Neurologic complications of immune checkpoint inhibitors

Abstract

Introduction:

Immune checkpoint inhibitors (ICI) are associated with a wide spectrum of neurologic immune-related adverse events (irAEs) including meningo-encephalitis, myasthenia gravis and various neuropathies. Although relatively rare, they often present significant diagnostic complexity and may be under-recognized. Permanent neurologic deficits and/or fatality have been described but improvement is noted in most cases with ICI discontinuation and immunosuppressive therapy.

Areas Covered:

This review highlights the most frequently reported ICI-associated neurologic toxicities with a particular focus on those that may be more severe and/or fatal. Data from case series and pharmacovigilance studies is leveraged to provide an overview of associated clinical features, expected outcomes and appropriate management. Various immunobiologic triggers have been proposed to explain why certain patients might develop neurologic irAEs and are also briefly discussed.

Expert Opinion:

All providers who care for patients with cancer should be made aware of common neurologic irAEs and able to recognize when prompt evaluation and consultation with appropriate specialists are indicated. Symptoms suggestive of encephalitis, myasthenia-gravis or an acute polyradiculopathy such as Guillain-Barre Syndrome (GBS) in patients exposed to these agents warrant immediate attention with a low threshold for hospitalization to expedite work-up and monitor for severe and/or life-threatening manifestations.

1. Introduction

The discovery, development and rapid implementation of immune checkpoint inhibitors (ICI) has unequivocally revolutionized the treatment of metastatic cancer over the last decade [1]. Encouraging response rates and long-term outcomes associated with these agents have unfortunately been complicated by the increasing recognition of a wide spectrum of associated immune-related toxicity [2]. Adaptive immune dysregulation plays an integral role in the development and progression of many malignancies, most notably in the setting of a high mutational burden or other immunogenic features, which are particularly common in melanoma. Tumors often directly or indirectly co-opt “immune checkpoints” including PD1/PDL1 and CTLA4 that function to maintain self-tolerance in healthy tissue in order to evade immune detection. Antibodies that specifically target these molecules promote immune surveillance and often lead to a robust anti-tumor immune response and host-mediated destruction of malignant cells [3].

The effects of checkpoint inhibition are however infrequently limited to the tumor microenvironment. PD1/PDL1 and CTLA4 are widely expressed across various tissue types and down-regulation can trigger a broad array of auto-immune toxicity. The most frequently noted immune-related adverse events (irAEs) involve inflammation of gastrointestinal, dermatologic, endocrine or pulmonary organs. Increasing use and awareness of ICIs has helped to establish characteristic features of these more common toxicities. Treatment of irAEs consists of three distinct pillars. First, ICI should be discontinued in severe cases. However, the long pharmacokinetic and pharmacodynamic effects (lasting weeks to months) makes this insufficient alone to mitigate the severe inflammation. Second, high-dose steroids or other immunosuppressants are used to dampen the ongoing inflammation. Organ specific second-line treatments may also be required, including infliximab for colitis and mycophenolate mofetil for hepatitis. Finally, supportive care is essential in some cases (for example, fluids and electrolyte replacement for colitis, oxygen for pneumonitis). This framework is useful when considering therapies for neurologic irAEs.

Neurologic irAEs may be particularly difficult to recognize and/or diagnose as symptoms are frequently non-specific. Data is limited primarily to case series that describe the onset of auto-immune or inflammatory conditions with a temporal relationship to checkpoint inhibition. Extrapolation from case reports and pharmacovigilance data suggests that neurologic toxicity occurs in 1–5% of patients treated with ICIs, which comprise a fairly broad spectrum of events involving the central, peripheral, and autonomic nervous systems individually or in combination [4, 5]. The true incidence is difficult to estimate but may be higher due to frequent under-recognition and/or under-reporting. Of note, while the general mechanisms of irAEs are fairly well understood (i.e., removal of key negative immune regulators), the specific reasons why individual patients experience neurologic or other irAEs are not known.

The most commonly reported neurologic irAEs include myasthenia gravis, encephalitis/meningitis, inflammatory polyradiculopathies such as Guillain-Barre syndrome, and peripheral neuropathy [6]. Although uncommon, these toxicities may be associated with permanent or long-term sequalae and occasional fatality. The risk of severe and/or permanent neurologic toxicity may be mitigated by prompt recognition and appropriate management. Further characterization and awareness of the spectrum of ICI-associated neurologic toxicity may therefore improve outcomes and decrease morbidity among the growing population of patients treated with checkpoint inhibitors.

2. Overview of immune-related adverse events

Immune checkpoint inhibitors function by blocking either PD1 or its ligand or CTLA4, two key receptors involved in immune regulation primarily via effects on T-cell activation and function. CTLA4 is a T-cell specific receptor that competes with CD28, a homologous T-cell specific receptor, to bind with stimulatory ligands on antigen presenting cells (APCs). In addition to interaction between the T-cell receptor and peptide-MHC complexes on APCs, T-cell activation requires a second signal triggered by binding between CD28 and associated ligands CD80/86 on APCs. The binding of CD80/86 to CTLA4, which competes with CD28 for these ligands, triggers down-regulation of T-cell activation. Persistent antigen exposure and immune activation leads to up-regulation of CTLA4, dampening the immune response and helping to prevent ongoing activation, particularly towards self-antigens [7].

While CTLA4 is primarily involved with the earlier stages of T-cell activation, primarily in lymphoid organs, PD1 works to control T-cell activity in peripheral tissues [3]. The interaction between PD1 and its ligands, PDL-1 and PDL-2, expressed primarily on APCs, inhibits T-cell function and proliferation, induces apoptosis and promotes differentiation into regulatory T cells. As with CTLA4, chronic antigen exposure and persistent immune activation may lead to the up-regulation of PD1 expression on the surface of T cells. Tumor cells and certain viral infections may also exploit this interaction by expressing PD-1 ligands to avoid immune detection [8]. Antibodies that block these inhibitory receptors or “immune checkpoints” bolster the anti-tumor response by promoting T-cell activation and function and increasing immune exposure to tumor specific antigens.

While these agents have led to remarkable responses in a wide variety of malignancies, immune checkpoints play an important role in promoting self-tolerance and preventing auto-immunity. Removing physiologic “brakes” on cell-mediated immunity can lead to widespread effects outside of the tumor microenvironment. Circulating activated T-cells targeting self-antigens and/or inflammatory cytokines may lead to inflammation and/or destruction in peripheral tissue, resulting in clinical symptoms of auto-immune disease. The majority of patients treated with a checkpoint inhibitor will develop some form of immune-related toxicity. Involvement of almost every organ system has been noted but the most frequently reported iRAEs involve the skin or the gastro-intestinal system (Table 1). While iRAEs have been reported in up to 90% of patients treated with anti-CTLA4 agents and 70% of those treated with anti-PD1/PDL1, grade 3 or 4 toxicities occur in an estimated 27% and 16% respectively [9–11]. Grade 3 or 4 toxicities are notably more frequent when both agents are used in combination and occur in about 55% of patients [11].

Table 1.

Most frequently reported immune-related adverse events with the combination of ipililumab/nivolumab or either agent alone among melanoma patients enrolled in the CheckMate-067 trial.

	Combination ICI		Anti-PD1		Anti-CTLA4		Clinical Features
	Any Grade	Grade 3–4	Any Grade	Grade 3–4	Any Grade	Grade 3–4
Cutaneous	62%	6%	46%	2%	56%	3%	Morbilliform eruption can be seen with ipilimumab, skin manifestations less severe with anti-PD1
Rash/Pruritis	53%	6%	37%	2%	51%	3%
Vitiligo	9%	0	9%	<1%	5%	0
Gastrointestinal	48%	15%	22%	4%	38%	12%	Diarrhea in almost all patients with colitis (92%); abdominal pain, nausea/vomiting
Colitis	13%	8%	2%	1%	11%	8%
Diarrhea	45%	9%	21%	3%	34%	6%
Endocrine	34%	6%	17%	2%	12%	3%
Hypothyroidism	17%	<1%	11%	0%	5%	0%	Hyperthyroidism is usually transient and followed by permanent hypothyroidism
Hyperthyroidism	11%	1%	4%	0%	1%	0%
Hypophysitis	7%	2%	1%	<1%	4%	2%	Hypophysitis presents with fatigue, headache, weakness, characteristic findings noted on MRI; can result in multiple hormone deficiencies; adrenal insufficiency generally permanent
Adrenal Insufficiency	4%	2%	1%	1%	1%	<1%
Hepatoxicity	33%	20%	8%	3%	7%	2%	Predominantly asymptomatic elevations in transaminases; rare severe auto-immune like hepatitis and liver failure
Pneumonitis	7%	1%	2%	<1%	2%	<1%	Dyspnea and cough most common symptoms but variable clinical and radiographic presentation

Despite a high incidence of associated immune-related toxicity, several large meta-analyses have indicated that ICIs are generally better tolerated than chemotherapy with lower overall adverse event rates, lower rates of higher-grade AEs and less frequent treatment discontinuation due to toxicity [12, 13]. A majority of iRAEs are low grade and can be managed effectively by withholding therapy and administering corticosteroids. Checkpoint inhibitors are however occasionally associated with rapidly progressive and life-threatening immune-related toxicities and patients should be closely monitored for suggestive symptoms.

Factors that influence which patients will experience immune-related toxicities, the severity of associated symptoms and which organ systems are involved are largely unknown and specific underlying pathogenic mechanisms are poorly understood. Tissue from organs affected by immune-related toxicity often demonstrates extensive infiltration by activated CD4+ and CD8+ T cells. Some studies have demonstrated a greater diversity in the number of productive TCR sequences among patients who experienced iRAEs compared with those without symptoms suggestive of immune-related toxicity. Mobilization of a wider spectrum of T-cells may increase the likelihood of self-reactivity and associated symptoms [14]. Extensive inflammation and cell lysis often triggered by checkpoint blockade may also contribute to epitope spreading [15]. Recruitment of additional T-cell populations that target new epitopes within the tumor microenvironment may enhance the anti-tumor response but may also promote cross-reactivity and/or recognition of self-antigens that ultimately trigger an auto-immune response.

One proposed mechanism of immune-related toxicity suggests that attenuation of an immune response directed against antigens shared by both tumor and non-malignant tissue may lead to self-directed immunity outside of the tumor microenvironment. Neoantigens encountered by immune surveillance following the destruction of tumor cells may also play a role in priming secondary immune responses, which are then activated when shared or homologous antigens are encountered in peripheral tissue [15]. Some studies have suggested that anti-CTLA4 agents may promote down-regulation of regulatory T-cells, impacting their role in suppressing cell-mediated immunity. PDL1/PDL2 have also been shown to be upregulated by regulatory T cells within the tumor microenvironment, suggesting that PD1 blockade might also impact Treg activity [15, 16].

Increasing evidence suggests that B-cells may also play a significant role in the pathogenesis of immune-related toxicities. Auto-antibodies known to contribute to non-ICI associated auto-immune conditions have been detected in a variety of ICI-associated conditions, including diabetes, thyroiditis/hypothyroidism as well as both myasthenia gravis and encephalitis [14]. Combination checkpoint blockade in particular was associated with distinct B-cell related changes in one study, including a decrease in overall circulating B-cell populations and an increase in plasmablasts. These changes were significantly more prevalent amongst those who developed iRAEs and correlated with both time to onset and severity of toxicity [17].

Genetic predisposition to auto-immune disease and/or sub-clinical autoimmunity have also been postulated to contribute to iRAEs. Pre-existing autoantibodies that might support this mechanism are rarely identified in pre-treatment samples but testing is limited to antibodies currently known to contribute to auto-immune disease [15]. Germline genetic variants that may confer an increased risk for immune-related toxicity are of significant interest. Assessment of genetic risk factors is limited by the large sample size required but will likely become more feasible as the population of patients treated with checkpoint inhibitors continues to increase.

Recent evidence also suggests that the intestinal microbiome may play a role in both response to ICI and risk of immune-related toxicity [18]. Alterations in the intestinal microbiome and dysregulation of gastrointestinal immunity have been noted amongst patients who experience immune-related colitis [19]. Although evidence of an underlying mechanistic link remains unclear, dysbiosis has been associated with auto-immune disorders involving both the gastrointestinal tract and various other organ systems and may play a similar role in some ICI-related conditions [20].

3. Neurologic immune-related adverse events associated with checkpoint inhibitors

3.1. CNS manifestations

3.1.1. Encephalopathy/encephalitis

Given the broad differential for neurologic symptoms in a patient with metastatic cancer, ICI-related encephalitis is often difficult to diagnose. Most patients described in case series present with altered mental status (AMS) [4]. Other common symptoms include headache, fever, weakness (generalized or focal), or other neurologic deficits, including seizures [4]. In addition to laboratory evaluation and CNS imaging to rule out metabolic or structural causes of AMS, suspicion for ICI-related encephalitis should prompt a lumbar puncture to evaluate for infectious and/or paraneoplastic causes. A lymphocytic pleocytosis has been described in the CSF in some reports of ICI-related encephalitis but is not particularly specific or sensitive [4]; we have also observed neutrophilic pleocytosis in some patients. MRI may reveal non-specific inflammatory changes such as temporal lobe enhancement but may also be unremarkable. Electroencephalography (EEG) may demonstrate seizures, temporal slowing, nonspecific patterns, or be normal. Limbic encephalitis and cerebellitis have also been described [21, 22]. The recent consensus clinical criteria for autoimmune encephalitis can be helpful in evaluating and diagnosing patients with ICI-related encephalitis [23].

As is observed in autoimmune encephalitis, CSF paraneoplastic and auto-immune antibody panels are often negative in patients with suspected ICI-related encephalitis, emphasizing the importance of clinical diagnosis rather than dependence on antibody detection [24]. In fact, most cases of encephalitis appear to be T cell mediated processes arising from the treatment (analogous to other irAEs) rather than paraneoplastic. However, at least one patient with metastatic melanoma developed anti-NMDA encephalitis after treatment with an ICI and various other onconeuronal antibodies have been identified among patients with other malignancies [25–27]. The GRIN2A gene, which encodes for a subunit of the NMDA receptor is frequently expressed and mutated in melanoma and could represent a shared antigen involved in paraneoplastic autoimmune encephalitis [28]. In addition, NMDA receptor complexes composed of NR1 and NR3B subunits have been observed in the nuclei of melanoma cells but not melanocytes [29]. Mouse models involving a neo-self-antigen expressed in both Purkinje cells and implanted tumor cells demonstrated significant cerebellar inflammation and T-cell infiltration after exposure to anti-CTLA4 antibodies. Cerebellar inflammation was present in 84% (27/32) of mice treated with anti-CTLA4 but was not found in any controls that were not exposed to anti-CTLA4 or in any non-neo-antigen expressing models that were treated with ICI [30].

In addition to unmasking new paraneoplastic processes, it is reasonable to assume that checkpoint inhibitors could potentiate or trigger paraneoplastic disorders that also occur sporadically. A recent study noted a significant increase in the frequency of anti-Ma2 associated paraneoplastic disorders, which generally result in limbic encephalitis, following the implementation of ICI [31]. Review of pharmacovigilance data found that encephalitis was most frequently reported in patients with lung cancer while several other neurotoxicities were more commonly noted in patients with melanoma [6]. There are several factors that may confound this observation, but the association between lung cancer (particularly SCLC) and other tumors of neuroectodermal origin and paraneoplastic CNS disorders is notable [32]. Paraneoplastic syndromes are also potentially more varied than previously appreciated, raising concern for under recognition [33]. Extensive diagnostic testing for known paraneoplastic antibodies has been negative in most cases, suggesting an underlying pathophysiology that is more likely distinct to ICI. The presence of one unifying pathologic driver in ICI-associated encephalitis is also unclear and further characterization may reveal several distinct subtypes with unique molecular and clinical features. Ultimately, the relationship between ICI-encephalitis and paraneoplastic syndromes remains to be elucidated.

Immune-mediated encephalitis can also occur in the post-infectious setting, usually following a non-specific viral infection [34]. The distinction between ICI-related and post-infectious encephalitis may be difficult to ascertain via history given the vague and relatively minor symptoms generally associated with preceding infectious causes. Some series have described a prodrome of URI or flu-like symptoms preceding the onset of ICI-associated encephalitis [25, 35]. Molecular profiling of one case of fatal ICI-induced encephalitis identified a clonal population of activated CD4+ memory cytotoxic T cells in addition to EBV+ lymphocytes and EBV specific T cell receptors in affected brain tissue, suggesting that viral infection or reactivation may have triggered the event (which was fueled by the anti-PD-1) [36]. This is supported by the association of primary HSV encephalitis with a variety of secondary autoimmune encephalitides, including anti-NMDA receptor encephalitis [37]. Cytotoxic CD4 cells play a role in anti-viral and anti-tumor response and have also been shown to contribute significantly to auto-immune disease [38, 39]. Further characterization of how and why these cells may drive inflammation in iRAE and chronic auto-immune disease is warranted and may provide insight into underlying triggers and molecular drivers.

Some authors have postulated that disruption of the blood brain barrier in patients with CNS metastases who undergo stereotactic radiosurgery (SRS) may increase the risk for immune-related CNS complications [25]. A recent report described a patient who was started on ipilimumab and nivolumab two weeks after undergoing SRS and developed a large ring-enhancing lesion at the surgical site with associated acute neurologic deficits. Although radiographically concerning for an intra-cranial abscess, this was felt more likely to be an irAE and the patient was started on dexamethasone with complete resolution of neurologic deficits and associated radiographic abnormalities [40].

ICI-related encephalitis is associated with significant morbidity and has a relatively high mortality rate [21, 41]. While some neurologic irAEs may be managed in the outpatient setting, patients with suspected ICI-associated encephalitis warrant inpatient monitoring and immediate high dose IV steroids [42]. Guidelines recommend an initial trial of 1–2 mg/kg of methylprednisolone per day with escalation to pulse dosing and/or IVIG for severe or progressive symptoms [42]. If patients fail to improve rapidly (within 2–3 days), therapy should be quickly escalated. Rituximab can be considered in patients with features typical for autoimmune encephalitis, such as positive CSF autoantibody testing, new onset seizures, or characteristic T2 hyperintensities of the medial temporal lobes or in those with no improvement despite 1–2 weeks of appropriate therapy [25, 42].

Many patients require prolonged hospitalization and rehabilitation but generally show some improvement with appropriate management [4, 25, 43, 44]. The degree of symptom improvement is not specified in most series but complete or near-complete recovery is possible with prompt diagnosis and treatment [44]. Data regarding outcomes with the use of agents other than corticosteroids is limited, particularly in regard to rituximab [25, 45]. Encephalitis associated fatality has been reported in several cases, potentially due to delayed diagnosis and/or failure to initiate appropriate treatment prior to clinical decompensation [4, 41, 44]. Checkpoint inhibitors are generally permanently discontinued in the setting of ICI-associated encephalitis given the associated morbidity and mortality risk [42].

3.1.2. Aseptic meningitis

ICI-associated meningitis has been noted in several different series yet remains poorly categorized. The distinction between meningitis and encephalitis is not always straightforward and may be particularly difficult in a patient with metastatic cancer or who is presenting with new seizures and cognitive impairment. While some series describe an ICI-associated meningoencephalitis, overlap is infrequently reported [6, 46]. Several features of ICI-associated meningitis are notably distinct from those associated with encephalitis and it remains unclear whether these are two distinct disease processes or a phenotypic spectrum with a shared mechanism.

While encephalitis has been noted more frequently in association with anti-PD1 agents, meningitis has primarily been described amongst patients treated with ipilimumab [6, 47]. Meningitis may also be more likely to present with additional non-neurologic iRAEs [6]. Hypophysitis, which is also more commonly associated with ipilimumab, is a relatively frequent co-diagnosis, although this raises the question of diagnostic accuracy, since hypophysitis may present with headaches and non-specific neurologic symptoms [6, 48, 49]. Guidelines recommend MRI with pituitary protocol as well as AM cortisol to rule out adrenal insufficiency in all patients with suspected aseptic meningitis [42]. It may be difficult to determine which process is driving symptoms in patients who also develop adrenal insufficiency although CSF studies may provide some insight [50]. This distinction may not be clinically relevant as patients with ICI-associated meningitis often demonstrate complete recovery following drug withdrawal and steroid initiation [47]. Patients should be evaluated and monitored in the inpatient setting in most cases but steroids can often be given orally with a taper over 4 weeks following symptom improvement [42]. NCCN guidelines recommend consideration of ICI re-challenge in patients who experience mild to moderate grade meningitis with complete symptom resolution prior to re-introduction [42].

3.1.3. Multiple sclerosis

Despite many shared pathologic features, multiple sclerosis is infrequently reported amongst patients treated with checkpoint inhibitors. Patients with pre-existing MS may experience flares when exposed to ICI but de novo CNS demyelination appears relatively uncommon [51, 52]. One patient with radiologically isolated MS progressed to clinically definite disease following exposure to ipilimumab [28]. When CNS demyelination is noted in association with ICI, radiographic features are typically inconsistent with multiple sclerosis [53, 54]. Other demyelinating disorders attributed to checkpoint blockade include acute demyelinating encephalitis and subacute tumefactive demyelination [53, 54]. CD4+ T-cells isolated from one patient who developed new CNS demyelination after receiving ipilimumab demonstrated a similar response to myelin antigen exposure as those from patients with known multiple sclerosis [55].

3.2. Peripheral nervous system manifestations

3.2.1. Myasthenia gravis

ICI-induced myasthenia gravis (iMG) is an increasingly recognized and feared complication of checkpoint inhibitor therapy. While some cases involve exacerbation of known myasthenia, most represent a de novo syndrome [56].. Antibodies to the acetylcholine receptor are identified in ~85% of patients with generalized myasthenia gravis yet are more frequently negative in patients with ICI-associated MG [57, 58]. When AChR antibodies are detected, titers are generally much lower than those seen in ICI naive patients [59, 60]. Of note, of those patients with generalized MG who are seronegative for AChR antibodies, 50–70% are seropositive for antibodies to muscle specific kinase (MUSK). To our knowledge, anti-MUSK antibodies have yet to be detected amongst patients diagnosed with iMG but further evaluation is warranted.

Unlike traditional myasthenia gravis, patients with iMG frequently exhibit symptoms suggestive of life-threatening bulbar or respiratory muscle involvement at the time of diagnosis [44]. While ocular symptoms such as ptosis and diplopia are often noted in both settings, dysphagia, dysarthria and dyspnea are much more frequent presenting symptoms of iMG [58, 60]. Symptoms generally progress rapidly with frequent decompensation to myasthenic crisis requiring respiratory support [5, 58]. Almost all reported patients with iMG required hospitalization with 40–50% requiring mechanical ventilation [5, 58]. Among a retrospective cohort of 65 patients diagnosed with iMG, the median time from symptom onset to respiratory failure requiring intubation was only 7 days [58]. This is notably distinct from non-ICI associated myasthenia gravis which has an estimated 15–20% lifetime risk of myasthenic crisis with about one-fifth of patients presenting in myasthenic crisis at the time of diagnosis [61, 62]. The prevalence of milder cases of ICI-associated myasthenia may, however, be higher than suggested as non-specific symptoms such as generalized weakness and fatigue may be difficult to recognize among patients with advanced malignancy.

ICI-associated myasthenia gravis also seems to represent a distinct phenotype amongst neurologic irAEs, which are otherwise in many ways similar to immune toxicities affecting other organ systems. Myasthenia seems to occur earlier than other neurotoxicities with a median time to onset of 29 days compared with a median of 61–80 days for other neurologic irAEs based on pharmacovigilance data [6]. Increased awareness of this unique toxicity and a more characteristic and/or severe presentation may influence time to diagnosis. MG is also one of the most frequently reported neurologic iRAEs although these factors may also limit extrapolation regarding comparative incidence.

Unlike other neurologic toxicities, which are often reported in isolation, myasthenia gravis also frequently presents with concurrent myositis and/or myocarditis [6, 58]. Myositis has been reported in up to one-third of cases [58]. Several reports of ICI-associated MG describe myalgia and/or elevated creatine phosphokinase (CPK) levels without a formal diagnosis of myositis, suggesting that the true incidence may be higher [6]. Among 49 of 65 patients with iMG who were screening for concurrent myositis, CPK levels were elevated in 41 (84%) cases with a median CPK level of 2,638 IU/L and a range from 418 to 19,794 [58]. However, only 24 of these 65 patients were formally diagnosed with concurrent myositis [58]. Skeletal muscle biopsy to confirm a diagnosis of myositis was rarely performed but did demonstrate characteristic inflammatory infiltrates in a majority of tested cases (5/7, 71%) [58].

Reports of ICI-associated myositis generally describe rapid onset of severe and often life-threatening symptoms that may be difficult to distinguish from myasthenia gravis in certain contexts. ICI-associated myopathy with bulbar or ocular involvement that may mimic symptoms of myasthenia gravis has also been reported [63]. It is unclear whether the CPK elevation frequently noted in iMG reflects a concurrent process or possibly a primary myositis in some cases. Electrodiagnostic testing to evaluate for features of MG may be useful but can be difficult to interpret in the setting of suspected muscle involvement [58, 59]. Characteristic EMG features of idiopathic myasthenia gravis are identified in less than half of patients with iMG [57, 58]. One series noted a relatively homogenous presentation of ICI-related myositis with frequent axial limb-girdle and oculomotor weakness in addition to myalgia in most patients [64]. Further characterization of the clinical, diagnostic and pathologic features of suspected myositis and myasthenia gravis may help distinguish between the two conditions as well as any overlap syndrome.

Concurrent myocarditis has been reported in about 8% of patients with ICI-associated myasthenia gravis [6, 58]. As with myositis, the frequency of concurrent diagnoses is likely influenced by growing awareness of this association and incidence is difficult to estimate. Review of pharmacovigilance data identified several reports of ICI-associated MG that also noted cardiac arrythmia and/or myocardial infarction but did not report myocarditis [6]. ICI-associated myocarditis carries an estimated 20–50% mortality and often progresses rapidly to involve life-threatening arrythmia and/or heart failure [65, 66]. Patients who present with myasthenia-like symptoms should thus be screened and monitored closely for signs of myocardial involvement [67]. We would recommend an ECG and troponin testing to assess for this diagnosis, as these appear to be the most sensitive mechanisms of ruling out myocarditis [68].

Outcomes appear consistently worse in most series among patients with overlapping myositis and/or myocarditis compared with those with isolated iMG [5, 6, 58]. Patients with more severe and/or systemic symptoms may be more likely to be evaluated for concurrent conditions but it is also reasonable to assume that clinical or diagnostic evidence of multi-organ system involvement portends a worse prognosis. First line treatment for suspected ICI-associated MG includes ICI discontinuation and either initiation of low dose corticosteroids with slow titration or methylprednisolone with the addition of PLEX and/or IVIG in severe cases [42]. Steroids are thought to result in a transient worsening of symptoms in some patients with myasthenia gravis who present with myasthenic crisis and are generally not given as monotherapy in this setting [69–71]. The decision regarding initial immunosuppression management is driven by a patient’s pulmonary function and clinical trajectory. A recent review noted much higher rates of symptom improvement amongst patients treated with IVIG or PLEX in the first line (with or without simultaneous corticosteroids; 18/19, 95%) compared with those treated with corticosteroids alone (24/38, 63%) speaking to the tenuous clinical status of patients presenting with iMG [58]. The use of first line IVIG or PLEX was triggered by early onset of severe respiratory or bulbar symptoms in almost all cases (17/19). Notably, none of the 12 patients who developed respiratory failure after initial treatment with steroid monotherapy had any improvement in symptoms with IVIG or PLEX [58]. Immediate initiation of PLEX and/or IVIG is often indicated in non-ICI associated myasthenic crisis given the need for rapid symptom control and should also be considered ICI-associated cases [70]. NCCN guidelines currently recommend addition of IVIG/PLEX in the second line for patients with ICI-associated MG and grade 3–4 symptoms that fail to improve with corticosteroids with consideration of immediate initiation in severe cases. Based on decades of clinical experience with myasthenia gravis in the setting of a paucity of evidence to guide checkpoint inhibitor associated MG, clinicians should have a low threshold to consider initiating IVIG or PLEX in patients with rapidly progressive or severe symptoms.

Fatality due to ICI-associated MG has been reported in ~20% of cases [6, 58]. Complete symptom resolution is relatively rare and patients are often maintained on prolonged steroid tapers following hospital discharge [58, 60]. Limited data suggests a variable response to cholinesterase inhibition in patients with iMG [58, 60]. Current guidelines recommend a trial of pyridostigmine with or without concomitant steroid therapy in patients with Myasthenia Gravis Foundation of America (MGFA) severity class I (ocular symptoms only) or class II symptoms (mild generalized weakness) [42]. Recommendations are particularly difficult in this setting given evidence is currently limited to case series, which generally describe only more severe or fulminant cases. Rare instances of iMG remission without the use of immunosuppressive agents have been noted but only amongst patients with mild symptoms limited to ocular or facial muscle involvement [72, 73]. Monotherapy with cholinesterase inhibition is usually avoided in non-ICI-associated generalized MG and should be used with caution in those with iMG.

Data related to outcomes and optimal management of patients with more severe symptoms that improve with therapy and ICI discontinuation is similarly sparse. ICI re-challenge can be considered in patients with mild to moderate symptoms (MGFA class I or II) that respond to steroids but is generally avoided in those with more severe symptoms including those that require hospitalization [42]. Longer term follow-up and re-evaluation of patients who develop ICI-associated myasthenia gravis is warranted to inform optimal management and may provide some insight into underlying pathophysiology.

Given associated severity and mortality and a notable overlap with myositis and myocarditis, the immunobiologic mechanisms behind iMG are of particular interest yet remain largely unknown. While some clinical features suggest that ICI-associated MG may be more fulminant than non-ICI associated MG, it is unclear whether this is due to differences in underlying pathophysiology or related to the demographics of affected patients and/or under-reporting of more mild cases of iMG. As with other iRAEs, one proposed mechanism involves unmasking of a pre-existing subclinical autoimmune disorder or tendency with checkpoint inhibition. Several recent series described elevated pre-treatment anti-AchR antibodies among patients who subsequently developed ICI-associated myasthenia gravis with increasing titers following symptom onset [58, 74, 75]. The presence of AchR antibodies, even at very low titers, is thought to be highly specific for MG. Although modern assays may be more sensitive, antibodies were almost exclusively undetectable amongst healthy controls at the time of assay development [76]. Pre-existing AchR antibodies may be paraneoplastic or may reflect underlying genetic or environmental risk factors and/or long-standing subclinical disease.

Although most cases in non-ICI exposed patients are idiopathic, paraneoplastic myasthenia gravis can be triggered by an underlying thymoma. Patients with thymoma-associated MG frequently exhibit severe symptoms and/or myasthenic crisis upon initial presentation. Concurrent myositis and/or myocarditis are also most frequently noted in patients with thymoma-associated MG [77]. Thymic abnormalities including thymic hyperplasia are prevalent in many types of idiopathic MG and characteristics that seem more specific to thymoma-associated MG are not well understood [78]. The presence of “myoid” cells that express AchR within the thymus and/or abnormalities involving central tolerance, including aberrant production of functional regulatory T cells are possible mechanisms to explain the role of the thymus in the pathogenesis of myasthenia gravis [78]. Further characterization of molecular or pathologic features common to checkpoint inhibitor toxicity and thymic abnormalities in MG, particularly associated with thymoma, may be useful. The association between thymoma-associated MG and myocarditis/myositis also warrants further evaluation. Several autoantibodies to proteins highly expressed in skeletal muscle have been identified in these patients but are of uncertain significance [79].

3.2.2. Peripheral neuropathy

Checkpoint inhibitors are associated with a relatively wide spectrum of neuropathic toxicity. Patients may present with pain, paresthesia or weakness due to involvement of sensory, motor or autonomic peripheral nerves. Cranial neuropathies and inflammatory polyradiculopathies, including Guillain Barre syndrome (GBS), are increasingly recognized [80, 81]. Guillain-Barre syndrome triggered by checkpoint inhibition is generally similar to non-ICI associated GBS in terms of presentation and clinical course [45]. Like non-ICI associated GBS, most cases can be categorized as an acute inflammatory demyelinating polyneuropathy (AIDP) although rare variants, including Miller-Fisher syndrome, have also been reported in association with ICI [82]. Albuminocytologic dissociation, a characteristic feature of both GBS and chronic inflammatory demyelinating polyneuropathy (CIDP), is often noted [81].

Chronic demyelinating polyneuropathies thought to be triggered by checkpoint inhibition have also been noted but are much less frequently reported than GBS [56]. CIDP may occasionally appear similarly to GBS in acute and early phases of the disease yet is distinguished by the time course of symptoms and response to corticosteroids. Rapid symptom onset suggestive of an acute demyelinating polyneuropathy has been reported in several cases that were ultimately diagnosed as ICI-associated CIDP following decompensation several weeks after initial improvement [56, 83]. CIDP is generally associated with a more indolent disease course with time to maximal weakness of at least 8 weeks. Symptoms may continue to progress or may demonstrate a relapsing/remitting course due to segmental demyelination and remyelination and associated changes in involved neurons. Although much is unknown regarding the pathogenesis of CIDP, both humoral and cell-mediated mechanisms are thought to be involved.

Some studies have suggested that patients with melanoma may be at higher risk for ICI-associated demyelinating polyneuropathy due to epitopes shared by melanocytes and Schwann cells [56, 84–86]. Although rare, melanoma is the solid tumor most frequently reported in malignancy-associated CIDP [84]. AIDP has also been attributed to BRAF inhibitor therapy in several patients with metastatic melanoma [87–89]. Although one of these patients was previously treated with an anti-PD1 agent that may have contributed, at least two cases have occurred following BRAF/MEK inhibitor initiation in treatment naive patients. BRAF inhibitors have many known immunomodulatory effects in metastatic melanoma and could theoretically potentiate an anti-tumor response that also aberrantly targets neural tissue [89–91].

While corticosteroids are of little benefit in GBS, impact in the setting of ICI exposure remains unclear. Current guidelines thus recommend consideration of a trial of corticosteroid therapy in conjunction with IVIG or plasmapheresis in patients with ICI-related GBS [42]. Fatality has been reported in patients with ICI-associated GBS, attributed to respiratory muscle involvement as well as to fulminant enteric neuropathy in two cases [56]. Non-ICI associated CIDP is often responsive to corticosteroids, which are indicated in the first line setting for ICI-associated CIDP [93]. Patients should be monitored closely for recurrence with weaning of therapy with a low threshold to start or increase immunosuppression due to new neurologic symptoms.

Electrodiagnostic studies in patients with ICI-associated neuropathies are frequently suggestive of an underlying demyelinating process although acute axonal processes have also been reported [79, 95]. Several cases of acute painful sensory neuropathy have been reported in association with ICI. Sensory nerve conduction studies performed in one series were suggestive of predominant axonal involvement in many of these cases [95]. While only 50% of patients with demyelinating neuropathy reported associated pain, all seven patients diagnosed with axonal neuropathy indicated that neuropathic pain was the primary symptom prompting initial presentation [95]. ICI-related acute painful neuropathy has also been attributed to sensory neuronopathies in some cases, characterized by predominant involvement of dorsal root ganglia [80]. Chemotherapy-induced peripheral neuropathy (CIPN) is also often characterized by involvement of dorsal root ganglia and/or axons of sensory nerves with pain as a predominant symptom. While much remains unknown about the pathogenesis of CIPN, most cases are attributed to direct neurotoxic effects on neurons by chemotherapeutic agents, which is unlikely with ICI. It remains unclear whether there is any synergistic effect among patients exposed to both ICI and chemotherapy [96]. Acute axonal neuropathies due to ICI exposure, particularly with predominant sensory involvement, are not uncommon and warrant further investigation.

Facial nerve palsy and trigeminal neuralgia are the most frequently noted cranial neuropathies attributed to checkpoint inhibitors [80, 81]. Small fiber or autonomic neuropathies resulting in orthostasis, anhidrosis, gastrointestinal dysmotility and/or urinary retention have also been reported [80, 81, 94]. Most peripheral neuropathies improve significantly with ICI discontinuation and/or corticosteroids although we have observed long-term persistence in some cases of painful sensory neuropathy [80]. ICI re-challenge can be considered in patients who experience grade 1–2 neuropathy symptoms that improve to grade 1 or below when treatment is withheld. If symptoms are reasonably well-controlled, re-challenge can also be considered in patients with persistent isolated painful sensory neuropathies [42]. Re-challenge was associated with an increased risk of recurrent neurologic iRAEs in one study yet this observation is limited by small sample size and warrants further evaluation [80].

4. Conclusion

ICI have transformed treatment paradigms in many types of cancer, although autoinflammatory toxicities may lead to morbidity and even mortality in some patients. Neurologic toxicities occur in 1–5% of treated patients and encompass a diverse range of events, most commonly meningo-encephalitis, myasthenia gravis-like syndrome, and peripheral neuropathies. Treatment of these events revolves around removing the source (ICI), suppressing overactive immune cells (with high-dose steroids and other immunomodulators), and supportive care.

5. Expert opinion

Neurologic toxicities present some of the most diagnostically challenging and clinically diverse spectrum of irAEs. Moreover, given the lack of familiarity by many oncologists, this set of events produces a significant amount of anxiety and concern. Thus, multidisciplinary management by oncology and neurology is critical.

Encephalitis requires a high index of suspicion and diagnostic acumen. The differential diagnosis should remain broad and include infectious (viral or bacterial) and neoplastic etiologies, and initial treatment must be broad until a diagnosis is established. Initial mechanistic studies by our group have been notable for a potential viral role in a single case, although it is unclear whether this is a more generalizable mechanism. One can speculate that a viral infection or reactivation would produce CNS inflammation, which would be countered by expression of PD-L1 on the inflamed cells. Blockade of PD-1 or PD-L1 however, would allow inflammation to persist unchecked, further worsening the clinical presentation. Underlying paraneoplastic processes may be involved in ICI-associated encephalitis in some cases with propagation and/or potentiation of a host-mediated anti-tumor response that also targets neuronal antigens.

Myasthenia gravis is also a highly concerning clinical manifestation of ICI. The overlap between myasthenia, myositis, and myocarditis is a symptom cluster not otherwise observed with ICI. Further, the subset of what clinically appears to be iMG but is associated with muscle inflammation and high CPK levels appears to be a distinct and perhaps completely novel clinical syndrome. Respiratory muscle involvement may lead to severe morbidity and mortality. Further, the specter of concurrent myocarditis (the toxicity with the highest mortality rate) makes close monitoring and aggressive treatment even more critical.

Finally, peripheral neuropathies may present in diverse fashion, ranging from Guillain Barre Syndrome to mononeuropathy of a single cranial nerve. While usually reversible, we have observed highly symptomatic and persistent cases that produce long-term sequelae. Appropriate neuro-imaging and neurology involvement is critical in these cases.

Article highlights.

• Immune-related adverse events (irAEs) involving the nervous system are an important toxicity to consider when evaluating patients who have been treated with immune checkpoint inhibitors (ICI).

• Neurologic irAEs occur in an estimated in 1–5% of patients treated with ICI and involve a wide spectrum of clinical and pathologic disorders of both the central and peripheral nervous systems. Meningo-encephalitis, myasthenia gravis and peripheral neuropathy have been most frequently described.

• ICI-associated encephalitis should be considered in patients who develop altered mental status or other acute neurologic deficits following ICI exposure. Work-up is often unrevealing and this is often a diagnosis of exclusion but is critical to recognize to prevent long-term neurologic damage and/or fatality.

• ICI-associated myasthenia gravis has gained increasing attention due to a notable overlap with myositis and/or myocarditis in some cases and a frequently fulminant life-threatening presentation requiring close monitoring and/or ventilator support.

• ICI-associated peripheral neuropathies involve a wide spectrum of clinical manifestations with involvement of sensory, motor and/or autonomic nerves. Cranial neuropathies and inflammatory polyradiculopathies, such as Guillain Barre syndrome, are most frequently reported.

• Treatment of neurologic irAEs is similar to those involving other organ systems and includes withholding or permanently discontinuing ICI and immunosuppressive therapy, usually in the form of corticosteroids. While most patients improve with these measures, long-term follow-up data is limited and the degree of neurologic recovery that can be expected is unknown.