive protein were observed in four patients, but they were probably related to the malignancy, as they were present before the development of rheumatic symptoms. One patient benefited from a periarticular steroid injection for left epicondylitis, seven patients received opioid and/or non-opioid analgesics, including NSAIDs, three patients recovered after physiotherapy and one patient did not require any treatment. Two patients received prednisone (15 mg/day and 30 mg/day, respectively) due to cancer progression with development of renal metastasis, and one was treated with steroids for concomitant interstitial lung disease, leading to complete resolution of rheumatic symptoms. Importantly, no modification of immunotherapy was necessary.

Other irAEs

Overall, non-musculoskeletal irAEs were observed in 137 patients (26.1%) (table 2) and were associated with rheumatic irAEs in 16 patients. Reported irAEs were mainly cutaneous (rash, vitiligo, psoriasis and bullous pemphigoid), endocrine (thyroiditis and dysthyroidism) and digestive (colitis). Some cases of pneumonitis and hepatitis were also observed, as well as rarer irAEs such as pancreatitis, sarcoidosis, myasthenia, nephritis or uveitis.

Tumour response according to irAEs

Tumour response to ICI treatment was abstracted from medical records for all patients, with data available for 472 patients. Overall, there was a tumour response in 248 patients (52.5%), including complete response (n=31; 6.6%), partial response (n=84; 17.8%) or stable disease (n=133; 28.2%), while progressive disease was observed in 224 patients (47.5%). Rheumatic irAEs were observed in 12.1% of responders and in 2.2% of non-responders.

In patients with rheumatic irAEs, the tumour response rate was higher compared with patients without irAEs (85.7% vs 35.3%; OR=8.8 (95% CI 3.2 to 29.8), P<0.0001), (table 2). The proportion of responders to ICI treatment was also higher in patients presenting with other irAEs compared with patients without irAEs (75.1% vs 35.3%; OR=5.4 (95% CI 3.3 to 9), P<0.0001). There was no statistical difference between rheumatic and non-rheumatic irAEs regarding the relationship with tumour response (85.7% vs 75.1%, P=0.18). Subgroup analyses according to tumour type confirmed this relationship between rheumatic or other irAEs and tumour response across different cancers.


Arthralgia and myalgia were by far the most commonly reported rheumatic irAEs in clinical trials, while IA was less commonly described.10 Our prospective series estimates the prevalence of rheumatic manifestations in patients treated with ICI at 6.6% since we know the exact number of patients receiving ICI at our institution and since there was a systematic referral to the department of rheumatology of patients with any rheumatic symptoms. This study illustrates the wide range of clinical presentations, with the confirmation that the most frequent inflammatory manifestations are PMR-like syndromes, RA or PsA (3.8%), while others developed non-inflammatory disorders (2.8%), which has not been described previously. Moreover, the most novel message of this series is the strong association of these manifestations with tumour response, as patients experiencing rheumatic irAE had a better prognosis concerning the cancer outcome.

In accordance with the current literature describing rheumatic irAEs, PMR and RA were the two major clinical entities observed, with mainly seronegative RA, and only one patient being anti-CCP-positive in our series.14–16 However, the systematic referral of any musculoskeletal symptoms revealed new phenotypes of rheumatic irAEs. First, we observed some cases of PMR with a typical clinical presentation (acute onset, bilateral scapular and/or pelvic pain and stiffness) but with no increase in inflammation markers or only mild elevation. Second, non-inflammatory disorders were also noted in several patients, sometimes occurring a few days after ICI initiation and therefore suspected to be attributable to the treatment. When using the Naranjo algorithm, the adverse drug reaction probability score was rated as possible or doubtful for such non-inflammatory manifestations due to improvement despite ICI continuation, with alternative causes being possible, and because of the lack of similar events reported in the literature.18 Of note, there was no difference between inflammatory and non-inflammatory manifestations regarding the relationship with tumour response. Future clinical reports and a better understanding of the pathogenesis of these disorders will help to answer this question.7

The distinction between inflammatory versus non-inflammatory irAEs is worthwhile for treatment decisions, as patients might not require systemic glucocorticoids. For non-inflammatory conditions, symptomatic treatment with analgesics, NSAIDs, local injection or physiotherapy were often effective. For inflammatory irAEs, low-to-moderate doses of glucocorticoids were generally sufficient to achieve improvement or remission, without ICI discontinuation. When needed, methotrexate was added, which was the case for two patients. In our experience, as in the series reported by Belkhir et al,16 classical anti-inflammatory treatment was able to control rheumatic disorders without any modification of the ICI regimen, conversely to the experience of American colleagues, who reported a requirement for higher steroid dosage (often >40 mg/day), the use of tumour necrosis factor inhibitors for some patients and usually ICI discontinuation.14 15 Furthermore, the increasing number of cases reported in the literature illustrates these differences regarding rheumatic irAE management, highlighting the need for developing either national or international recommendations.

The higher prevalence of rheumatic irAEs with anti-PD-1/anti-PD-L1 exposure, either as a single agent or in combination with anti-CTLA-419–21 is a common observation from case reports and case series and was also recently reported from a single-centre pharmacosurveillance database.22 All patients referred to our department except one were receiving anti-PD-1 or anti-PD-L1 treatment. Even if ipilimumab is less commonly used currently, it is clear that anti-PD-1 and anti-CTLA-4 therapies have a distinct spectrum of irAEs.23 24 Of note, rheumatic irAEs were rarely reported with CTLA-4 blockade when ipilimumab was the major ICI used in melanoma patients, raising the question of whether the PD-1 pathway is the one mainly involved in ICI-related rheumatic manifestations.25 Animal models support this hypothesis, with arthritis being observed in PD-1 knockout mice but not in CTLA-4 knockout mice.26 27 Moreover, heterozygous CTLA-4 mutations in humans are associated with severe immune disorders such as thrombocytopaenia, hypogammaglobinaemia and lung, brain and gastrointestinal lymphocytic infiltrates but no clear rheumatic manifestations.28 Interestingly, some CTLA-4 and PD-1 polymorphisms are associated with RA,29 30 and recently, a breakdown of PD-1/PD-L1 checkpoints has been reported in giant cell arteritis.31 Unifying research efforts in our classical inflammatory/autoimmune diseases and in the ones induced by ICIs would result in a better understanding of costimulation pathways and would benefit the management of both diseases.

This study is the first to demonstrate that patients experiencing rheumatic irAEs had a higher rate of response to ICI treatment compared with those without irAEs. In our series, this strong association with tumour response was also observed with non-rheumatic irAEs, as described previously in the majority of case reports describing irAEs. Given the poor prognosis of advanced cancers, this clinical observation is of major interest and might be related, in part, to the fact that ICI treatment was continued in all but one patient. The role of the rheumatologists in this situation is to manage the rheumatic side effect with their clinical experience of our classical rheumatic diseases, using the drugs at disposition, namely, NSAIDs, glucocorticoids, non-biological DMARDs or even biologicals, and with the main objective of pursuing ICI therapy.32 Furthermore, a longer follow-up is needed and is ongoing for this study to assess whether the occurrence of irAEs affected patient survival, which represents a more clinically relevant efficacy outcome. The correlation between irAE occurrence and patient outcomes has been actively investigated but not fully elucidated and has yet to be validated in dedicated prospective studies. Among two retrospective analyses, one reported a better overall survival for nivolumab-treated melanoma patients experiencing irAEs compared with those without irAEs, while the other study, which involved ipilimumab-treated patients, concluded that tumour response or survival were not affected by the occurrence of irAEs.33 34 So far, the main hypothesis is that the occurrence of irAE could be the signal of a well-activated immune reaction leading to an efficient and durable antitumour response, as tumour neoantigens and normal tissue antigens could be cross-reactive.35 36 Rheumatologists, as specialists of rheumatic disease but also inflammatory systemic diseases, are closely and historically associated to immunology. Thereby, their expertise and their active participation in this evolving field of the cancer immunotherapy and its unique immune toxicity would be of great value for the oncology community.32


Since ICIs are used with increasing frequency and represent a new standard of care in some advanced cancers, knowledge of irAEs and their management is of major interest. Rheumatic irAEs occurred in 6.6% of treated patients, mainly associated with anti-PD-1 or anti-PD-L1 therapies and were easily manageable without ICI discontinuation. The key messages for the practising rheumatologist are summarised in the box. Collaboration between oncologists and rheumatologists is worthwhile to achieve a better understanding of these rheumatic irAEs and to define appropriate treatment algorithms.


Key messages for the practising rheumatologist

  1. Musculoskeletal symptoms are likely to occur in 5%–10% of cancer patients being treated with immune checkpoint inhibitors (ICIs).

  2. There are two distinct clinical manifestations: inflammatory arthritis (3.8%), mainly rheumatoid arthritis (RA), polymyalgia rheumatica (PMR) or psoriatic arthritis (PsA) , and non-inflammatory musculoskeletal conditions (2.8%).

  3. Treatment options comprise:

    • For the inflammatory arthritis: non-steroidal anti-inflammatory drugs (NSAIDs), glucocorticoids and conventionnal synthetic disease-modifying antirheumatic drugs (csDMARDs). A biological DMARD may be considered if there is no improvement.

    • For the non-inflammatory conditions: NSAIDs, analgesics and/or physiotherapy.

  4. Usually there is no need to stop or even to modify the ICI treatment.

  5. Treatment of these musculoskeletal manifestations must be based on a shared decision between the rheumatologist and the referring oncologist.

  6. Patients experiencing rheumatic manifestations or other immune-related adverse event(s) are likely to respond to ICI treatment (85%).



  • MK and LR contributed equally.

  • Handling editor Josef S Smolen

  • Contributors All the authors contributed to the manuscript: conception and design (LR, NM, MK and TS), collection of data (all authors) and statistical analysis (TB). MK, LR and TS drafted the manuscript, and all the authors critically reviewed and approved the final version of the manuscript.

  • Competing interests MB-B reports consulting and advisory boards for BMS and MSD France. AR reports being a member of Global, European and/or French Advisory Board in GU tumours and/or immunotherapy for Pfizer, Novartis, BMS, Roche, Astra Zeneca and MSD and received travel support from Pfizer, BMS, Roche, Astra Zeneca and MSD. SP reports consulting for BMS and travel support from MSD. AP-L has received honoraria and travel support from BMS and MSD. RV reports being one of the investigators for a clinical trial from BMS, consulting and advisory boards for BMS and MSD and travel support from BMS and MSD. All the others authors declared no conflict of interest for this work.

  • Patient consent Detail has been removed from this case description/these case descriptions to ensure anonymity. The editors and reviewers have seen the detailed information available and are satisfied that the information backs up the case the authors are making.

  • Ethics approval The study was approved by the local ethics committee (CE-GR-2017/007).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement We are publishing all data either in the paper or in the online supplementary material.