Autoimmune encephalitis and CNS vasculitis

• Autoimmune encephalitis is a non-infectious, immune-mediated inflammation of the brain that usually develops over days to weeks and presents with memory loss, behavioural change, psychiatric symptoms, seizures or movement disorders.
• Anti-N-methyl-D-aspartate receptor (anti-NMDA) encephalitis is the best known of these disorders; it affects young women most often and is strongly associated with ovarian teratoma.
• Diagnosis rests on a compatible clinical syndrome plus evidence of inflammation, with reasonable exclusion of infection; neuronal antibody testing in both serum and cerebrospinal fluid is required for definite diagnosis.
• Treatment is urgent: remove any tumour, give first-line immune therapy (high-dose corticosteroids, intravenous immunoglobulin or plasma exchange), and escalate to second-line agents such as rituximab or cyclophosphamide if there is no response within about a month.
• Central nervous system (CNS) vasculitis, especially primary angiitis of the CNS, is a separate and rarer diagnostic problem in which inflamed vessels cause stroke, headache or cognitive change; biopsy remains the gold standard and treatment is immunosuppressive.
• Always exclude infection before immunotherapy. Herpes simplex encephalitis can mimic or even trigger autoimmune encephalitis, so repeated viral testing is part of the work-up. ^[3]

By the end of this page you should be able to:

1. Recognise the clinical patterns of autoimmune encephalitis and name the major surface and intracellular antibodies.
2. Apply the 2016 Graus diagnostic algorithm to classify a case as possible, probable or definite autoimmune encephalitis.
3. Plan tumour screening according to the clinical picture and antibody type.
4. Choose and dose first-line and second-line immunotherapy, and explain when to escalate.
5. Distinguish autoimmune encephalitis from infectious encephalitis and from central nervous system vasculitis.
6. Outline the diagnosis and management of primary angiitis of the CNS using current guidance. ^[1][3][6]

Autoimmune encephalitis is a heterogeneous group of non-infectious, immune-mediated inflammatory disorders that primarily affect the brain but may also involve the meninges, spinal cord, peripheral nervous system or other organs. ^[3] It usually has a subacute onset (rapid progression over less than 3 months) and produces a polysyndromic picture because the underlying inflammation is diffuse rather than confined to a single area. ^[1][3]

Clinicians usually classify autoimmune encephalitis in three ways, and more than one classification can apply to the same patient:

• Anatomical classification groups cases by the main site of inflammation: limbic, cortical or subcortical, striatal, diencephalic, brainstem, cerebellar, meningoencephalic or encephalomyelitic. This guides the differential diagnosis, because different regions of the brain produce different syndromes. ^[3]
• Serological classification separates cases by antibody: antibodies to intracellular antigens, antibodies to neuronal surface antigens with high clinical relevance, antibodies of uncertain clinical relevance, and seronegative autoimmune encephalitis.
• Aetiological classification separates idiopathic, paraneoplastic, post-infectious and iatrogenic cases. The same antibody can appear in any of these settings. ^[3]

The distinction between surface antibodies and intracellular antibodies matters greatly. Surface or synaptic antibodies are thought to be directly pathogenic, so the resulting disease is often responsive to immunotherapy. Intracellular antibodies are usually markers of T-cell-mediated immunity driven by a tumour, so the disease is less responsive to immunotherapy and treatment of the tumour is central. ^[1][3]

This topic does not cover stiff person syndrome, progressive encephalomyelitis with rigidity and myoclonus or autoimmune cerebellar ataxia, which are related disorders with different presentations. ^[1]

Recent studies suggest that autoimmune encephalitis is probably at least as common as infectious encephalitis in many high-income settings, with an estimated prevalence rate of 13.7 per 100 000. ^[3] The reported incidence of encephalitis overall is about 5 to 10 per 100 000 inhabitants per year in high-income countries. ^[1] These figures vary between regions and studies because case definitions and the availability of antibody testing differ.

Anti-NMDA receptor encephalitis is the commonest autoimmune subtype in most series. It predominantly affects young individuals: in a cohort of 577 patients, 95 percent were under 45 years and 37 percent were under 18 years, with a female-to-male ratio of about 4 to 1. ^[1][5] Infectious encephalitis remains relatively more common in older adults and in low-resource settings where antibody testing is not available; the cross-check is therefore to run the infectious and autoimmune differential diagnoses in parallel rather than in sequence. ^[3]

Autoimmune encephalitis has no single cause. The broad categories are:

• Paraneoplastic disease, in which a systemic tumour triggers an immune response against neuronal antigens. The common tumours are small cell lung cancer, thymic neoplasms, breast cancer, ovarian teratoma or carcinoma, testicular teratoma or seminoma, neuroblastoma and lymphoma. ^[3]
• Idiopathic disease, especially in younger patients or those with a personal or family history of another autoimmune disorder.
• Post-infectious disease. Herpes simplex encephalitis is the classic trigger: about 20 percent of patients with herpes simplex encephalitis can develop a relapsing, autoimmune phase weeks to months later, most often anti-NMDA receptor encephalitis. ^[1]
• Iatrogenic disease, for example in patients treated with immune-checkpoint inhibitors or TNF-alpha inhibitors, which can accelerate a paraneoplastic encephalitis. ^[3]

Risk factors for a paraneoplastic aetiology include older age, smoking, rapid unintentional weight loss and a known cancer history. A preceding viral prodrome, fever or infection is common in many patients. ^[3]

In surface-antibody disease, the antibody binds to a neuronal receptor or synaptic protein and disrupts its function without destroying the neuron; this explains why the disease is often reversible and improves with immunotherapy. In anti-NMDA receptor encephalitis, antibodies against the GluN1 subunit cause internalisation of the receptor and impairment of NMDA-receptor-dependent synaptic plasticity; this accounts for the memory, behaviour, seizure and autonomic features. ^[1][5]

In intracellular-antibody disease, the antibody is a marker of cytotoxic T-cell attack against neurons driven by a tumour; the neuronal loss explains the poor response to immunotherapy and the dependence on cancer treatment. ^[1][3]

Inflammation in both groups is diffuse. A CSF lymphocytic pleocytosis, raised IgG index or oligoclonal bands indicates intrathecal antibody production. MRI may be normal in the first few days; in limbic encephalitis the characteristic pattern is bilateral medial temporal lobe T2 or fluid-attenuated inversion recovery (FLAIR) hyperintensity. ^[1][3]

The hallmark is a new, rapidly progressive encephalopathy with psychiatric and neurological features that evolves over days to weeks. Typical anti-NMDA receptor encephalitis in a teenager or adult begins with behaviour change, psychosis, delusions, agitation, insomnia or catatonia, followed by speech dysfunction, memory impairment, dyskinesias, seizures, autonomic instability and a decreasing level of consciousness. ^[1][5]

In a large cohort of 577 patients, onset and early progression differed by age: adults were more likely to present with behavioural and psychiatric features, whereas children under 12 years more often presented with seizures or movement disorders. Despite this, by the end of the first month 87 percent of patients had developed four or more of the major symptom groups, so the overall syndrome converges. ^[5]

The six major symptom groups used for the probable anti-NMDA receptor criteria are: abnormal psychiatric behaviour or cognitive dysfunction; speech dysfunction; seizures; movement disorder, dyskinesias or rigidity and abnormal postures; decreased level of consciousness; and autonomic dysfunction or central hypoventilation. ^[1]

Other surfaces to recognise:

• Limbic encephalitis: rapid confusion, short-term memory loss as the hallmark, mood change and seizures, often with CSF lymphocytic pleocytosis and bilateral medial temporal lobe MRI changes. ^[1]
• LGI1 antibody encephalitis: faciobrachial dystonic seizures, which are very brief, stereotyped dystonic movements of the face and arm, often preceding memory loss. CSF can be normal in these patients. ^[1][3]
• Faciobrachial dystonic seizures can look like tics; the combination of subacute memory loss and hyponatraemia should raise clinical suspicion. ^[3]
• Other movement or sleep disorders occur with antibodies such as CASPR2 (Morvan-type presentations), DPPX (diarrhoea and hyperekplexia) and IgLON5 (sleep disorder). ^[1]

Clinical pearl: because symptoms are polysyndromic, a patient who presents with isolated new psychiatric symptoms and otherwise normal neurological examination is unlikely to have autoimmune encephalitis until focal neurological, seizure or autonomic signs appear. ^[3]

Autoimmune and infectious encephalitis can look identical at the bedside. Both can produce subacute confusion, fever in the early phase of autoimmune disease, seizures and a CSF lymphocytic pleocytosis. Herpes simplex virus encephalitis is the most important mimic because it is common and treatable, and because it can itself trigger a later autoimmune encephalitis. ^[1][3]

Key clinical and laboratory separators:

• Very early CSF can be falsely negative for HSV. If suspicion remains high, the test should be repeated. ^[1][3]
• Bacterial meningitis and other infections usually produce a neutrophil-dominant cerebrospinal fluid, low CSF glucose and a markedly raised protein. Unexplained CSF glucose should not be ignored. ^[3]
• High fever with positive blood cultures or a known immunocompromising condition should push interpretation toward infection, and empiric antimicrobials should be continued until infection is excluded. ^[3]
• Central nervous system lymphoma or glioma can mimic limbic encephalitis on imaging; unilateral or mass-like MRI changes, or an atypical cell population in the CSF, require tissue diagnosis. ^[1][3]
• Vascular, metabolic and toxic causes of encephalopathy must be excluded before immunotherapy.

Read this alongside the infectious-disease topic on encephalitis, whose differential, empirical antimicrobial approach and timing of imaging are complementary; the two diagnoses should be investigated in parallel rather than in sequence. ^[1][3]

The table condenses the antibody table of the 2016 position paper and should be read as a guide rather than a rule; seronegative cases and overlapping antibodies are well described. ^[1][3]

The aim of investigation is to prove inflammation, identify the antibody and its tumour, and exclude infection and structural disease.

• MRI of the brain with contrast is preferred in autoimmune encephalitis. It may be normal early, especially in anti-NMDA receptor disease. The classic definite limbic encephalitis pattern is bilateral T2 or FLAIR hyperintensity restricted to the medial temporal lobes. Single or mass-like lesions should raise alternative diagnoses. ^[1][3]
• Lumbar puncture is the most important test and should be done early unless there is a contraindication. Send CSF for cell count and differential, protein, glucose ratio, IgG index, oligoclonal bands, viral PCR (at minimum HSV-1, HSV-2 and varicella zoster virus), bacterial and fungal studies when relevant, cytology and flow cytometry, and a paired neuronal antibody panel in CSF and serum. ^[1][3]
• CSF commonly shows a mild to moderate lymphocytic pleocytosis, commonly 20 to 200 cells per mm3 and in some cases as high as 900, hyperproteinorrachia and elevated IgG index or oligoclonal bands. In LGI1 antibody disease CSF can be normal, which may wrongly suggest a non-inflammatory encephalopathy. ^[1][3]
• EEG is used to exclude non-convulsive status epilepticus and to support a focal or multifocal encephalitis. Extreme delta brush is characteristic but uncommon in anti-NMDA receptor encephalitis. ^[1][3]
• Fluorodeoxyglucose PET of the brain may reveal abnormalities when MRI is negative and clinical suspicion is high, but its specificity is limited. ^[3]
• Blood tests for infection, systemic autoimmunity, metabolic and toxic causes should be guided by the clinical picture. Blood must be drawn before intravenous immunoglobulin or plasma exchange to avoid assay interference. ^[3]
• Neuronal antibody testing should use cell-based assays and should include CSF, not just serum. CSF is more sensitive for NMDA and GFAP antibodies; serum is more sensitive for onconeuronal, LGI1 and aquaporin-4 antibodies. Antibody dynamics correlate imperfectly with symptoms, so do not treat antibody titres rather than the patient. ^[1][3]

The 2016 position paper by Graus and colleagues provides a syndrome-based diagnostic framework that does not depend on antibody results or immunotherapy response at the bedside. Three levels are used: possible, probable and definite. ^[1]

All three of the following must be met:

1. Subacute onset (rapid progression of less than 3 months) of working memory deficits, altered mental status or psychiatric symptoms.
2. At least one of: new focal CNS findings; seizures not explained by a previously known seizure disorder; CSF pleocytosis with more than five cells per mm3; or MRI features suggestive of encephalitis. ^[1]
3. Reasonable exclusion of alternative causes. ^[1]

For clinically recognisable syndromes (limbic encephalitis, anti-NMDA receptor encephalitis, Bickerstaff brainstem encephalitis, acute disseminated encephalomyelitis) the criteria allow definite or probable classification before antibody results arrive. ^[1]

For probable anti-NMDA receptor encephalitis, there must be rapid onset of at least four of the six major symptom groups (or three groups with a systemic teratoma), plus either an abnormal EEG or CSF pleocytosis or oligoclonal bands, after excluding other disorders. Definite anti-NMDA receptor encephalitis requires one or more of the six symptom groups together with IgG GluN1 antibodies in the correct clinical setting. ^{[1] [5]}

For definite autoimmune limbic encephalitis, the criteria are subacute limbic symptoms, bilateral medial temporal lobe T2 or FLAIR changes, and either CSF pleocytosis or temporal EEG abnormalities, with reasonable exclusion of alternatives. ^[1]

Antibody-negative but probable autoimmune encephalitis has its own four-point criteria and should be diagnosed only after a thorough antibody panel in both CSF and serum. ^[1][2]

• Detection rates: CSF NMDA antibodies can be present when serum is negative (up to 14 percent in anti-NMDA receptor encephalitis), so both fluids must be tested. ^[1]
• IgG antibodies against the GluN1 subunit are specific. IgM or IgA antibodies against the same antigen are found in a broad range of disorders and in healthy individuals and should not be used to diagnose autoimmune encephalitis. ^[1]
• Molecular precision matters: antibodies to linear epitopes of other NMDA receptor subunits have uncertain significance, and voltage-gated potassium channel antibodies that are not directed against LGI1 or CASPR2 are not syndrome specific. ^[1]
• A later 2023 critical review warned that increased awareness has produced frequent misdiagnosis, largely from failure to meet the clinical requirements for possible autoimmune encephalitis, inadequate assessment of MRI and CSF inflammation, and underuse of confirmatory tissue-based assays. For possible or probable antibody-negative disease, clinicians should adhere closely to published criteria, especially excluding alternatives, and should be able to substantiate the absence of neural antibodies in both serum and CSF. ^[2]

Tumour screening is required in most adults with autoimmune encephalitis, because both paraneoplastic and non-paraneoplastic cases present similarly and the antibody is often unknown at first presentation. ^[3]

• Screening usually starts with CT of the chest, abdomen and pelvis with contrast, which is cheaper and provides structural detail for biopsy planning. Its limitations include low sensitivity for early breast and testicular cancer, and it is not preferred in children or pregnant women. ^[3]
• Mammography, and breast MRI if suspicion remains high, should follow a negative initial screen in women. ^[3]
• Young and middle-aged patients with a clinical picture suggestive of anti-NMDA receptor disease need targeted imaging: pelvic ultrasound (or MRI if equivocal) in women, testicular ultrasound in men. ^[3]
• Whole-body FDG-PET can be used when the initial CT is negative or inconclusive and suspicion is high. ^[3]
• After a teratoma is found, removal is associated with faster improvement and fewer relapses. ^[5]

In last-resort or difficult cases, the best-practice recommendations advise oncology referral and prolonged surveillance for intracellular antibody or limbic encephalitis, with semiannual to annual cancer screening for at least 4 years if no tumour was found. ^[4]

The extent of screening should reflect age, sex and antibody. In anti-NMDA receptor encephalitis, tumours are uncommon in children of both sexes and in males, whereas about 52 percent of females aged 12 years or older had an ovarian teratoma in one large cohort; screening similar to paraneoplastic disease (for example pelvic MRI every 6 months for 4 years) is appropriate in females aged 12 years or older, while the need for repeat screening in young children and males is unclear. ^{[5] [1]}

There are no robust randomised trials comparing modalities, so practice is based on observational evidence and expert consensus. ^[3] The 2016 criteria emphasise commencing immunotherapy once autoimmune encephalitis is highly suspected and infection is excluded. ^[1][3]

• High-dose corticosteroids: intravenous methylprednisolone 1 g per day for 3 to 7 days is a common reasonable approach. ^[3]
• Intravenous immunoglobulin: 2 g per kg over 2 to 5 days, especially when corticosteroids are contraindicated or when the picture is likely antibody-mediated (probable or definite NMDA receptor encephalitis). ^[3]
• Plasma exchange is the alternative fast immunomodulatory option, particularly for antibody-mediated disease, and is often used when corticosteroids are unsuitable or ineffective. ^[3]

Empirical choice varies by syndrome: corticosteroids alone or combined with other agents were selected by most specialists for a general encephalitis presentation, for faciobrachial dystonic seizures and for NMDA receptor encephalitis; for paraneoplastic disease associated with classical onconeuronal antibodies, corticosteroids are preferred theoretically but immunotherapy is often ineffective and cancer therapy matters more. ^[3]

If there is no sustained improvement within 4 weeks of first-line therapy or tumour removal, escalation rather than repetition is recommended. Roughly half of patients with anti-NMDA receptor encephalitis fail first-line therapy within the first month. ^[3][5]

• Rituximab is preferred by most clinicians as the second-line agent, and is less toxic than cyclophosphamide. ^[3]
• Common rituximab regimens are 375 mg per m2 weekly for 4 weeks, or two doses of 1000 mg 2 weeks apart. ^[3]
• Common cyclophosphamide dosing is 600 to 1000 mg per m2. ^[3]

In the observational cohort of 501 patients, 94 percent underwent first-line immunotherapy or tumour removal and 53 percent improved within 4 weeks. Among those who failed first-line therapy, patients who received second-line immunotherapy had better outcomes than those who did not, with an odds ratio of 2.69. Second-line therapy was also associated with fewer subsequent relapses. ^[5]

Refractory cases may respond to bortezomib, interleukin-6 inhibition with tocilizumab or, in children, escalation third-line therapy. These options are reserved for specialist centres. ^[3]

Maintenance therapy is individualised by antibody and by clinical response. Options include oral prednisone taper, periodic intravenous methylprednisolone or intravenous immunoglobulin, and bridging with rituximab. Relapses are uncommon (about 12 percent within 2 years in one large anti-NMDA cohort) and are usually milder than the initial episode. ^[4][5]

Patients who develop relapses despite maintenance, particularly in surface-antibody disease, need repeat courses of therapy and reconsideration of tumour status. ^[4][5]

In classical onconeuronal antibody disease the recommendations are different: treat the tumour, treat relapses with intravenous methylprednisolone or cyclophosphamide as required, and avoid long-term immunosuppression. ^[4]

Anti-NMDA receptor encephalitis has a good outcome (modified Rankin score 0 to 2) in about 81 percent of patients by 24 months, with a median follow-up of 24 months in the largest observational cohort; recovery can continue for up to 18 months after onset. Predictors of good outcome were early treatment and not needing intensive care. ^[5]

Relapse risk within 2 years was about 12 percent in that cohort; 67 percent of relapses were milder than the initial episode. Second-line immunotherapy was associated with fewer relapses, especially in patients without a teratoma. ^[5]

Mortality in untreated NMDA receptor disease historically approached 7 percent at 24 months with imputation; modern early treatment has improved this. Other autoimmune syndromes such as intracellular antibody encephalitis depend on cancer treatment and have a worse prognosis. ^[5]

CNS vasculitis is a different problem from autoimmune encephalitis, although both can produce cognitive change and seizures. It is inflammation of the blood vessels of the brain and spinal cord, presenting with headache, encephalopathy, multifocal neurological deficits or stroke-like events. It is classified as:

• Primary angiitis of the central nervous system (PACNS): isolated involvement of the CNS vessels, without systemic vasculitis.
• Secondary CNS vasculitis: vessel inflammation due to a systemic vasculitis or a trigger, such as antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis, systemic lupus erythematosus, Behcet disease, granulomatosis with polyangiitis, infections (varicella zoster, HIV, syphilis, tuberculosis), or drug exposure. ^[6]

PACNS is rare, with an estimated incidence of 2.4 cases per 1 000 000 persons per year, a male-to-female ratio of about 1 to 1, and a median age at diagnosis of about 50 years. ^[6] It has a long prodromal period (mean about 170 days from onset to diagnosis) in many patients and can present acutely with stroke. Headache, psychosis, cognitive decline and multiple territorial infarcts are typical patterns. ^[6]

The gold standard is histopathological confirmation on brain or leptomeningeal biopsy. ^[6] Biopsy is relatively safe but the risk-benefit ratio requires discussion, and non-diagnostic results occur. Non-invasive tests support the diagnosis but cannot confirm it on their own:

• MRI of the brain may show infarcts, leukoencephalopathy or contrast enhancement, but there is no pattern specific to PACNS. ^[6]
• Digital subtraction angiography (DSA) is more sensitive than magnetic resonance or CT angiography for medium- and large-vessel involvement; high-resolution vessel wall imaging is promising but not yet validated. ^[6]
• Cerebrospinal fluid is frequently abnormal (pleocytosis or raised protein) but normal cerebrospinal fluid does not exclude PACNS. ^[6]
• Blood tests and systemic screening are required to exclude systemic vasculitis and the many mimics (reversible cerebral vasoconstriction syndrome, intracranial atherosclerosis, intravascular lymphoma, moyamoya disease, secondary vasculitis, infection). ^[6]

The widely cited criteria are those of Calabrese and Mallek (1988), which require an acquired unexplained neurological deficit, classic angiographic or histopathological features of angiitis within the CNS, and no evidence of systemic disease or a mimic. Birnbaum and Hellmann (2009) revised these into definite (biopsy or autopsy proof) and probable (high-probability angiogram with abnormal MRI and CSF consistent with PACNS) categories. ^[6]

The 2023 European Stroke Organisation guideline issued expert consensus statements across 17 questions. Its key practical messages are:

• Be cautious interpreting non-invasive vascular imaging and consider digital subtraction angiography when magnetic resonance angiography is normal. ^[6]
• We suggest CSF examination is performed during work-up, not to prove PACNS but to exclude differential diagnoses. Normal CSF alone cannot exclude PACNS. ^[6]
• We suggest considering an immunosuppressant in addition to glucocorticoids in most patients, given disease severity and long-term steroid toxicity. ^[6]
• Glucocorticoid plus cyclophosphamide or mycophenolate mofetil are suggested for induction, with mycophenolate mofetil preferred for maintenance because of lower toxicity. ^[6]
• Platelet inhibition with aspirin may be included, especially with large or medium vessel involvement. ^[6]
• Intravenous thrombolysis and endovascular thrombectomy are not contraindicated in PACNS-related stroke if otherwise indicated. ^[6]
• Maintenance therapy should generally continue for at least 2 years before considering withdrawal in patients without relapse. ^[6]

Secondary CNS vasculitis should be sought and treated as the underlying condition dictates. Infection-related (for example varicella zoster or tuberculosis) or systemic disease-related vasculitis may improve with treatment of the trigger, with immunosuppression added when appropriate. ^[6]

• Seizures and status epilepticus, which can be refractory in autoimmune encephalitis.
• Raised intracranial pressure and cerebral herniation.
• Intensive care admission for dysautonomia, hypoventilation or encephalopathy; outcomes are worse when intensive care is needed. ^[5]
• Epilepsy, cognitive impairment and psychiatric sequelae after recovery.
• Medication toxicity: high-dose steroids, cyclophosphamide and immunoglobulin all have important adverse effects.
• Tumour-related morbidity, especially from ovarian teratoma in anti-NMDA receptor disease.
• Missed diagnosis, particularly HSV encephalitis presenting then relapsing as autoimmune disease. ^[1][3]

There is no primary prevention for most autoimmune encephalitis. Important secondary prevention steps include:

• Maintain herd immunity and vaccinate against measles, mumps, rubella, varicella and influenza, since infections (especially HSV) can trigger encephalitis and later autoimmunity.
• Display a low threshold for considering post-HSV autoimmune encephalitis in the months after recovery; investigate relapsing symptoms with CSF NMDA antibody testing. ^[1]
• In patients with anti-NMDA receptor disease and teratoma, tumour removal reduces relapses. ^[5]
• Consider malignancy screening as outlined above in most adult patients and extended surveillance in paraneoplastic forms. ^[3][4]

• Children. Anti-NMDA receptor encephalitis often presents first with seizures or movement disorder rather than psychiatric change. ^[5] Tumours are rare in children. Paediatric evaluation and dosing of immune therapy require specialist oversight; large paediatric cohorts showed comparable outcomes to adults but with small numbers for second-line therapy. ^[3][5]
• Pregnant women. Diagnosis is difficult because psychiatric change can be misattributed; management of tumour screening and immune therapy must be multidisciplinary. CT contrast is relatively safer than gadolinium in pregnancy, which affects choice of screening modality. ^[3]
• Older adults. Paraneoplastic causes and infection are more common; presentations can be misdiagnosed as dementia or delirium.
• Immunocompromised patients. Opportunistic infection must be excluded carefully, and aciclovir is continued for longer when immune function is impaired. ^[3]

Where antibody panels, MRI, digital subtraction angiography or second-line agents are unavailable, the following still improve outcomes:

• Recognise the syndromic patterns, especially anti-NMDA receptor encephalitis, from the clinical history and examination. ^[1]
• Exclude infection systematically: basic CSF studies, HSV and VZV testing by the best locally available method, and continuation of aciclovir until these are negative. ^[3]
• CT of the chest, abdomen and pelvis is often more accessible than PET and is an acceptable first neoplasm screen; pelvic ultrasound can screen for teratoma. ^[3]
• First-line therapy (intravenous methylprednisolone, intravenous immunoglobulin or plasma exchange, chosen by availability) should be given when the syndrome is convincing and infection has been reasonably excluded; delay in first-line immune therapy is an independent predictor of poor outcome. ^[5]
• For suspected primary CNS vasculitis, biopsy may not be available; treat vascular risk factors and consider immunosuppression only after a multidisciplinary discussion and exclusion of mimics, because antiplatelet and anticoagulant decisions differ from stroke caused by atherosclerosis. ^[6]
• Arrange early transfer of deteriorating patients to a centre with intensive care and neuroimaging.

• Autoimmune encephalitis is a major and possibly equally common alternative to infection as a cause of encephalitis in high-income settings; the cumulative 2016 Graus criteria classify cases as possible, probable or definite. ^[1][3]
• Antibodies are divided into pathogenic surface antibodies (immunotherapy responsive, often tumour-independent) and intracellular antibodies (tumour-driven, less responsive). ^[1]
• Anti-NMDA receptor encephalitis: young women, female-to-male ratio about 4 to 1; roughly half of women over 18 have an ovarian teratoma; children and males rarely have tumours. ^[1][5]
• Diagnosis requires both CSF and serum neuronal antibody testing; CSF is more sensitive for NMDA receptor antibodies. ^[1]
• HSV encephalitis must be excluded and can relapse as autoimmune encephalitis weeks later. ^[1][3]
• First-line therapy is intravenous methylprednisolone 1 g daily for 3 to 7 days, or intravenous immunoglobulin 2 g per kg over 2 to 5 days, or plasma exchange; escalate at about 4 weeks if there is no sustained response. ^[3]
• Second-line agents are rituximab or cyclophosphamide; rituximab is generally preferred. ^[3][5]
• In the largest anti-NMDA cohort, about 81 percent had good outcome by 24 months and relapse risk was about 12 percent within 2 years. ^[5]
• Tumour removal speeds recovery and reduces relapses; screen most adults and follow paraneoplastic cases for at least 4 years. ^[3][4][5]
• CNS vasculitis, especially PACNS, is rare, biopsy-confirmed and treated with glucocorticoids plus an immunosuppressant for induction, with maintenance for at least 2 years. ^[6]
• Antibody titres correlate poorly with symptoms; treat the patient, not the laboratory number. ^[1]

Encephalitis; meningitis; stroke; seizures and status epilepticus; herpes simplex infection; vasculitis; autoimmune connective tissue disease; ovarian teratoma; delirium; brain tumours; post-infectious demyelination; neuropsychiatric disorders in young women.