Autoimmune Encephalitis Diagnosis
Autoimmune Encephalitis Diagnosis
No single test makes this diagnosis, and waiting for one is the commonest way it is missed. Autoimmune encephalitis is diagnosed on a pattern: a change that is new and subacute, in a recognisable domain, supported by Findings from the spinal fluid, the scan or the EEG, and arrived at after the alternatives have been excluded. This page sets out the criteria in plain language, what each investigation contributes and what each one cannot be trusted to rule out, why the antibody panel is sent on spinal fluid as well as blood, when a tumour has to be hunted for and how hard, and what happens when the panel comes back negative.
The criteria, and what the timeline contributes
Autoimmune encephalitis is diagnosed the way several difficult neurological diagnoses are made: by a set of criteria rather than by a single positive result. That is deliberate. It exists so that treatment can begin before a laboratory confirmation arrives, because in this illness the delay costs more than the uncertainty.
The criteria have three parts. An entry feature: a subacute onset — conventionally within about three months — of a change in short-term memory, in mental state, or in behaviour and psychiatric symptoms. At least one supporting feature: new seizures that are not explained by a previously known epilepsy, new focal neurological findings, inflammation in the cerebrospinal fluid, or MRI changes consistent with encephalitis. And a precondition: reasonable exclusion of the alternative causes.
What the label is worth: meeting all three is enough to start treatment. It is not the same as a confirmed subtype. Where a well-characterised antibody is identified, the diagnosis is definite; where the criteria are met and no antibody is found, it is treated as antibody-negative autoimmune encephalitis — a real diagnosis, not a holding position. Where the criteria are not met, the honest response is to keep looking rather than to treat on hope.
| Part | Requirement | Why it is there | Common error |
|---|---|---|---|
| Entry | Subacute change in working memory, mental state, or psychiatric symptoms and behaviour | Separates this from degenerative disease, which is slower, and from stroke, which is instant | Treating an isolated psychiatric symptom as automatically sufficient, or dismissing it as automatically psychiatric |
| Support | At least one of: new seizures, new focal findings, CSF inflammation, MRI changes consistent with encephalitis | Anchors the diagnosis in the nervous system rather than in the history alone | Requiring all four, or stopping the assessment when the first test is normal |
| Precondition | Reasonable exclusion of alternatives — infection, metabolic and toxic causes, structural lesions, and the other inflammatory diseases | Because the treatments differ completely and because infection is immediately life-threatening | Excluding infection after treatment starts, or excluding it so slowly that immunotherapy is delayed |
| Subtype | Identification of the antibody, where one is found | Names the family — cell-surface or intracellular — and therefore whether a tumour must be hunted for | Waiting for this before starting, or over-reading a single low-titre serum-only result |
And the document that does more work than anything else in the file: the timeline. What the person was like before, what changed first, in what order, and from roughly which date — written down by someone who knew them. It establishes the subacute pace and often points to the subtype before any test is sent. The symptoms page describes the shapes the timeline takes.
What has to be excluded first — infection above all
Exclusion is not a formality in this disease; it is the part that keeps people alive. The list of things that can look like autoimmune encephalitis is long, and the top of that list is occupied by one item that is treated differently, immediately, and on suspicion.
- Herpes simplex encephalitis. The single most important alternative. It presents with fever, headache, altered behaviour and temporal lobe changes, and it is treated with intravenous aciclovir before it is confirmed — while the cerebrospinal fluid PCR is awaited rather than after it returns. It is also, importantly, a cause of the thing it mimics: a proportion of people who recover from herpes encephalitis then develop anti-NMDA receptor encephalitis weeks later, with new psychiatric symptoms, abnormal movements or a fall in consciousness. Anyone who deteriorates after apparently recovering from herpes encephalitis should be assessed for this rather than assumed to be relapsing.
- Other infections. Bacterial and tuberculous meningitis and encephalitis, HIV, syphilis, Lyme disease, and the opportunistic infections that occur in immunosuppressed people. Cerebrospinal fluid is sent for cell count, protein, glucose and microbiological studies including PCR and culture as part of the same tap, not as a separate decision.
- Metabolic and toxic causes. Thyroid disease, vitamin B12 deficiency, liver or kidney failure, low or high sodium, hypoglycaemia, and the effects of medicines — including the anticholinergic and sedative load, which can produce a picture that is indistinguishable at the bedside.
- Structural lesions. Tumour, abscess, subdural collection, and the vascular causes. This is what the contrast-enhanced scan is for, and it is why the scan is read for the patterns of encephalitis as well as for a lesion.
- Other inflammatory and autoimmune diseases. Multiple sclerosis and related demyelinating disease, neuromyelitis optica spectrum disorder, vasculitis of the central nervous system, and systemic autoimmune disease with neurological involvement. Each is treated differently.
- Primary psychiatric illness. Which remains possible, may coexist, and should be a conclusion reached after the assessment rather than a label applied before it.
- Prion disease and the rapidly progressive dementias. Considered where the course is relentless, where the spinal fluid is acellular, and where the specific imaging and EEG patterns point that way.
The practical shape of this is: blood tests, spinal fluid studies and imaging are sent together, on the first day, not in sequence over a fortnight. Aciclovir goes in with them. The exclusion work and the immunotherapy then run on parallel tracks, and the absence of a completed exclusion is not, on its own, a reason to withhold immunotherapy when the criteria are met.
Imaging and EEG: what they show, what they miss
Both of these tests are supportive rather than decisive, and both are routinely over-trusted in the wrong direction — a normal result is taken as reassurance when it should not be. Knowing what each one is actually for is more useful than knowing what it can show.
| Test | What it can show | How it is used | What a normal result means |
|---|---|---|---|
| Contrast-enhanced MRI at 3.0T | Signal change in the medial temporal lobes in limbic encephalitis; scattered cortical or subcortical change in some subtypes; and the structural alternatives that must be excluded | To support the criteria, to point to a limbic process, and to exclude a lesion, an abscess or a collection | Very little. Normal imaging is common, especially in NMDA receptor disease, and it must not slow the assessment down |
| Routine EEG | Diffuse or focal slowing consistent with encephalopathy; epileptiform discharges that indicate a seizure tendency; and, in some cases, patterns associated with a particular subtype | To support the diagnosis, to identify subclinical seizures, and to establish a baseline before treatment | It does not exclude seizures and it does not exclude encephalitis. A single short recording catches what happens to be happening during it |
| Long-term video EEG | The brain activity during an actual episode, watched alongside the movement | To settle the question of whether a movement is a seizure — which decides whether the treatment is an antiseizure drug or immunotherapy | The definitive answer in most cases of doubt; worth persisting with when the first routine recording was uninformative |
| Sleep study, where indicated | Disordered sleep architecture, breathing pauses in sleep, and the dream-enactment behaviour that points to certain subtypes | To characterise the sleep disorder when it dominates the picture, as in IgLON5 disease | Not a screening test; ordered for a specific question rather than routinely |
| PET-CT | Metabolic change in the brain where MRI is normal, and — separately — a whole-body search for an occult tumour | Second-line brain imaging where the diagnosis is uncertain, and as part of the tumour screen when the first screen is negative | A normal study reduces probability without excluding anything; in the tumour search, a negative scan is a reason to repeat, not to stop |
The one sentence to hold on to: in this disease, a normal scan and a normal EEG are common, and neither is reassurance. The diagnosis lives in the clinical picture and the cerebrospinal fluid. Why the mechanism produces this pattern is set out on the causes page.
The antibody panel: serum and cerebrospinal fluid
The antibody panel is what names the illness, and naming it changes the plan. It is also the slowest part of the assessment, and the part most often sent incompletely. Three things are worth being precise about.
It is sent on both, not either. Antibody testing is performed on serum and on cerebrospinal fluid, using cell-based assays, with tissue-based assay as an adjunct for confirming a pattern or catching something the panel did not target. A result found only in blood, particularly at low titre, may be incidental; the same antibody found in the spinal fluid — where the immune response is happening — carries substantially more weight. Blood alone is not an adequate panel.
The family matters more than the name. Antibodies against proteins on the cell surface or at the synapse — NMDA receptor, LGI1, CASPR2, AMPAR, GABAA receptor, DPPX, IgLON5 — are usually directly pathogenic and usually respond to immunotherapy. Antibodies against proteins inside the cell — Hu / ANNA-1, Yo, Ri, Ma2, CRMP5, amphiphysin, GAD65 — are markers of an immune-cell-mediated attack rather than the weapon, frequently indicate a tumour, and respond less well. That distinction, not the individual name, is what decides whether the tumour search is urgent and repeated.
A negative result is not the end. The panel is finite and assays vary between laboratories. Where the criteria are met, the diagnosis of antibody-negative autoimmune encephalitis stands, and the sequence that follows is defined: confirm the panel was broad enough and run on both fluids, repeat the spinal tap if the first was taken very early or after steroids were started, complete the exclusion list, finish the tumour screen, and document the response to treatment. Response to treatment is itself a diagnostic observation, and it is worth recording as one rather than as an anecdote.
Two practical cautions. Steroids given before the tap can reduce the inflammatory findings and make the picture harder to read, which is a reason to take the fluid before the first dose wherever the person is stable enough for it. And a positive result in isolation — an antibody with no clinical picture behind it — is a laboratory finding, not a diagnosis; it is read against the person, not accepted on its own.
The tumour search, and how hard to look
Some cases of autoimmune encephalitis are the immune system's response to a tumour that happens to express the same protein as the nerve cell. Finding that tumour is not an incidental part of the workup: in the intracellular antibody group it is the step most likely to change the outcome, because removing the source of the immune stimulus is more reliably effective than suppressing the immune response it provokes.
How hard to look depends on the antibody. With a cell-surface antibody such as NMDA receptor, the search is real but bounded — pelvic imaging looking for an ovarian teratoma in women, and screening of the chest, abdomen and pelvis, with the search repeated if the person does not respond as expected. With an intracellular antibody such as Hu or Ma2, the search is thorough and it is repeated: a tumour may be small, may be found only on PET-CT, and may not be present on the first screen. In that group, screening is repeated at intervals over the following years rather than once.
| Antibody group | Typical tumours | Screening | Follow-up if negative |
|---|---|---|---|
| NMDA receptor | Ovarian teratoma, most often in younger women; occasionally others | Pelvic ultrasound or MRI, with CT or PET-CT of chest, abdomen and pelvis | Repeat imaging if the clinical course does not follow the expected path; the search is driven by the response as much as by the calendar |
| LGI1, CASPR2 | Thymoma for CASPR2; tumour is uncommon in LGI1 disease | Chest imaging with attention to the thymus, extended to the abdomen and pelvis | A negative screen is usually sufficient; re-screen if the course is atypical or relapsing |
| AMPAR, GABAA receptor | Thymus, lung, breast | CT or PET-CT of chest, abdomen and pelvis, plus mammography where indicated | Repeat at intervals; relapse is described and is accompanied by a renewed search |
| Intracellular group (Hu, Yo, Ri, Ma2, CRMP5, amphiphysin) | Small cell lung cancer, thymoma, breast, ovarian, testicular — depending on the antibody | Whole-body PET-CT in addition to cross-sectional imaging; testicular ultrasound and mammography as indicated | Repeated at intervals over the following years. A negative first screen is the beginning of the search, not the end of it |
| GAD65 | Tumour is often absent | Screening is performed, but the yield is lower than in the groups above | A negative screen is more likely to be genuine; the focus shifts to long-term immunotherapy and symptom management |
One thing is worth saying to families directly: a tumour being found is not a worse outcome than none being found. It is frequently the better finding, because it gives the treatment a target and a mechanism. How tumour removal fits into the treatment order is set out on the treatment page.
Waiting for results without waiting to treat
The two are not the same, and separating them is the discipline this page exists to install. Waiting for a result is unavoidable — the antibody panel takes days to weeks, and confirmatory testing takes longer. Waiting to treat is a decision, and in this disease it is usually the wrong one.
What cannot wait. Intravenous aciclovir, started on suspicion while infection is excluded. The lumbar puncture, taken ideally before the first dose of steroid, because steroids can blunt the inflammatory findings that the diagnosis depends on. And — where the criteria are met — first-line immunotherapy, because treatment started earlier is associated with a more complete recovery. None of these requires the antibody result.
What legitimately waits. Second-line therapy, which is usually reserved for a documented inadequate response to first-line treatment rather than given alongside it. The decision to commit to long-term maintenance immunosuppression, which is made once the subtype, the response and the tumour status are known. And any irreversible step, which should never be taken on a single unconfirmed result.
What is recorded while waiting. A daily note of consciousness, seizures, movements, and the specific functions that matter to this person — because the recovery curve is measured against the worst point and against the baseline, and a chart started on day one is worth more than any recollection later. If the diagnosis is later revised, that record is what makes the revision possible rather than speculative.
The closing principle: a diagnosis here is a working position that gets sharper, not a verdict that arrives complete. It is revised as the fluid, imaging, antibody panel and response accumulate — and a revision is the system working, not the first clinician failing. What happens once treatment starts follows on the treatment page.