September 2026 Research Update

Across these papers, diagnosis was framed as putting together several types of information rather than relying on one test alone: symptoms and exam, MRI and EEG tests, spinal fluid (CSF) results, and antibody testing. Several studies tested scoring tools meant to estimate how likely autoimmune encephalitis is compared with infectious encephalitis, with moderate accuracy in the studied groups. Some newer laboratory markers and prediction models looked highly accurate in single-center datasets, but the papers emphasized that outside confirmation is still needed. [1] [2] [3] [4] [5] [6] [7]

In children with acute encephalitis, one center created a weighted clinical score to separate autoimmune from infectious causes using sex, fever, a specific MRI enhancement pattern, and season, with internal validation sensitivity 81.5% and specificity 72.3%. The same dataset showed fever and higher CSF white blood cell counts were more common in infectious encephalitis, while abnormal MRI and abnormal neurologic exams were more common in autoimmune encephalitis. [1]

For pediatric anti-NMDA receptor encephalitis, a retrospective validation study found that the number and pattern of symptoms mattered, with abnormal behavior, seizures, and movement disorders helping predict which children had NMDA receptor antibodies in CSF. Different symptom thresholds traded off sensitivity and specificity, suggesting that symptom-based screening can help in settings where testing resources are limited but will not be perfect. [2]

Another pediatric study tested whether an adult encephalitis prediction model could be used in children and found acceptable overall performance in that pediatric cohort. In exploratory pediatric-specific modeling, older age pointed toward autoimmune causes, while a CSF protein cutoff behaved in the opposite direction than might be expected (CSF protein at or above 30 mg/dL was linked with lower odds of autoimmune encephalitis in that dataset). A different score (DEE) performed poorly for predicting autoimmune encephalitis in this group. [3]

In adults referred to a tertiary neuroimmunology clinic, the APE2 score separated autoimmune encephalitis from non-autoimmune cases better than the Graus criteria in this retrospective dataset (AUC 0.894 vs 0.804). In this study, an APE2 cutoff of 4 or higher gave sensitivity and specificity around 80%, and performance was similar in antibody-negative autoimmune encephalitis as defined by clinician review. The authors also explored a modified model (APE2SM), but described it as post hoc and needing prospective validation. [5]

One adult case-control study reported that total IgA in CSF was higher in autoimmune encephalitis than in symptomatic non-inflammatory controls, and that adding CSF IgA to routine lab measures produced a six-variable model with very high discrimination (AUC 0.990) in that single-center dataset. The study only included patients who were antibody-positive for autoimmune encephalitis and stated that external validation is needed before this could be relied on more broadly. [6]

An interdisciplinary psychiatry-neurology consensus paper described a stepwise diagnostic evaluation when autoimmune encephalitis is suspected as a cause of psychiatric syndromes, combining clinical assessment with MRI, EEG, blood tests, CSF testing, and antibody testing in both blood and CSF. For unclear cases, it discussed additional tools like brain PET scans and more advanced antibody tests, while also noting limits of each modality and the risk of both missed cases and overdiagnosis. [4]

A single-patient report of a 17-year-old with anti-NMDA receptor encephalitis linked to an ovarian teratoma illustrated that early studies such as CT, EEG, or initial assessments may be inconclusive, while antibody testing in both serum and CSF confirmed the diagnosis. In that patient, MRI showed leptomeningeal enhancement, and tumor imaging identified the ovarian mass later confirmed on pathology. As a case report, this describes one person and may not apply to others. [7]

For patients and caregivers, these studies support the idea that diagnosing autoimmune encephalitis often depends on a bundle of clues (symptoms over time, neurologic and psychiatric features, MRI and EEG results, spinal fluid testing, and antibody tests) rather than one definitive early test. Scoring tools and added lab markers may help clinicians sort autoimmune from infectious causes or decide who is more likely to have a specific antibody, but many tools were developed or tested at single centers and may not perform the same way everywhere. Important uncertainties include how well newer models work in different hospitals, in people with antibody-negative disease, and in patients whose main early symptoms are psychiatric.

Across the papers, efgartigimod (a treatment that lowers immunoglobulin G, or IgG) was mainly studied in people with antibody-positive autoimmune encephalitis, including anti-NMDAR and anti-LGI1 encephalitis. The overall pattern was clinical improvement reported alongside drops in blood (and sometimes spinal fluid) antibody levels and total IgG, but the evidence is still early because most data came from uncontrolled studies or individual cases. [8] [9] [10] [11]

In an 18-person multicenter prospective real-world study, people with anti-NMDAR encephalitis and anti-LGI1 encephalitis had improved disability and symptom scores after efgartigimod, along with reduced antibody levels and lower total IgG in the blood. A small subset also had brain inflammation imaging that decreased in some brain regions at the same time symptoms improved, while other imaging results were negative. Side effects were limited in this report, with one case of mild pneumonia. [8]

A systematic review summarizing 12 reports (77 people total) described clinical improvement in 76 out of 77 people treated with efgartigimod, and it did not report serious side effects. The review also emphasized that the included studies were quite different from each other, so results could not be combined into a single pooled estimate. [9]

Two single-patient reports described efgartigimod use after limited response to earlier immune treatments, with improvement reported and no relapse during follow-up (6 months in one case and 12 months in the other). One case involved anti-NMDAR encephalitis with rapid cognitive improvement and a drop in IgG, and the other involved anti-mGluR1 encephalitis treated with efgartigimod followed by low-dose rituximab, with large drops in antibody levels and improved walking and coordination scores. Because these are individual cases and treatments were sometimes combined, they cannot show what would happen for most patients or which treatment component mattered most. [10] [11]

For patients and caregivers, these studies mainly add early, real-world and case-based information about efgartigimod as an option that aims to lower harmful antibodies in certain antibody-positive autoimmune encephalitis types. The reports suggest improvements in day-to-day brain functions (like thinking, alertness, and seizures) can occur alongside measurable antibody/IgG reductions, but it remains uncertain how well this works compared with other treatments because most evidence here did not include a control group. Safety signals in these papers were limited, yet the number of treated people was still small, so uncommon side effects may not be captured.

This group of papers covered several approaches used when symptoms were relapsing, hard to control, or complicated by seizures, intensive care needs, or severe psychiatric symptoms. Evidence ranged from one moderate-strength prospective cohort study to small case series and single-patient reports, so results are uneven and not equally generalizable. [12] [13] [14] [15] [16] [17] [18]

In a 76-person prospective cohort of adults with relapsing anti-NMDA receptor encephalitis, those who received ofatumumab as a second-line immune treatment had higher seizure reduction and seizure freedom rates over follow-up than those receiving other second-line approaches, and they also had better functional scores. The same study reported fewer relapses in the ofatumumab group and described that earlier initiation related to better seizure outcomes, while psychiatric symptoms at onset and abnormal brain MRI were linked with worse seizure outcomes. Reported safety events were mostly mild, including one injection-related reaction in the ofatumumab group and no serious adverse events in either group. [12]

Tocilizumab (an interleukin-6 receptor blocker) was described as an added immune therapy in difficult anti-NMDAR encephalitis situations, but only in small reports without a comparison group. In a single-center pediatric series of six children treated as a third-line option after other therapies (often including rituximab), all had favorable functional outcomes at 12 months, while reported side effects included varicella-zoster infection, mild low blood counts, and changes in blood fats. [13]

One ICU case report described a ventilated adult with refractory anti-NMDAR encephalitis who improved after tocilizumab was introduced as an early second-line treatment, with later rituximab also given, and she had good longer-term function reported at 25 months. The report noted a bacterial abscess identified after tocilizumab started, though it did not establish whether tocilizumab caused it. Because this describes one person with multiple treatments and intensive supportive care, it cannot show which step drove the improvement. [14]

Several case-based papers focused on other treatment bundles commonly used in autoimmune encephalitis, including steroids, intravenous immunoglobulin (IVIG), plasma exchange, and rituximab, sometimes in the setting of severe seizures or prolonged ICU care. A resource-limited setting report of a teenager with suspected autoimmune encephalitis and super-refractory status epilepticus described improvement after a long ICU course that included immunotherapy and plasma exchange, alongside significant complications such as ventilator-associated pneumonia and weakness needing rehabilitation. A separate pediatric anti-NMDAR case described combined immunotherapy plus structured nursing and rehabilitation supports, with functional recovery and infection challenges during care. [17] [18] [16]

One report addressed severe catatonia in seronegative paraneoplastic encephalitis (linked to a testicular germ cell tumor) after multiple immunotherapy and medication attempts had not worked. After 12 electroconvulsive therapy sessions, the patient had sustained improvement with resolution of catatonia and related behavioral and psychotic symptoms, though this is a single case and may not apply broadly. [15]

For patients, this theme shows that studies in 2026 included both immune-targeting medicines used beyond first-line treatment (such as ofatumumab or tocilizumab) and non-immune supportive interventions aimed at specific complications (such as electroconvulsive therapy for catatonia, or ICU-level seizure and rehabilitation care). The most comparative information in this set came from the ofatumumab cohort in relapsing anti-NMDAR encephalitis, while the tocilizumab and electroconvulsive therapy reports were small and mainly provide background on what has been tried and what side effects or complications were seen. What remains uncertain is which patients benefit most from these approaches and how they compare head-to-head with other options in controlled trials.

Across these studies, relapse risk and longer-term disability were not predicted by a single factor. Instead, patterns emerged linking outcomes to (1) what is seen on brain imaging, (2) early clinical features like seizures or psychiatric symptoms, and (3) immune-related signals measured in blood or spinal fluid. Most evidence here is observational, meaning it can show links but cannot prove what causes relapse or poor recovery. [19]

In a large multicenter cohort, 1-year relapse happened in about 1 in 4 people, whether antibodies were found or not, but the predictors differed by antibody status. In antibody-positive autoimmune encephalitis, a specific seizure type (faciobrachial dystonic seizures) was linked to higher relapse risk, and in LGI1 autoimmune encephalitis, older age at onset was linked to relapse. In antibody-negative autoimmune encephalitis, frequent daily seizures predicted relapse. [19]

The same multicenter study found that moderate-to-severe diffuse cortical atrophy and medial temporal atrophy on MRI were linked to worse disability at 12 months, in both antibody-positive and antibody-negative groups. This suggests that visible shrinkage in key brain regions may carry prognostic information beyond antibody status alone. [19]

In anti-LGI1 encephalitis, one single-center study found psychiatric symptoms were common and marked a more severe course: people with psychiatric symptoms had worse initial severity scores, more blood-brain barrier disruption, more low sodium, higher relapse rates, and worse 1-year outcomes. In that study, having multiple MRI lesion areas at the start also related to higher relapse, and greater lesion burden related to poorer cognitive scores at 12 months. [20]

Two imaging-biomarker studies suggested additional MRI features might help estimate future risk. In a cohort of people with definite autoimmune encephalitis with neuronal surface or synaptic antibodies, a larger choroid plexus volume at baseline was linked to a higher chance of developing autoimmune encephalitis-associated epilepsy and to worse 12-month disability, and a radiomics model based on choroid plexus features showed good prediction in a test set. [21]

In LGI1 autoimmune encephalitis, another MRI study found widespread changes in brain structure measured by a complexity metric (fractal dimensionality), often picking up changes not seen with volume alone. Early reductions in certain deep brain regions were linked to less improvement in disability and symptom severity scores, and over time, longer disease duration and relapses were linked to further declines in some subcortical measures. [22]

Beyond MRI, immune-marker studies pointed to possible blood and spinal-fluid signals related to severity and outcomes. In anti-NMDAR encephalitis, a protein pattern including lower blood CCL11 and CXCL6 and higher spinal-fluid MMP10 helped predict severe disease, and blood CCL11 alone helped predict 6-month outcome, including an external validation using absolute (ELISA) measurements. [23]

A separate exploratory study focused on the behavior of patient antibodies in lab tests for LGI1 and CASPR2 autoimmune encephalitis. Antibody-driven cell-killing and cell-engulfing functions were common, and the combined ADCC+/ADCP+ profile was the only independent predictor of poor 1-year outcome in that cohort, while complement-driven cell damage was not detected. [24]

For patients and caregivers, this group of studies suggests that relapse and recovery may be estimated using a mix of clinical features (such as certain seizure patterns or early psychiatric symptoms) plus MRI signs (such as brain atrophy or other structural measures), and in some settings, blood or spinal-fluid markers. Some of these tools (like MRI ratings and certain protein tests) are closer to real-world use than others (like specialized radiomics or antibody function lab assays), and it is still uncertain how well any single marker will apply to an individual person across different hospitals and treatment approaches.

Across these mainly single-patient reports plus one small cohort study, autoimmune encephalitis showed a wide symptom range, often mixing brain symptoms (behavior and thinking changes, seizures, memory problems) with physical or whole-body complications. Some people also had nerve and muscle weakness outside the brain and spinal cord. A smaller subset had gastrointestinal (gut) inflammation or gastrointestinal cancer around the time neurologic symptoms began, and outcomes varied depending on the underlying gut disease. [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39]

Psychiatric and behavior changes were common in these reports, including acute psychosis and catatonia in anti-NMDA receptor encephalitis, severe behavioral regression in a child with suspected antibody-negative autoimmune encephalitis, and neuropsychiatric symptoms in a toddler with CASPR2 autoimmune encephalitis after mercury exposure. Several reports also described seizures, including status epilepticus in a possible drug-triggered limbic encephalitis case. [25] [31] [36] [39]

Memory and thinking problems also stood out, sometimes as selective deficits rather than a global decline. One anti-LGI1 case began with olfactory hallucinations followed by memory impairment, and one anti-GAD65 case showed a severe amnestic syndrome with loss of semantic knowledge alongside difficult-to-control temporal lobe epilepsy. In an adolescent with anti-NMDA receptor encephalitis linked to ovarian teratomas, mild cognitive impairment persisted even after other symptoms improved. [28] [34] [32]

Some papers highlighted complications affecting the body beyond the brain. One anti-NMDA receptor encephalitis case was complicated by blood clots in both lungs and drug-induced liver injury during the illness course. In an anti-mGluR1 case with brainstem involvement, the patient developed a lower lung infection and a heart attack during hospitalization and later died from respiratory and circulatory failure. [25] [38]

Peripheral nervous system involvement (problems affecting nerves outside the brain and spinal cord) was described in a retrospective cohort of 46 people with CSF-confirmed anti-NMDA receptor encephalitis. Nine had peripheral nerve involvement, and limb weakness was much more common in that group; nerve testing abnormalities were frequent among those tested. In this study, having peripheral nerve involvement did not clearly change overall recovery after first-line immune therapy. [26]

A multicenter case series focused on 16 people with autoimmune encephalitis who also had gastrointestinal inflammatory conditions, including inflammatory bowel disease and gastrointestinal tumors. Gastrointestinal symptoms or imaging findings were present at the start of neurologic symptoms in most patients, and three deaths occurred only among those with metastatic gastrointestinal cancer. Many others had good function at last follow-up. [27]

Several reports emphasized that autoimmune encephalitis can occur in special settings or alongside other conditions. Examples included anti-NMDA receptor encephalitis together with MOG antibody-associated disease in a young child with an ovarian teratoma found later, GAD65 autoimmune encephalitis after traumatic brain injury, and anti-GAD65 epilepsy as part of chronic graft-versus-host disease after stem cell transplant. Another report described recurrent catatonia and focal epilepsy in a woman previously labeled with antibody-negative autoimmune encephalitis, where later evaluations did not show clear evidence of ongoing autoimmune inflammation. [30] [35] [37] [33]

Rare but notable brain imaging patterns were described in individual cases. One anti-NMDA receptor encephalitis patient developed progressive, Rasmussen-like asymmetric cortical shrinkage and later had residual spastic gait despite seizure control. Another case described limbic (mesial temporal) swelling on MRI during status epilepticus in a possible medication-associated limbic encephalitis. [29] [39]

For patients and caregivers, this set of papers reinforces that autoimmune encephalitis can look very different from person to person, even within the same antibody type, and it can involve more than mood or memory symptoms. Some complications (like blood clots, infections, or cancer-related illness in a gut-cancer subgroup) came from single cases or small series, so it is uncertain how often they occur. The strongest generalizable evidence here comes from the small anti-NMDA receptor cohort showing that peripheral nerve involvement and limb weakness can occur and may not necessarily mean a worse overall outcome, but larger studies are still needed.

Across the clinical studies, a repeated pattern was that the hippocampus (a memory-related brain region) showed weaker communication with other brain areas, and this mapped onto specific thinking and motivation-related symptoms rather than being a vague, nonspecific change. Both studies also suggest that standard measures of brain connectivity may not capture all the ways brain activity becomes less well-coordinated in autoimmune encephalitis. [40] [41]

In LGI1-antibody limbic encephalitis, patients mainly made "misbinding" errors on an object-location memory task (mixing up what was where), more than simple guessing, pointing to a specific short-term memory binding problem. In the imaging subset, weaker left hippocampus-to-prefrontal connectivity went along with more misbinding errors at short delays and with higher apathy scores, and right hippocampal volume was smaller in patients. [41]

In anti-NMDAR encephalitis, patients had reduced hippocampal connectivity, including links with a major resting brain network and the thalamus, but the study also found higher "pattern incongruency" (a measure of how mismatched moment-to-moment brain signal patterns are) even when another complexity measure looked similar. The usual relationship between these complexity measures seen in healthy people was disrupted in patients, and the pattern incongruency measure related to cognition beyond what standard connectivity explained. [40]

For patients and caregivers, these studies help explain why some symptoms can be very specific, such as mixing up details of recent events or reduced motivation, by linking them to hippocampus connections with frontal brain areas. This is mainly background research now: it does not change care by itself, but it may guide future work on better ways to measure brain recovery and to track symptom-relevant brain network changes. Important uncertainties include that both studies were cross-sectional (one time point), and one had a small imaging subset, so it is not yet clear how these measures change over time or how well they predict day-to-day function for an individual.

Together, the preclinical and exploratory studies point to two connected ideas: immune attacks can rapidly change how nerve cells signal at synapses, and immune-driven seizures and inflammation can be followed by longer-lasting reshaping of hippocampal circuits. Separately, measures of brain-signal "entropy" were explored as a way to summarize long-term changes in brain activity related to seizures, but the evidence is early. [42] [43] [44]

In a neuron culture experiment, a human antibody against the NMDAR (N-methyl-D-aspartate receptor) did not change the amount of receptor measured on the surface at 2 hours, but it did trigger widespread changes in phosphorylation (a common on/off chemical signaling process) in synapse-containing preparations. The pattern of changes fit with downstream signaling that looks more like NMDAR activation than inhibition, suggesting antibodies may affect brain cells by shifting internal signaling pathways even before large receptor loss is seen. [42]

In a mouse model designed so CD8+ T cells would target inhibitory interneurons in the CA1 hippocampus, immune cell infiltration was followed within days by acute seizures, and later by chronic spontaneous recurrent seizures along with hippocampal sclerosis, ongoing glial activation, granule cell dispersion, and mossy fiber reorganization. Reduced Reelin expression appeared during the chronic stage and was reported alongside these structural circuit changes, offering one potential link between inflammation, loss of inhibition, and lasting seizure-prone remodeling. [43]

In a small retrospective human study of people with epilepsy who became seizure-free after surgery, earlier seizure onset was linked to lower average cortical entropy, and entropy did not differ between tissue that was removed and tissue that was not. In a related rat experiment using LGI1 monoclonal antibodies to model autoimmune encephalitis, the treated animals showed lower entropy after seizures began, consistent with the idea that recurrent seizures may reduce the richness or variability of brain activity over time. [44]

For patients, this group of studies mainly adds biological context for why seizures and memory-area symptoms can persist: immune effects may act quickly at synapses and may also set the stage for longer-lasting changes in hippocampal circuitry in animal models. The entropy work is exploratory and does not yet provide a validated test for autoimmune encephalitis; it is better viewed as early research into ways to summarize long-term seizure effects on brain activity. What remains uncertain is how closely these lab and animal mechanisms match the range of human autoimmune encephalitis syndromes, and whether any of these measures can reliably track an individual persons recovery over time.