Article Text

Download PDFPDF

Paraneoplastic neurological syndromes: a practical approach to diagnosis and management
  1. Sophie Binks1,2,
  2. Christopher Uy1,3,
  3. Jerome Honnorat4,5,
  4. Sarosh R Irani1,2
  1. 1 Oxford Autoimmune Neurology Group, Nuffield Department of Clinical Neurosciences, Oxford University, Oxford, UK
  2. 2 Department of Neurology, Oxford University Hospitals NHS Foundation Trust, Oxford, UK
  3. 3 Department of Medicine (Division of Neurology), University of British Columbia, Vancouver, British Columbia, Canada
  4. 4 French Reference Centre on Paraneoplastic Neurological Syndromes and Autoimmune Encephalitis, Hospices Civils de Lyon, Hopital Neurologique, Lyon, France
  5. 5 SynatAc Team, Institute NeuroMyoGene INSERM U1217/CNRS UMR 5310, Universite de Lyon, Universit Claude Bernard Lyon 1, Lyon, France
  1. Correspondence to Prof Sarosh R Irani, Oxford Autoimmune Neurology Group, Nuffield Department of Clinical Neurosciences, Oxford University, Oxford OX3 9DS, UK; sarosh.irani{at}


Paraneoplastic neurological syndromes (PNS) are the immune-mediated effects of a remote cancer and are characterised by an autoantibody response against antigens expressed by the tumour. Classically, well-characterised ‘onconeuronal’ antibodies target intracellular antigens and hence cannot access their antigens across intact cell membranes. The pathogenic mediators are likely to be neuronal-specific T cells. There is a variable response to immunotherapies and the clinical syndrome helps to direct the search for a specific set of tumours. By contrast, many newly emerging autoantibodies with oncological associations target cell surface epitopes and can exert direct pathogenic effects on both the central and peripheral nervous systems. Patients with these cell-surface directed autoantibodies often clearly respond to immunotherapies. Overall, the clinical, serological and oncological features in an individual patient help to determine the clinical relevance of the syndrome and hence guide its management. We summarise current knowledge and a practical approach to the investigation, diagnosis, treatment and outcomes of patients with suspected PNS.

  • immunology
  • tumours
  • neuroimmunology
  • paraneoplastic syndrome
  • clinical neurology

Statistics from


Paraneoplastic neurological syndromes (PNS) describe the remote neurological immune-mediated consequences of a systemic cancer. They affect ~1:300 patients with tumours, yet population-level epidemiology suggests an incidence rate of only between ~1 and 8/100 000 person-years,1 2 indicating ongoing under-recognition. Distinctive clinical and serological features (tables 1–3) typically direct the search for a tumour, which is subsequently detected in around 65% of cases. Rarely this tumour emerges only months or years after the neurological syndrome, demanding ongoing clinical vigilance.1 3 A key feature of PNS is that the cancer triggers the immune response, and so it should express the autoantigen to which the immune response (including an autoantibody) is directed: this ensures a direct biological link between the cancer and PNS.

Table 1

Demographic, tumour, clinical, treatment response and prognosis in onconeuronal antibody-associated syndromes

Table 2

Demographic, tumour, clinical, treatment response and prognosis in cell surface antibody-mediated syndromes where an underlying tumour is often detected in a minority of cases

Table 3

Paraclinical features of onconeuronal and surface antibody-associated syndromes

Expert guidelines in 2021 have redefined aspects of these disorders in light of the description of novel antibodies and the most robust emergent clinical–serological–oncological associations.4 Among other benefits, these observed relationships avoid the spurious attribution of common cancers to neurological presentations with an alternative explanation, and hence encourage accurate diagnosis and prognostication. In this classification, clinical presentations and associated autoantibodies may be broadly defined as ‘high’ or ‘intermediate’ risk of paraneoplastic aetiology (box 1).4 5 High-risk clinical presentations are reflected by epidemiological studies which consistently identify autoantibody patterns with subacute cerebellar degeneration, encephalomyelitis, limbic encephalitis and sensory neuronopathy as the leading PNS in European cohorts.1–3 The intermediate-risk groups show recognised, but less reliable, clinical–serological associations.

Box 1

Classic paraneoplastic presentations4 5

A. High risk, frequently associated with tumours

Central nervous system

  • Encephalomyelitis.

  • Limbic encephalitis.

  • Subacute cerebellar degeneration.

  • Opsoclonus–myoclonus.

Peripheral nervous system and neuromuscular junction/muscle

  • Subacute sensory neuronopathy.

  • Chronic gastric pseudo-obstruction.

  • Lambert-Eaton myasthenic syndrome.

  • Dermatomyositis.

B. Intermediate risk, less frequently associated with tumours

Central nervous system

  • Encephalitis.

  • Brainstem encephalitis.

  • Isolated myelopathy.

Peripheral nervous system and neuromuscular junction/muscle

  • Polyradiculopathy.

  • Stiff-person syndrome.

  • Myasthenia gravis.

Paraneoplastic antibodies in context

Traditionally, PNS have been associated with ‘well-defined onconeuronal’ antibodies, which almost always target intracellular proteins and hence show limited direct pathogenic significance. Nevertheless, their presence often indicates a robust (sometimes nearly 100%) association with a tumour in addition to distinct clinical links (table 1), for example, Yo antibodies and cerebellar degeneration.6 Such associations clinically guide cancer identification and can expedite and focus treatments.

More recently, an explosion of scientific discovery in the field of cell surface-directed neuroglial antibodies, which show clear pathogenic potential, has yielded several further clinical–serological associations amongst PNS. However, the frequency of cancer associated with these newer autoantibodies varies substantially according to the antigenic target, and the age and sex of the patients. Examples include N-methyl-D-aspartate receptor (NMDAR)-autoantibodies and ovarian teratomas,7 which occur in young women between ~18 and 35 years of age, but very rarely in young children or men; α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor-antibodies which are associated with ~50% rate of cancers, especially small cell lung cancer (SCLC and thymoma), but mainly in the older patients8; and γ-aminobutyric acid (GABA)B receptor-autoantibodies which associate with a ~50% rate of SCLC in patients more than 50 years of age (table 1).9 Also, several surface neuroglial antibodies are associated with very low overall rates of cancer.10

By contrast to intracellular-directed antibodies, those targeting cell-surface proteins—which have access to their native targets in the blood/cerebrospinal fluid (CSF)—are considered directly pathogenic as they bind in vivo and can trigger deleterious effects, including complement fixation, receptor internalisation and induce direct functional modifications to their target proteins.11 12 We have detailed these causative antibodies in our sister review.13 Also, herein, we will neither focus on autoantibodies in inflammatory myopathies, which have recently been reviewed in this journal,14 nor the potential triggering of PNS after checkpoint inhibitor drugs,15 as these likely show different underlying mechanisms to traditional PNS.16 Myasthenia gravis may be associated with an underlying tumour, typically a thymoma, but also is not further considered herein due to its absence from the updated 2021 guidelines.4


Role of onconeuronal antibodies

The exact role of intracellular-directed antibodies remains unresolved, but they are unlikely to be direct mediators of disease. Although uptake and pathological effects of Hu-antibodies by Purkinje cells have been shown in vitro,17 their infusion into various in vivo models does not consistently replicate disease despite achieving high antibody titres in the experimental animals.18 Experiments with Ma2-antibodies yielded similar negative results.19 However, some antibodies classically denoted as ‘onconeuronal’ do recognise extracellular domains of neuroglial proteins and thus could have a direct role in causation: for example, Tr antibodies are closely associated with Hodgkin’s lymphoma and have been recognised to target the extracellular portion of the delta and notch-like epidermal growth factor-related receptor (DNER).20

The relatively high rate of many of these antibodies in patients with tumours but no accompanying neurological syndrome, for example, ~16% of patients with SCLC with Hu- 21 or Zic4-antibodies,22 and the frequent co-occurrence of more than one paraneoplastic antibody in an individual,3 cautions against interpretation of the antibody result in isolation. It may also suggest that the polyclonal immune response protects against tumour growth. Indeed, tumours may be more indolent in patients with a PNS.23

In PNS, the likely origin of immunisation is the tumour itself, given immunohistochemical evidence of relevant antigen expression in cancers including SCLC,24 testicular seminoma25 and ovarian carcinoma.26 The accompanying peripheral immune response may translocate to the central nervous system (CNS) in some patients with, overall, CSF studies indicating intrathecal synthesis of antibodies6 22 25 in CNS-predominant syndromes (vs absent intrathecal synthesis in those with PNS-only involvement).27 This suggests an influx of lymphocytes to the CSF is a key triggering event for CNS syndromes, a finding that aligns with clonal CD8+ T cells and restricted T-cell receptor repertoires in postmortem brain pathology.28 It is likely that these cell types are key disease effectors.

Genetic contributions

Major histocompatibility complex (MHC) molecules are considered essential to antigen presentation, T-cell activation and autoantibody generation. Early—although limited—immunohistochemical observations suggested that tumours of patients with Hu-antibody PNS might have increased expression of MHC proteins, when compared with tumours in patients without autoantibodies.29 More recently, it was demonstrated that patients with Hu-antibody with PNS were significantly more likely to carry the class II molecules HLA-DQ2 and HLA-DR3 than ethnically-matched healthy controls.30 By contrast, with Yo-antibodies, a risk haplotype DQA1*01:03-DQB1*06:03-DRB1*13:01, was reported as tumour-specific and observed mainly in patients with ovarian but not breast cancer, whereas there is a protective effect of DRB1*04:01 across all patients.31

A different genetic predisposition has been suggested by somatic mutations in antigenic proteins. Taking Yo-antibody cerebellar degeneration, it has been shown that associated ovarian tumours, but not those found in patients without Yo-antibodies, expressed numerous genetic lesions in the antigenic cerebellar degeneration-related protein 2-like (or Yo) protein. It is plausible that these create neoantigens that drive the disease.26 A differing mechanism has been elucidated in patients harbouring NMDAR-autoantibodies and ovarian teratoma. Pieces (‘explants’) of tissue from these—usually benign—tumours, and isolated intrateratoma B cells, were able to secrete NMDAR-autoantibodies in vitro. Furthermore, tumour immunohistochemistry revealed structures housing T cells, B cells and the NR1 subunit—the key autoantibody epitope.32 In this scenario, the tumours may act as ectopic lymphoid organs and initiate the directly pathogenic autoantibody response.

Taken together, these early observations support a local tumour response, facilitated by genetic and other as-yet confirmed factors in at-risk patients, which gains traction in the CNS with either autoantibodies or cytotoxic T cells as the key pathogenic effector. Future efforts to map steps along this pathogenic pathway systematically will offer clearer insights into disease pathogenesis and biology.

Autoantibody testing: practical considerations

Commercial testing for specific onconeuronal antibodies is typically performed using immunodot or blot methods.33 34 With these techniques, the antigens of interest are immobilised within a fixed band on a nitrocellulose paper, patient serum or CSF is applied, and if ‘antibody-positive’, an intensity is visualised at the known location of the antigen (figure 1). Several recent studies have cast doubt on the accuracy of this approach, consistently showing only ~40% of line blot-positive results are verified by more robust methods such as immunofluorescence or cell-based assays.33 34 Positive predictive values have been as low as 39%.35 While bleak, these overall results mask some even more worrying findings for individual antigens—for example, Déchelotte et al almost never confirmed positive line dots for amphiphysin, Ma1- and Yo-antibodies, whereas for Hu-antibodies, confirmation by other methods was observed in up to ~88%.34 A recent investigation into patients with collapsin response-mediator protein-5 (CRMP5)/CV2-antibodies showed commercial kits failed to detect ~8% of patient sera found positive by immunofluorescence or cell-based assay (figure 1).36 Two of these samples were from individuals with SCLC, showing a crucial diagnosis could have been missed.

Figure 1

Detection of paraneoplastic antigens. Use of commercial line or dot blots miss clinically relevant reactivities identified by immunohistochemistry or cell-based assay. (A) Immunohistochemistry on sections of paraformaldehyde-perfused rat cerebellum incubated with a CV2-positive serum (dilution 1:5000) that did not react with commercial immunoblots despite robust immunoreactivity with the cytoplasm and processes of oligodendrocytes in the GCL and WM. Scale bar=25 µm. (B) Line blot neuronal antigen profile. Antigens by columns: a, titin; b, SOX1; c, recoverin; d, Hu; e, Yo; f, Ri; g, Ma2; h, CV2; and i, amphiphysin. Strips were incubated with the sera of patients with CV2 antibodies determined by immunohistochemistry, two positives and one negative control. Serum 3 showed very mild immunoreactivity at 1/200 dilution but was negative at the manufacturer’s recommended dilution (1/1000). This serum was also immunoreactive with amphiphysin. (C) Serum of patients with CV2 antibodies by immunohistochemistry, diluted at 1/2000 and 1/200, did not react with blot strips. The localisation of the appropriate CV2 band detected with the positive control (+) is indicated with an arrowhead. Serum 3 showed a weak immunoreactivity with SOX1 and amphiphysin. (D) HEK293 cells transfected to express GFP-tagged CRMP5 were incubated with serum of patients with CV2 antibodies that did not react with commercial immunoblots or control (−) serum. Patient’s serum, but not control serum, stained the cells (red) that specifically express CRMP5 (seen in green). Both reactivities are shown merged in the bottom row (yellow). Nuclei counterstained with 4′,6-diamidino-2-phenylindole (DAPI) . Scale bar=20 µm. Reproduced from Sabater et al 36 with permission from Elsevier. GCL, granular cell layer; GFP, green fluorescent protein; WM, white matter.

Therefore, expert consensus suggests that a positive onconeuronal result by commercial kits should be reinforced by a second method, and that both positive and negative results are interpreted in the context of the patient presentation. Research laboratories with an interest in these disorders and routine laboratories with close links to experienced clinicians may be able to help with difficult cases. The importance of research-level assays is further emphasised by the rapid expansion of in newer entities (eg, Kelch-like protein 11 (KLHL11)), for which availability remains on a research-only basis despite its description in 2019.37 Overall, to maximise sensitivity and specificity, we recommend sending both serum and CSF for antibody testing as both biosamples show differing diagnostic characteristics across CNS autoimmune illnesses and, in combination, provide optimised diagnostic accuracy.38

Practical approach to testing: which antibodies should I suspect…

…at the population level

Among 979 cases in a 2010 European cohort, the most frequently detected antibodies were against Hu (~39%), Yo (~13%), CRMP5/CV2 (~6%) and Ri (~5%). The rate of seronegative PNS in the cohort was ~18%, and—overall—the most common tumours were SCLC (~38%), ovarian (~10%) and breast (~9%).3 More recently, a 2020 population-based Italian study identified Yo (30%), Hu (26%) and Ma2 (22%) as the leading antigenic specificities.1 Overall, these authors found PNS coexisted with 1 in 334 cancers, most commonly lung (17%), breast (16%) and lymphoma (12%). Modest variations between these two cohorts may be ascribed to ancestry, chronological trends or local environmental factors. However, it is important to note that, although some are especially typical, a variety of tumours can be associated with PNS (tables 1 and 2).

…if the patient has (forms of) encephalitis

Onconeuronal antibodies that target the ‘Hu’ antigens are classically associated with a limited ‘limbic’ encephalitis. These patients can also develop a multifocal neurological presentation, including spinal cord involvement and rhombencephalitis.24 A striking peripheral manifestation is a sensory neuronopathy, with clinical and electrophysiological studies also showing sensory or sensorimotor neuropathies.24 39 Typically, patients are in their mid-60s with a slight male predominance, but Hu-antibodies also have been detected in children with neuroblastoma in connection with a brainstem syndrome including opsoclonus–myoclonus29 40 and an aggressive—but non-paraneoplastic—paediatric limbic encephalitis with a limited response to immunotherapies.40

Several surface neuronal antibodies associate with autoimmune encephalitis and tumours. Again, these patients show clinical characteristics that help refine the pretest probabilities of a tumour. In addition to the age/sex bias of teratomas in patients with NMDAR-antibody encephalitis, patients with Morvan’s syndrome and contactin-associated protein-like 2 (CASPR2) autoantibodies have a 40% rate of thymoma compared with a <5% rate in patients with CASPR2-autoantibodies and limbic encephalitis.41 A diencephalic/brainstem-centred encephalitis is a hallmark of Ma-antibodies, particularly antibodies to Ma2, also known as Ta. Almost all affected patients are male and commonly have testicular germ cell tumours. A distinctive aspect of this disease localises to the hypothalamus with features including gelastic seizures and daytime somnolence (~30%), and even frank narcolepsy/cataplexy.19 42

KLHL11-antibody encephalitis was first described as a rhombencephalitis with prominent features of vertigo, diplopia, dysarthria, ataxia and auditory dysfunction (tinnitus and sensorineural hearing loss). The syndrome was originally described exclusively in men with a typical age of onset in the mid-40s and a strong association with testicular tumours (~65%). An early association has been proposed with the class II HLA alleles DQB1*02:01 and DRB1*03:01.37 43 By contrast, a laboratory-based study using only a cell-based assay detected KLHL-11 positivity in equal proportions of men and women with a wider phenotype including isolated germ cell tumours and patients with NMDAR-antibody encephalitis.44 Hence, the full clinical spectrum associated with KLH11-antibodies requires further study.

…if the patient has prominent psychiatric features

NMDAR-antibody encephalitis is the exemplar of antibody-mediated neuropsychiatry, and its most prominent characteristics include behavioural changes, psychosis mood disturbances, catatonia and sleep disturbances.7 The diagnosis should be suspected in new-onset acute to subacute psychopathology, especially in young women (~80%), typically with neurological accompaniments such as memory disturbances, seizures and/or movement disorders.45–48 Initial presentation to psychiatric services remains common,and patients may have received prior erroneous diagnoses of primary psychosis, bipolar disorder or depression.

The Ophelia syndrome is associated with surface-directed mGluR5-antibodies and seen in patients with Hodgkin’s lymphoma. Psychosis, hallucinations, mood disorder, behavioural and personality change, and sleep disturbance are among the psychiatric features reported.49

Overall, we recommend antibody screens in new-onset pyschosis for individuals fitting the aforementioned descriptions, as well as others with atypical demographic, phenotypical or clinical features when compared with primary psychiatric diagnostic categories, in particular, those with additional traditionally ‘neurological’ features.

…if the patient has a cerebellar syndrome

Several onconeuronal antibodies are strongly associated with paraneoplastic cerebellar degeneration. In this condition, the deterioration is usually acute to subacute, pancerebellar—causing limb and truncal ataxia plus nystagmus—and is frequently irreversible even with successful treatment of the underlying neoplasm, likely secondary to established Purkinje cell destruction.6 50 51 A cerebellar ataxia developing over the time course of a neurodegenerative process is rarer but can be observed, for example in patients with Ri- or glutamic acid decarboxylase- (GAD) antibodies although GAD-antibodies are rarely tumour-associated.52 53 Yo-antibodies are the most well-established association of a subacute cerebellar syndrome, almost always occuring in women with breast or gynaecological tumours,6 50 but one case series reported seven men, five of whom had gastrointestinal malignancy.54 Also, paraneoplastic cerebellar degeneration may occur with Zic4-antibodies, frequently detected alongside other onconeural antibodies in SCLC,22 and with Tr-antibodies, mainly detected in male patients with Hodgkin’s lymphoma.51 Further, around 20% of patients with Hu-antibodies develop a cerebellar syndrome,24 in addition to ~60% of those with Ri-antibodies, particularly women with breast cancer.52 Also to be considered in this setting are Ma1-antibodies (between 27% and 77% in small cohorts),55 56 CRMP5/CV2-antibodies57 and mGluR1-antibodies, the latter with superadded psychiatric and cognitive features.58 Finally, a subgroup of patients with voltage-gated calcium channel (VGCC) antibodies (P/Q type) have ataxia with SCLC, both with and without Lambert-Eaton myasthenic syndrome (LEMS).59

…if the patient has another movement disorder

Ri-antibodies may occur in combination with opsoclonus–myoclonus60 and, more recently, movement disorder presentations including tremor, parkinsonism and stiff-person syndrome, mainly in female (~80%) patients. Some have a stepwise progression and are misdiagnosed with atypical Parkinson’s disease, multiple sclerosis or a functional disorder.52 An important ‘not-to-miss’ association of Ri-antibodies is jaw dystonia and laryngospasm, which may be severe enough to precipitate nutritional deficiency or airway compromise.61 Amphiphysin-antibodies are associated with a paraneoplastic stiff-person syndrome and a female bias (~60%).62 In contrast, chorea should prompt consideration of CRMP5/CV2-antibodies, which show a male bias (~75%).57 63 Table 1 describes other features associated with these antibodies. Autoantibody-associated movement disorders have been reviewed in greater detail elsewhere and provide valuable clues as to the nature of the autoimmune illness.64

…if the patient has a peripheral nerve or muscle problem

Two classic peripheral PNS are sensory neuronopathy (or Denny-Brown syndrome) and LEMS.5 More than 50% of patients with Hu-antibodies have a sensory neuronopathy or peripheral neuropathy, characteristically involving both large and small fibres with dominant axonal involvement and only upper limb involvement in ~25%,24 39 with or without additional neurological features. The neuropathy occurring in connection with amphiphysi-antibodies may be immunotherapy-responsive, increasing the urgency of early and accurate detection.65 In CRMP5/CV2-antibody disease, the neuropathy is often a painful, asymmetrical, sensorimotor polyradiculopathy which may be steroid-responsive and commonly found alongside additional features such as cerebellar ataxia and myelopathy.66 A few (~10%) women with Yo-antibodies may present with a peripheral neuropathy of upper limb onset,67 and a painful peripheral neuropathy is recognised in association with CASPR2-autoantibodies.41

LEMS is accompanied by VGCC-antibodies in ~85% of patients and is paraneoplastic in ~70%.68 SOX1-antibodies are a useful biomarker of malignancy in patients with LEMS, being found in as many as ~60% of tumour cases but far less commonly in non-paraneoplastic cases.69 Paraneoplastic myeloneuropathy70 is a syndromic description associated chiefly with antibodies to amphiphysin, Hu and CRMP5/CV2, and in smaller numbers with other specificities (table 1).

…if the patient has lymphoma

Whereas limbic encephalitis (eg, with mGluR5-antibodies) and paraneoplastic cerebellar degeneration (eg, with antibodies to Tr/DNER or mGluR1) mainly occur with Hodgkin’s lymphoma,49 51 58 the more frequent associations of non-Hodgkin's lymphoma are sensorimotor neuropathies and dermatomyositis, although Tr/DNER and Ma2 specificities may occur in this context.71 72 It is noteworthy that histopathological efforts did not find DNER was expressed by the malignant cells of Hodgkin’s lymphoma, suggesting these patients represent a potential exception to the paradigm of tumour-driven immunisation.51

…if the patient has dysautonomia

Dysautonomia can cause potentially dangerous complications in several of the paraneoplastic syndromes. For example, in patients with Ri-antiboies, dysautonomia may lead to cardiac arrhythmias and central respiratory failure.52 Another manifestation is gastric pseudo-obstruction, especially seen with Hu-antibodies.5

Of the cell-surface antibodies, autonomic features are prominent in LEMS where they typically include dry mouth/eyes, visual disturbance, erectile dysfunction, constipation, impaired sweating and orthostatic hypotension, and are rarely life-threatening.73 Dysautonomia is a prominent feature of patients with NMDAR-antibodies. This tends to occur in ~50% of cases and characteristically becomes apparent 2–4 weeks into the illness, with manifestations including tachycardia–bradycardia, labile blood pressure and cardiac asystole.46–48 Finally, almost all (>90%) patients with Morvan’s syndrome, seen with CASPR2- ±LGI1-autoantibodies, have dysautonomia, most commonly hyperhidrosis (>85%) and cardiovascular instability (tachycardia, labile blood pressure in ~50%).41

…if the patient has visual loss

Among their many manifestations, antibodies to CRMP5/CV2 have a recognised association with paraneoplastic optic neuropathy, as well as posterior uveitis.57 63 Other antibodies can be more closely associated purely with visual syndromes. For example, melanoma-associated retinopathy has a characteristic presentation of night blindness, photopsias and reduced visual acuity, often occuring in patients with a pre-existing diagnosis of cutaneous melanomas and its appearance may signal a relapse. Onset can be abrupt and antibodies are found against bipolar layer retinal cells, with—as yet—undefined targets.74 Another syndrome with antibodies to retinal tissue, recoverin or cancer-associated retinopathy, causes painless visual loss in patients with known cancer of various types (table 1), as well as in some non-paraneoplastic cases.75 76


Overall, the early recognition of a paraneoplastic cause is essential to guide oncological investigation and optimise tumour management, which is also a cornerstone in addressing the associated PNS.73 Imaging forms the mainstay of investigations—as directed by the clinico-serological associations (tables 1–3). Tumour markers (eg, CA125 where ovarian cancer is suspected) may also be measured, especially when tumours are small and not definitively detected with imaging.6 As an overview, recommendations suggest body imaging to include CT of the chest, abdomen and pelvis,73 with additional—more focused—modalities based on the specific predicted associations, for example, testicular ultrasound scanning in men with Ma2-antibodies42 or pelvic/transvaginal ultrasound or magnetic resonance (MR) scanning in young women with NMDAR-antibody encephalitis.38 Selection of ultrasound scan or MRI to visualise an ovarian teratoma depends on local expertise and the mental state of patients, who are often extremely agitated in the acute phase of their illness. In our experience, some ovarian teratomas can be radiologically small and challenging to detect (figure 2). Expert radiological help is strongly recommended with equivocal scans. However, without the radiological suggestion of a teratoma, we do not recommend empirical oophrectomy as overall yields appear low. In some cases, such as in women with subacute cerebellar ataxia, confirmed circulating Yo-antibodies and negative radiological explorations, a surgical examination may be warranted.73

Figure 2

Body imaging in paraneoplastic encephalitides. (A,B) Pelvic MRI of a 19-year-old woman with N-methyl-D-aspartate receptor-antibody encephalitis. A right ovarian teratoma is visible as a dark cystic area (arrow) on a fat suppressed sequence (A). This was previously seen on T1 imaging (B) (arrow) but required fat suppressed sequences for the lesion to be clearly identifiable as a dermoid, rather than haemorrhagic, cyst.

To detect delayed tumour presentations, it is often recommended to follow-up a negative body CT with FDG-PET scan and surveillance imaging for several years thereafter.73 This decision – and the duration of follow-up - should rest with clinical acumen and perceived likelihood of detecting a tumour.

Paraclinical investigations: neurological

Cerebrospinal fluid

CSF is recommended to explore important differentials. CSF autoantibodies are frequently detected in PNS.38 77–79 In certain syndromes, such as NMDAR-antibody encephalitis, CSF NMDAR-autoantibodies are highly specific and often considered mandatory for a diagnosis.38 In most of the antibodies discussed within this review, broader CSF markers are abnormal (table 3) with common themes including lymphocytic pleocytosis, CSF-specific (‘unmatched’) oligoclonal bands and elevated protein. Over the range of these conditions, glucose is usually normal, and cytospin does not reveal malignant cells (table 3). One notable aspect in Ma2-antibody encephalitis is the possible finding of reduced hypocretin in patients with narcolepsy/somnolence.42 However, it should be highlighted that a bland CSF does not rule out PNS and, in certain entities such as CASPR2-autoantibodies with Morvan’s and in GABAAR-antibody encephalitis, CSF is quite often unremarkable (table 3).


MRI appearances of the brain or spinal cord are often characteristic in particular syndromes (table 3). For example, abnormalities in hypothalamus, diencephalon and rhombencephalon in patients with Ma2-antibody encephalitis,42 basal ganglionitis in those with chorea and CRMP5/CV2-antibodies,80 and multifocal cortical and subcortical abnormalities in patients with GABAAR-antibodies.81 In the spinal syndromes, paraneoplastic myelopathies often show a predominance for corticospinal tracts and are frequently longitudinally extensive.82

Again, as with CSF studies, it is important to note that MR scan of the brain and spine can be normal in PNS. This is an especially common scenario in NMDAR-antibody encephalitis,46 47 in patients with amphiphysin-antibodies52 and in Morvan’s syndrome with CASPR2-autoantibodies.41 Therefore, in a clinically and serologically appropriate setting, an unremarkable MRI should not deter from the diagnosis of some PNS. Moreover, in most cases of paraneoplastic cerebellar degeneration, initial MRI or CT is unrevealing with cerebellar atrophy visible only in later scans (figure 3).

Figure 3

Central nervous system imaging in paraneoplastic encephalitides. (A) CT scan of the head showing chronic cerebellar atrophy in a patient with Yo antibodies. (B–D) FLAIR hyperintensities in a patient with AMPAR-antibody encephalitis with a thymoma. The images demonstrate multifocal inflammatory T2 hyperintensities of bilateral hippocampi with other discrete cortical lesions in the right orbitofrontal cortex, left anterior cingulate, right posterior superior temporal sulcus, right anterior and inferior insula, right temporal operculum, left posterior temporal region, left occipital and left calcarine cortex. AMPAR, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor; FLAIR, fluid-attenuated inversion recovery.


Patients with encephalopathies will typically manifest electrophysiological abnormalities (table 3). One distinctive electroencephalogram (EEG) pattern is extreme delta brush in NMDAR-antibody encephalitis83—but is usually apparent only in the profoundly encephalopathic patient, hence having limited diagnostic value.46 An EEG study in patients with Hu-antibody showed surprisingly widespread abnormalities including in those without overt seizures or focal clinical signs.84

Electromyography is a valuable investigation in patients with CASPR2-autoantibodies where frank neuromyotonia is common.41 85 Peripheral neurophysiology in PNS may delineate relevant findings as described previously, but less specific findings of sensory or sensorimotor neuropathy are also possible (table 2). In suspected LEMS, decreased compound muscle action potentials are expected; and the electrical hallmark is increment after high-frequency repetitive nerve stimulation.73 Specific abnormalities consistent with bipolar cell dysfunction on electroretinogram may assist in the diagnosis of patients with melanoma-associated retinopathy antibodies.74


Tumour identification and appropriate oncological management are key to medical and neurological stabilisation, especially as many PNS associated with cancers display a limited immunotherapy response.73 Median survival is frequently aligned to the underlying tumour prognosis and delineated for the individual antibodies within tables 1 and 2. However, the PNS field is limited by small and highly selected cohorts, few randomised trials, reports of spontaneous improvement and frequently biased interpretations of treatment benefits.86 An analysis of deaths in 403 paraneoplastic patients with differing antibody specificities concluded that 109 (27%) died of their neurological syndrome, 150 (37%) of tumour progression and the remainder of unknown or other causes. The risk of death from the paraneoplastic syndrome was highest for patients with dysautonomic features; in this subgroup, nearly 60% died of neurological disease.3 More than one comparative observational cohort study has shown that patients with Hu-antibodies have a worse prognosis than those with CRMP5/CV2-antibodies.57 66

Treatment and immunotherapy outcomes: onconeuronal antibodies

While many cases with onconeuronal antibodies are immunotherapy-resistant or only partially responsive, it is important to recognise certain syndromes in which there is stronger evidence for immune-directed treatment. For example, young men with Ma2-antibodies and testicular cancer may show some recovery with tumour removal and immunotherapy,42 and an immunotherapy response is also documented in CRMP5/CV2- and amphiphysin-antibody-associated neuropathies.65 66 Women with amphiphysin-antibody stiff-person syndrome can also make worthwhile functional gains with cancer treatment and corticosteroids.87

On the other hand, there is limited evidence for a symptomatic benefit of immunotherapy in Hu-antibody neuropathy, and the main predictors of outcome among all patients with this specificity are disability/performance status, age and multifocal disease.24 One trial of triple immunotherapy (intravenous methylprednisolone, intravenous immunoglobulin and cyclophosphamide) in patients with Hu- and Yo-antibodies harbouring a variety of presentations achieved transient stabilisation in neuropathy patients but no improvement in other symptomatic groups.88 There were similarly disappointing results in a trial of intravenous immunoglobulin alone.89 Many cases of paraneoplastic cerebellar degeneration remain refractory to treatment even if the underlying malignancy is successfully addressed, a situation which is particularly observed with Yo-antibodies.6 50 67 Treatment response in some of the rarer syndromes is available only in case reports.90 The literature for treatment in paraneoplastic retinopathies is dominated by case reports, but it is generally believed that in cancer-associated retinopathy, cancer treatment alone is ineffective for visual recovery, and that additional immune treatments are beneficial.91

Treatment and immunotherapy outcomes: surface neuronal antibodies

In marked contrast, overall, the vast majority of patients with surface neuronal antibodies are immunotherapy-sensitive. In addition to oncological management, these patients merit a trial of immunotherapy, often extending to second-line or even third-line agents based on initial responses (table 2, Uy et al 13). More than 80% of treated NMDAR-antibody encephalitis can be expected to have a good outcome at 24 months,47 a figure likely to have increased since better recognition and more effective treatment protocols are commonplace. In this condition, paraneoplastic and non-paraneoplastic cases are likely to do equally well, but shorter time to oophorectomy is key in securing clinical improvement in patients with a teratoma.46 In our experience, this benefits from close multidisciplinary discussions between radiology and gynaecology colleagues to provide a balanced approach regarding considerations around fertility preservation in young women of childbearing potential.

There is a contrasting situation for patients with CASPR2-autoantibodies, where patients with associated thymomas show demonstrably worse outcomes compared with non-paraneoplastic patients.41 In an observational cohort, the mortality rate of the thymoma group was 50%, mainly due to tumour-related causes such as respiratory failure and tumour invasion, and the modified Rankin Scale score was static despite immunotherapies. However, recently in our centres, we have experienced some clinical benefits with rituximab in such patients (SR Irani and J Honnorat, unpublished data, 2021).

The syndromes associated with surface antibodies to mGluR1 and mGluR5, despite common associations with lymphoma, are typically immunotherapy responsive, and >50% of patients with mGluR5-autoantibodies can make a full recovery.49 58 In VGCC antibody-positive patients, the LEMS—but less so the cerebellar syndrome—responds to immunotherapy and 3,4-diaminopyridine can be used for symptomatic treatment.73 A report of global improvement following rituximab of a patient with VGCC with LEMS and cerebellar degeneration again highlights the possible opportunities of newer therapies in recalcitrant disease entities.92 Within this group, the poorest outlook is reserved for patients with GABABR-autoantibodies, whose functional improvement is in many cases partial and, despite immunotherapy gains, many die due to an underlying SCLC and its related complications.77 79

Key points

  • Early recognition of paraneoplastic syndromes is crucial for successful identification and management of the underlying tumour.

  • Serum results should be interpreted with a critical, clinical eye, given the pitfalls of line blot/immunodot testing.

  • Syndromes mediated by surface neuronal autoantibodies usually respond to prompt immunotherapy, in addition to oncological therapies; some syndromes associated with ‘onconeuronal’ antibodies may not respond, but there are important exceptions.

  • There is limited high-quality trial evidence, meaning much practice is based on observational or cohort studies.

Further reading

  • Graus F, Vogrig A, Muñiz-Castrillo S, et al. Updated diagnostic criteria for paraneoplastic neurologic syndromes. Neurol - Neuroimmunol Neuroinflammation 2021;8:e1014. doi:10.1212/nxi.0000000000001014

  • Makuch M, Wilson R, Al-Diwani A, et al. N-methyl-D-aspartate receptor antibody production from germinal centre reactions: Therapeutic implications. Ann Neurol 2018;83:553–61. doi:10.1002/ana.25173

  • Ruiz-García R, Martínez-Hernández E, Saiz A, et al. The diagnostic value of onconeural antibodies depends on how they are tested. Front Immunol 2020;11:1–6. doi:10.3389/fimmu.2020.01482

Ethics statements

Patient consent for publication



  • Twitter @ANG_Oxford

  • Contributors SRI and SB conceived the study and produced the first draft. CU and JH edited the manuscript for intellectual content. All authors approved the final text.

  • Funding SRI is supported by the Wellcome Trust (104079/Z/14/Z), Medical Research Council(MR/V007173/1), BMA Research Grants - Vera Down grant (2013), Margaret Temple (2017), Epilepsy Research UK (P1201), the Fulbright UK-US commission (MS Society research award) and by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre. This research was funded in whole, or in part, by the Wellcome Trust (grant number 104079/Z/14/Z). JH is supported by a public grant overseen by the French National Research Agency (ANR), as part of the second 'Investissements d'Avenir' programme called BETPSY (reference ANR-18-RHUS-0012, For the purpose of open access, the author has applied a CC BY public copyright licence to any author accepted manuscript version arising from this submission. SB has received salary support from the NIHR and is currently supported by the Wellcome Trust. CU is supported by the Friedman Award for Health Scholars (University of British Columbia) and received salary support from the UBC Division of Neurology.

  • Disclaimer The views expressed are those of the authors and not necessarily those of the NHS, the National Institute for Health Research, the Department of Health, UBC, or Vancouver Coastal Health. The funders had no role in the preparation, review or approval of the manuscript, and decision to submit the manuscript for publication.

  • Competing interests SRI is a coapplicant and receives royalties on patent application WO/2010/046716 (UK patent number, PCT/GB2009/051441) entitled ‘Neurological Autoimmune Disorders’. The patent has been licensed commercially for the development of assays for LGI1 and other VGKC complex antibodies. SRI and SB are coapplicants on a patent application entitled 'Diagnostic Strategy to Improve Specificity of CASPR2 Antibody Detection' (PCT/GB2019/051257, publication number WO/2019/211633 and UK1807410.4). SRI has received honoraria from UCB, MedImmun, ADC therapeutics and Medlink Neurology, and research support from CSL Behring, UCB and ONO Pharma. CU and JH declare no competing interests with respect to this publication.

  • Provenance and peer review Commissioned. Externally peer reviewed by Neil Anderson, Auckland, New Zealand.

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.