Neuromuscular adverse events associated with anti-PD-1 monoclonal antibodies

Neuromuscular adverse events following cancer treatment with anti-programmed cell death protein 1 (PD-1) monoclonal antibodies are relatively rare, yet potentially fatal. We performed a systematic review to characterize the clinical presentation, diagnostic workup, and management of neuromuscular disorders (NMDs) in patients treated with nivolumab or pembrolizumab monotherapy or concurrent with other immunologic agents, such as ipilimumab. Sixty-one publications on 85 patients (mean age 66.9 years [range 34–86]; male/female 2.6:1; 59% metastatic melanoma) were identified from selected indexing databases until June 2018. Forty-eight patients had received nivolumab and 39 pembrolizumab. The mean number of PD-1 inhibitor treatment cycles prior to onset of symptoms was 3.6 (range 1–28). Symptoms included oculomotor (47%), respiratory (43%), bulbar (35%), and proximal weakness (35%), as well as muscle pain (28%). Diagnoses were categorized as myasthenia gravis (27%), neuropathy (23%), myopathy (34%), or a combination of these (16%). After a critical review of the data, however, evidence did not support the stated NMD diagnosis in 13% of cases, while up to 25% of patients had signs of additional NMDs. Cardiac complications occurred in more than 30% of patients diagnosed with myasthenia gravis or myositis. Mortality was high in these patients, despite adequate treatment strategies including corticosteroid, IV immunoglobulins, and plasma exchange. The clinical presentation of NMDs associated with PD-1 inhibitors is often atypical, with considerable overlap between myasthenia gravis and myopathy, and cardiac/respiratory complications are common.


Introduction
Nivolumab and pembrolizumab are immune checkpoint inhibitors (ICI) that inhibit the programmed cell death protein 1 (PD-1), thereby stimulating the immune system to react against cancer cells. 1 They have been approved for the treatment of, for example, metastatic melanoma, non-small cell lung cancer, urothelial carcinoma, and renal cell carcinoma. [1][2][3] The discovery of ICI was awarded the Nobel Prize in Physiology or Medicine in 2018. 4 However, it has become apparent that many patients experience immune-related adverse events (AE). [5][6][7] For instance, the estimated frequency of neurologic immune-related AEs is 0.8%-7.7%. [5][6][7][8][9] Immune-related AEs more often involve the peripheral nervous system than the CNS, [8][9][10][11] and the number of case reports on neuromuscular AEs following nivolumab or pembrolizumab has steadily increased, in particular those on side effects suggesting myasthenia gravis (MG), myositis, or peripheral nerve diseases. However, these disorders can cooccur in an individual patient, and the phenotypes might differ from classic neuromuscular disorders (NMDs) in terms of clinical presentation, laboratory workup, treatment response, and prognosis. [12][13][14][15][16][17][18] The use of PD-1 inhibitors has rapidly increased since their first approval in 2014. Neurologists and other clinicians are therefore likely to encounter more oncologic patients with neuromuscular symptoms. 5,8,10 We performed a systematic review to characterize the symptoms, clinical findings, and laboratory results of patients with cancer treated with anti-PD-1 monoclonal antibodies and treatment-associated neuromuscular disease.

Methods
We conducted a systematic review of the literature according to standard systematic review methodology (Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA] 19 ).

Standard protocol approvals, registrations, and patient consents
The complete review protocol, as well as the PRISMA checklist, can be accessed in the online supplementary file and is registered with PROSPERO (crd.york.ac.uk/prospero/).

Primary and secondary objectives
Using the patients, intervention, comparison, outcome (PICO 20 ) approach, we phrased the following primary research question: c In patients treated with an anti-PD-1 monoclonal antibody for disseminated cancer or melanoma, who develop neuromuscular symptoms or who have preexisting neuromuscular symptoms that become more pronounced following anti-PD-1 monoclonal antibody therapy (P), do neurologic examination and laboratory workup, including neurophysiology and antibody titers, (I) sufficiently discriminate between neuropathies and myopathies, including myositis, on one hand (C) and neuromuscular junction failure compatible with myasthenia gravis on the other (O)?
We further phrased 2 secondary research questions: c Do patients treated with an anti-PD-1 monoclonal antibody for disseminated cancer or melanoma (P), who fulfill clinical and laboratory criteria for anti-PD-1 monoclonal antibody-associated myasthenia gravis (I), respond better, similar, or worse to treatment with pyridostigmine and immunomodulation (C) than patients with myasthenia gravis in general (O)? c Do patients treated with an anti-PD-1 monoclonal antibody for disseminated cancer or melanoma (P), who fulfill clinical and laboratory criteria for anti-PD-1 monoclonal antibody-associated myasthenia gravis (I), have a better, similar, or worse prognosis with respect to their myasthenia (C) than patients with myasthenia gravis in general (O)?

Search strategy
We evaluated all case reports, cross-sectional or longitudinal, retrospective or prospective observational studies, as well as interventional trials reporting on neuromuscular symptoms following treatment with the anti-PD-1 monoclonal antibodies nivolumab or pembrolizumab.
We included only articles that allowed assessment of patient data at the single-subject level. We excluded articles that concerned patients already used in another article by the same authors (or the same institution). We included studies published in English and listed in Medline (PubMed), Cochrane Central Register of Controlled Trials (The Cochrane Library), and ClinicalTrials.gov up until June 30, 2018. The full search string can be found in supplementary files (doi.org/10.5061/dryad.84q73s8).
We evaluated all abstracts and identified eligible studies based on their full text. We included non-English literature only if an English abstract was available and the article could reliably be translated into English. The reference list of relevant articles was manually searched to identify additional articles.
S.J.C. and A.J. performed the initial selection and further review. Following the identification of relevant studies, A.J. and S.J.C. independently extracted the relevant information from each study. Disagreement was settled by D.K. Figure 1 visualizes our data extraction points.

Participants
We included adults (age ≥18 years) with anti-PD-1 monoclonal antibody therapy for disseminated cancer or melanoma, regardless of prior or concomitant therapy, presenting in health care facilities with neuromuscular symptoms, and who were diagnosed with a disorder of the peripheral nervous system. Patients were included Glossary AChR-ab = acetylcholine receptor antibody; AE = adverse event; CPK = creatine phosphokinase; ICI = immune checkpoint inhibitors; IVIg = IV immunoglobulin; MG = myasthenia gravis; MuSK-ab = muscle-specific kinase antibody; NCS = nerve conduction studies; NMD = neuromuscular disorder; OR = odds ratio; PD-1 = programmed cell death protein 1; PEX = plasma exchange; PRISMA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RCT = randomized controlled trial.
irrespective of comorbidities and history of neuromuscular diseases.

Target conditions
The target condition was defined as signs of neuromuscular disease compatible with MG, myopathy, or neuropathy. Based on the reported findings, we sorted each patient case into the following categories: (1) diagnosis likely; (2) diagnosis likely, but signs of codiagnosis; (3) diagnosis possible, but not corroborated; (4) diagnosis less likely, but signs of other NMDs, based on the criteria listed in table 1.

Systematic literature search
A total of 112 articles were identified in the database searches and 2 additional articles from reference lists. See the supplementary files (doi.org/10.5061/dryad.84q73s8) for a flowchart of the literature search and supplemental list of references. Sixty-one articles 8,9,11-18,21-60,e61−e70,e88 with 85 individual patients met the inclusion criteria; the first article was published in 2014. 51 Most articles were case reports (n = 57, 93%); only a few were retrospective  clinical database studies (n = 4). Prospective data were unavailable. We also included data from 1 additional patient who was admitted to our clinic (illustrated case description; Figure 2).
Odds ratios (ORs) for selected parameters for nivolumab vs pembrolizumab are shown in table 2. The only statistically significant finding was the proportion of patients diagnosed with metastatic melanoma, which was smaller for nivolumab than for pembrolizumab; OR 0.29 (95% confidence interval 0.12; 0.73, p = 0.009). Fourth line monotherapy with nivolumab was initiated in 2017 because of progressive metastatic disease. Six weeks later, the patient was admitted because of dyspnea and muscle weakness, which fluctuated and worsened in the evenings. Examination revealed left-sided abducens nerve palsy and ptosis, dysphagia, proximal weakness, and normal sensation. EKG showed a left branch block. Creatine phosphokinase was normal (33 U/L; reference 40-280 U/L) and myoglobin slightly increased (92 μg/L; reference 24-77 μg/ L). Tests for acetylcholine receptor antibody and muscle-specific kinase antibody were negative. However, an ice pack test was positive; that is, following application of an ice package on both eyes, ptosis was less pronounced in both eyes (A: before ice pack test; B: after). EMG revealed myopathic changes with reduced amplitudes of the motor unit potentials, while nerve conductance velocities and amplitudes were normal. Repetitive nerve stimulation was without abnormal decrement. Pyridostigmine was commenced without apparent effect. Treatment with IV immunoglobulin was planned, but the patient died the following day due to acute respiratory insufficiency and aspiration.  Neuromuscular adverse events Table 3 shows selected data points related to symptoms and laboratory investigations based on localization in the peripheral nervous system. We classified the diagnoses as MG (n = 23, 27%; including ocular, bulbar, or generalized MG), neuropathy (n = 20, 23%; typically Guillain-Barré syndrome, but also 2 cases of vasculitic neuropathy, 1 chronic inflammatory demyelinating polyneuropathy, and 1 Bell palsy), or myopathy (n = 29, 34%; mostly unspecified myositis or rhabdomyolysis). A combination of NMDs was seen in 14 patients, of whom 11 (79%) were diagnosed with MG and myopathy, 2 (11%) with myopathy and neuropathy, and 1 (7%) with MG, myopathy, and neuropathy. 12 Of the 23 patients diagnosed with MG, 8 (35%) had a history of preexisting MG, 11,29,31,45,46,49,56,e71 and these were diagnosed with either MG exacerbation or reactivation. They had been diagnosed 4-25 years prior to ICI treatment based on relevant symptoms (6/8 explicitly had ocular symptoms; 6/8 were acetylcholine receptor antibody [AChR-ab]positive). All were in spontaneous or pharmacologic remission at the time of ICI treatment. Three of the 8 patients were treated with nivolumab, 5/8 were male, and this subgroup did not differ from the group of de novo MG in terms of symptomatology, laboratory workup, treatment, or outcome.
The frequencies of symptoms are presented in table 3. Of note, unspecified weakness occurred in many patients, making it difficult to identify a pattern of affected muscle groups. Although a hallmark of MG, fatigable weakness was only reported in 7 (30%) cases of pure MG and in none of the cases with >1 NMD diagnosis. Head-drop and respiratory symptoms were most prevalent among patients with a combination of NMDs. Cardiac involvement was reported in 3 (13%) MG cases and in 11 (38%) and 6 (43%) myopathy and combination cases, respectively, while not seen in any of the patients with neuropathy. Concurrent noncardiac immune-related AEs were reported in 9/31 (31%) patients with myopathy, 5/20 (25%) with neuropathy, 2/14 (14%) in the combination group, and 1/ 23 (4%) with MG.

Laboratory workup Neuromuscular junction antibodies
AChR-ab status was available for 20/23 patients diagnosed with MG, of whom 50% had a positive antibody titer. Ten patients had been tested for muscle-specific kinase antibody (MuSK-ab) as well, but all were negative. In 29 patients with myopathy, 2/10 tested had a positive AChRab titer, while none of the 7 patients tested was anti-MuSK positive.
In patients diagnosed with more than one type of NMD, all 14 had been tested for AChR-ab, and the titer was positive in 8/ 12 patients with a co-diagnosis of MG. MuSK-ab were reported for 7/14 patients and positive in 1 patient (with a titer of 0.02 nmol/L [normal range 0.00-0.01 nmol/L]) 52 ; this patient had a history of ocular AChR-ab+ MG and was subsequently diagnosed with both myositis and AChR-ab+ MG. Ca + -channel-ab were tested in 2 patients (diagnosed with MG) and were negative.  (table 3).

Neurophysiologic examination
Data on neurophysiologic examination were less often reported (table 3). Only a few articles provided information on specific neuromuscular junction examinations, that is, single-fiber EMG (for MG; 4/23 patients, all with abnormal jitter) and repetitive nerve stimulation (8/23 patients; 5 pathologic). In the combination group, single-fiber EMG was reported in 2/14 patients (1 showed abnormal jitter), while repetitive nerve stimulation was normal in all 6 cases tested.  Copyright © 2019 American Academy of Neurology. Unauthorized reproduction of this article is prohibited.

Treatment and outcome
Corticosteroids were the most widely used treatment (table  3). IVIg and plasma exchange were also frequently used, especially in patients with a combination of NMDs. Pyridostigmine was initiated in 11 (48%) diagnosed with MG and 9 (64%) with >1 diagnosis.
Most patients improved to some degree following treatment onset; 19 (95%) patients with neuropathy, 19 (83%) with MG, 20 (69%) with myopathy, and 8 (57%) in the combination group. The time to follow-up was often not explicitly stated, and thus mortality rates should be interpreted with caution, but it is noteworthy that there were no deaths in the neuropathy group. Data on the final cause of death were insufficient to estimate the percentage of mortality attributed to neurologic immune-related AEs.
Data were deemed insufficient for a more detailed metaanalysis.

Inconsistencies in the diagnosis of neuromuscular symptoms
Evaluating each case report with respect to the criteria formulated in table 1, we found that for 67 (78%) of the patients included, clinical and laboratory findings were sufficient to deem the diagnosis likely, with no sign of other NMD pathology (  Abbreviations: AChR-ab = acetylcholine receptor antibody; CPK = creatine phosphokinase; IVIg = IV immunoglobulin; NCS = nerve conduction study. The diagnostic categories refer to the diagnosis stated by the authors of the individual case reports. a This group includes 11 with myasthenia gravis/myopathy, 2 with myopathy/neuropathy, and 1 with myasthenia gravis/myopathy/neuropathy. b Unless otherwise stated, percentage refers to number of patients with symptoms out of the total number of patients in the group. c Data available for n = 21 (myasthenia gravis), n = 17 (neuropathy), n = 26 (myopathy), and n = 14 (combination). d Data available for n = 15 (myasthenia gravis), n = 12 (neuropathy), n = 22 (myopathy), and n = 13 (combination). e Including ptosis, ophthalmoplegia, and diplopia. f Including dysphagia and dysarthria. g Any form of cardiac symptom or laboratory finding. h Positives as the percentage of number tested.  Myasthenia gravis c Five cases had signs of other NMD pathology; all had elevated CPK in the range 1,156-6,566 U/L, pointing to muscle involvement, and one had myopathic changes in the EMG, 37 while EMG data were not reported for the remaining 4 patients. 27,33,41,49 c Three cases lacked specific clinical (signs of fatigable weakness) and laboratory findings corroborating the diagnosis; 2/3 had tested negative for AChR-ab and MuSK-ab, 50,53 while antibody testing was not reported in the third case. 31 Single-fiber EMG and CPK results were not reported for either of them, but one had normal nerve conduction studies (NCS) (including repetitive nerve stimulation) and EMG.
c Finally, 2 cases lacked support for the diagnosis but showed signs of other NMD; again, symptoms could be attributed to MG, but fatigability was not mentioned. In one case, AChRab titer was normal, 8 but not reported in the other. 45 The first case had NCS findings that were interpreted as critical illness neuropathy (and a normal EMG), while the latter had slightly elevated CPK, pointing to muscular involvement.
Neuropathy c Three case reports did not present results supporting the diagnosis; one patient was diagnosed with cranial polyneuropathy, but symptoms were not described, and NCS/EMG was reported as normal. 11 For 2 cases of polyradiculitis 8 and one of Bell palsy, e69 diagnostic workup was unknown.
c For one patient with meningo-radiculitis, 8 diagnostic data were not reported in full, but the patient had muscle pain, which could indicate some degree of muscle involvement.

Myopathy c
Three cases had signs of other NMD pathology; one patient treated with nivolumab for a thymoma was diagnosed with myositis and rhabdomyolysis based on a CPK of >40,000 U/L and biopsies showing inflammation, but also had a positive AChR-ab titer. 28 It is unclear whether the AChR-ab were related to the underlying thymoma e70 or nivolumab treatment, and whether they were implicated in the pathology of this particular patient. Another patient 47 with anti-striated muscle-ab titer of 1:30,270, CPK of 444 U/L, had bulbar symptoms and a positive AChR-ab titer of 0.10 nmol/L. The third patient 9 had elevated CPK and normal EMG, but NCS showed signs of length-dependent peripheral neuropathy. A muscle biopsy showed necrotic fibers without inflammation.

Combination of diagnoses
c Two patients with MG-concurrent myositis 13,e63 did not report any symptoms or clinical findings suggestive of fatigable weakness. Both patients had a negative AChR-ab test, while single-fiber EMG and repetitive nerve stimulation were not reported. CPK was moderately elevated, and NCS showed signs of polyneuropathy in one of the patients.

Diagnostic workup
We investigated if patients with neuromuscular symptoms after nivolumab or pembrolizumab treatment could be classified into conventional diagnostic groups, that is, NMJ/MG, myopathy, or neuropathy (primary research question). Comprehensive diagnostic workup including neurophysiology, antibody titers, muscle enzymes, and biopsies was only rarely available (and if so, only from single case reports). Hence, the presented data are too heterogeneous to support a clear analysis of the sensitivity and specificity of clinical and laboratory findings.
Data were compatible with the reported diagnosis (i.e., without signs of another NMD) in 78% of patients (table 4); diagnostic precision was lowest in the group diagnosed with MG (57%) and highest in myopathy (90%). Nine percent of patients had findings suggesting codiagnosis of another NMD (5 with MG, 3 with myopathy), and in 8%, the stated diagnosis appeared uncertain (3 with MG, 3 with neuropathy, 1 with >1 diagnosis). In 5%, we found that the reported diagnosis was not supported by the diagnostic procedures, which instead pointed to a different NMD.
These numbers indicate that a broad and thorough diagnostic workup is needed when patients treated with PD-1 inhibitors show signs of neuromuscular adverse events. In particular, clinicians should note that up to 25% of patients show signs of overlapping NMDs, especially MG and myopathy. This finding is consistent with data from a retrospective case-control study on MG after nivolumab treatment. e72 Findings also seem to differ from idiopathic NMDs in other regards; for example, there is a lower proportion of AChR-ab positive MG e73 and a high prevalence of oculomotor and bulbar symptoms in patients with myopathy. The latter can easily prompt a diagnosis of MG, but if the element of fatigability is absent and diagnostic workup does not support neuromuscular junction failure, other explanations must be sought, and neurologists specialized in NMDs should be consulted. Indeed, often neurologists are not involved at all, but patients are investigated and treated by their attending oncologists alone; in 32 (37%) of the patient cases, none of the authors was affiliated with a neurologic department. We also found a high prevalence of respiratory symptoms/complications in patients with MG and myopathy (often requiring ventilatory assistance), likely contributing to the high mortality rate for these patients. This again shows the need for an accelerated and broad diagnostic workup to ensure that treatment is instituted quickly.
A high number of patients (all with MG or myopathy) showed signs of cardiac involvement in the form of elevated myocardial enzymes, 13-15,23,24,28,33,40,44,48,52,e61,63 EKG changes, 28,33,48 abnormal echocardiography, 44 cardiac arrest, 16 or cardiac symptoms prior to death. 11,43,44,50 The coexistence of MG, myopathy, and myocarditis was also noted by Suzuki et al. e74 The association seems to be stronger with myositis (table 3), but clinicians should be aware of the risk of cardiac complications in both groups.
In patients with myopathy, muscle biopsies showed presence of inflammatory cells (e.g., CD8+ cells), necrosis, and, to a lesser extent, macrophages. This resembles the features of polymyositis and is contrary to necrotizing autoimmune myositis, where predominately macrophage-mediated necrosis is seen. e75 We identified an unbalanced sex ratio in the reviewed studies with 72% male patients. This is markedly different from the reported sex ratio for melanoma e76 and non-small cell lung cancer, e77 which accounted for 76% of the cancer types in this review. The reason for this skewing is unclear, but one might speculate that, for unknown reasons, male oncologic patients are more prone to neuromuscular adverse events or female patients are less often treated with anti-PD-1 therapy. This is an important area for future research.

Treatment and outcome
Corticosteroids were the most frequent treatment, followed by IVIg and plasma exchange (PEX), reflecting the common approach when managing severe cases of MG and other NMDs. Although corticosteroid treatment for MG has been in use since the 1970s, e78 no randomized controlled trial (RCT) on idiopathic MG has been conducted. e79 Likewise, no RCTs exist with sufficient power for either IVIg or PEX to clarify the efficacy of these treatments on idiopathic MG. e79−e81 Furthermore, few authors of the articles included in this review gave a thorough description of the timing of these treatments, making it impossible to separate and compare treatment effects (secondary research question).
The available data also cannot sufficiently answer whether clinical outcome differs between patients with idiopathic MG and patients with ICI-related MG (secondary research questions). However, in an epidemiologic study on MG during the 1940-2000 period, the average lifespan following diagnosis was 10.3 years, e82 while in another study e83 researchers found the 3-year survival rate to be 85%. Although our study lacks systematic and detailed follow-up data, 11 of 23 patients with isolated MG died (2 due to cancer progression and one due to "unrelated cardiac issue" 50 ), perhaps suggesting a more severe course of PD-1related MG. Further, in a retrospective case-control study e72 researchers found neuromuscular decompensation to be more pronounced in patients with anti-PD-1 treatment. However, other factors likely contribute; increased overall morbidity/ mortality due to cancer, higher age (mean age 71.9 years in our study vs 45.3 and 35.3 years in idiopathic MG e82 for men and women, respectively).
Standardized prospective multicenter studies will allow for more accurate estimates on treatment response and clinical outcome, but given the relatively low incidence of neuromuscular adverse events compared to other autoimmune adverse events, 5,7,e72,e84 this could prove difficult.

Limitations
Some limitations should be acknowledged. Of note, we may have underestimated the true frequency of NMD characteristics, because the lack of reported signs and symptoms in individual case reports is not necessarily evidence for their factual absence. We would expect findings supporting the diagnosis given by the respective authors to be presented, but diagnostic information was often insufficient in detail. Thus, we cannot conclude with confidence if diagnostic inconsistencies are due to inadequate workup or misinterpretation of results. Further, the frequency of overlapping NMDs may be underestimated. Also, we were unable to quantify treatment effects and mortality compared with idiopathic MG because the data lacked sufficient detail and because a gold standard for treatment response in idiopathic MG is not established. Moreover, the effects of metastatic cancer itself are difficult to distinguish from severe autoimmune disease.

Future directions
Reviewing the current data raises important questions, not the least about the pathophysiology of anti-PD-1-associated MG and NMDs; for example, why do some patients experience symptoms after just one dose, while others go through several cycles before symptom onset? As for MG, does anti-PD-1 treatment initiate or unmask latent disease? Is there a paraneoplastic component? Are there any unidentified molecular targets in patients with atypical presentation or laboratory findings?
The diagnosis of PD-1 inhibitor-associated neuromuscular adverse events requires thorough work-up ( figure 3). Incorrect or incomplete diagnoses are made in more than 20% of cases. Clinical presentation is often atypical with a considerable overlap between MG and myopathy. Cardiac and respiratory complications are frequent and associated with a more severe disease course. Of note, neurologists are not always consulted but should get involved as early as possible, since immediate recognition of neuromuscular adverse events is crucial. Treatment delay may otherwise lead to increased morbidity and mortality. We suggest continuous screening for neuromuscular symptoms and signs, including basic investigations such as CPK, troponin levels, and EKG, in patients started on PD-1 inhibitor treatment, in particular in those with a history of a neuromuscular disorder.