Brain 18F-FDG PET for the diagnosis of autoimmune encephalitis: a systematic review and a meta-analysis

To consolidate current understanding of detection sensitivity of brain 18F-FDG PET scans in the diagnosis of autoimmune encephalitis and to define specific metabolic imaging patterns for the most frequently occurring autoantibodies. A systematic and exhaustive search of data available in the literature was performed by querying the PubMed/MEDLINE and Cochrane databases for the search terms: ((PET) OR (positron emission tomography)) AND ((FDG) OR (fluorodeoxyglucose)) AND ((encephalitis) OR (brain inflammation)). Studies had to satisfy the following criteria: (i) include at least ten pediatric or adult patients suspected or diagnosed with autoimmune encephalitis according to the current recommendations, (ii) specifically present 18F-FDG PET and/or morphologic imaging findings. The diagnostic 18F-FDG PET detection sensitivity in autoimmune encephalitis was determined for all cases reported in this systematic review, according to a meta-analysis following the PRISMA method, and selected publication quality was assessed with the QUADAS-2 tool. The search strategy identified 626 articles including references from publications. The detection sensitivity of 18F-FDG PET was 87% (80–92%) based on 21 publications and 444 patients included in the meta-analysis. We also report specific brain 18F-FDG PET imaging patterns for the main encephalitis autoantibody subtypes. Brain 18F-FDG PET has a high detection sensitivity and should be included in future diagnostic autoimmune encephalitis recommendations. Specific metabolic 18F-FDG PET patterns corresponding to the main autoimmune encephalitis autoantibody subtypes further enhance the value of this diagnostic.


Introduction
Encephalitis is defined as a debilitating neurological disorder that develops as a rapidly progressive encephalopathy caused by brain inflammation [1]. Even though this pathology is relatively rare [2], its prognosis is poor, entailing serious irreversible sequelae and death in up to 7-12% of cases [3,4].
Autoimmune encephalitis is characterized by the presence of autoantibodies (aAbs) against neuronal targets [10]. We reviewed encephalitis by aAb subtype, in terms of onconeuronal and non-onconeuronal Abs, since this classification appears to more closely reflect the clinical presentation [11].
The diagnosis of autoimmune encephalitis is currently based on the clinical and paraclinical criteria defined by Graus et al. in 2016 [1]. Clinical criteria alone are often inadequate to diagnose autoimmune encephalitis, due to the lack of specificity of symptoms presented by patients [12]. A number of paraclinical tools have therefore been recommended to initially evaluate suspected autoimmune encephalitis cases. These evaluations involve standard biochemistry and immunology tests, to measure the intrathecal synthesis of anti-neuronal-Abs from lumbar punctures. However, it is not uncommon to obtain results several weeks after sampling; moreover, there is no guarantee that these tests will detect any aAbs in cerebrospinal fluid (CSF) at all. This has prompted a search for adjunct diagnostic tools such as electroencephalogram (EEG) and magnetic resonance imaging (MRI). But EEG results are often non-specific and MRI has a limited 25-50% sensitivity [10,13]. Because encephalitis patients may suffer severe and sometimes irreversible neurological sequelae, the early initiation of specific treatments and early follow-up of responses to these treatments [14] are key. This need underscores the importance of finding an early biomarker. 18 F-Fluoro-deoxy-glucose ( 18 F-FDG) position emission tomography (PET) is a functional brain imaging technique used to visualize neuronal glycolytic metabolic activity which increases during brain inflammation. Importantly, neurological alterations are associated with lower metabolic activity in specific areas of the brain [15]. 18 F-FDG PET has also been shown to be superior to morphological imaging in the early diagnosis of autoimmune encephalitis [15]. The increasing number of publications studying the results of brain 18 F-FDG PET in autoimmune encephalitis over the past decade is in dire need of consolidation [13][14][15][16][17][18][19][20][21][22][23][24]. The literature is saturated with isolated case reports and retrospective studies conducted on a small number of patients, leaving the role of PET in the initial assessment of the disease unclear.
Our current systematic and exhaustive literature search completed by a meta-analysis aims to consolidate current understanding of detection sensitivity of 18 F-FDG PET brain scans in patients diagnosed with autoimmune encephalitis, according to the PRISMA guidelines [26], and to better define specific metabolic imaging patterns associated with the most commonly occurring autoantibodies.

Materials and methods
We queried PubMed/MEDLINE and Cochrane databases, from inception until January 22, 2021, for the search terms: ((PET) OR (positron emission tomography)) AND ((FDG) OR (fluorodeoxyglucose)) AND ((encephalitis) OR (brain inflammation)) to address the specific review question: the diagnostic sensitivity of brain 18 F-FDG PET in cortical autoimmune encephalitis. Our systematic and exhaustive search, and meta-analysis, were performed by two independent observers (MB and AV) and conducted according to the PRISMA statement [26].
The references of articles retrieved from the systematic review were also reviewed to find any other potentially relevant articles.
Studies had to satisfy the following criteria: (i) include at least ten pediatric or adult patients (case reports, reviews, and letters to the editor were excluded) suspected or diagnosed with cortical autoimmune encephalitis (all aAbs including Rasmussen's encephalitis because of its highly probable autoimmune mediation, and aAb associated paraneoplastic syndromes) according to the currently recommended criteria and independently of PET results; (ii) specifically present 18 F-FDG PET and/or morphologic imaging findings.
Encephalitis unrelated to cortical autoimmune etiologies (please see Fig. 1 for details) were excluded. All studies satisfying these inclusion criteria were included in the systematic review.
The final selection was performed manually to ensure that all inclusion criteria were satisfied by two independent observers (MB and AV).
Since no prospective study involving encephalitis and brain 18 F-FDG PET is currently reported to be in clinical trials (www.clinicaltrials.gov. at the date of January 22, 2021), we decided to only include publications with at least 10 patients which fulfilled the high-quality QUADAS-2 tool [27] criteria, to limit any potential publication bias.
We report the number of 18 F-FDG PET and MRI scans used from each publication, based on one initial diagnosis scan per patient. A true positive scan was defined as specific abnormalities detected in a patient with suspected encephalitis, and the remaining negative scans were considered false negatives. This allowed the autoimmune encephalitis detection sensitivity to be calculated for both 18 F-FDG PET and MRI. A systematic compilation of 18 F-FDG PET detected abnormalities was subsequently determined for each individual aAb. Extractions were repeated, and a final consensus analysis performed (MB and AV).
The quality of each study was evaluated according to the quality assessment of diagnostic accuracy studies 2 (QUADAS-2) tool [27] by 2 independent investigators (MB and AV); discrepancies were resolved by consensus. Random-effect models were performed for the overall analysis and for the aAb analysis. Sensitivities and 95% confidence intervals are summarized in forest plots. We used the metafor package of R software version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Heterogeneity was evaluated with I 2 and considered significant if above 50% or when the p value of Cochran's Q was significant.

Study design
The search initially identified 469 publications. One hundred fifty-seven publications were subsequently added, after screening the references of these 469 publications, resulting in a total of 626 publications. Three hundred three publications involving non-autoimmune-mediated encephalitis or reporting non-original cases (reviews, clinical and epidemiologic descriptions) were excluded. Three hundred twentythree full-text publications were finally reviewed and considered for the systematic review; of these, 176 were subsequently excluded because 18 F-FDG PET was not performed or described, and an additional 126 because the number of patients having undergone a 18 F-FDG PET was less than 10. Finally, 21 full-text publications  representing 444 patients were reviewed and considered in the meta-analysis. For analysis of specific metabolic patterns of the most commonly occurring aAbs, 4 publications were subsequently excluded because they lacked descriptive details of aAb-related anomalies [29][30][31]42], leaving 17 studies, corresponding to 267 patients, in the analysis.
The study design is summarized as a flowchart (Fig. 1). QUADAS-2 [27] results are shown in Supplemental Fig. 1. Table 1 summarizes the patient characteristics, the autoimmune encephalitis diagnostic methods used, and the overall performances of 18 F-FDG PET and MRI for the publications included in the meta-analysis.
Brain 18 F-FDG PET imaging reports vast regions of hypermetabolism for LGI1 encephalitis entities, as detailed in Table 2. The majority of hypermetabolism are nevertheless located within the basal ganglia and mesial temporal lobes. It is interesting to note that hypometabolism, principally affecting mesial temporal lobes, is more frequently described in  2 Forest plots of the metaanalysis for the detection sensitivity of brain 18 F-FDG PET Table 2 Summary of encephalitis cases with PET abnormalities by aAb subtype identified from the publications included in the meta-analysis BG basal ganglia, MTL mesial temporal lobe, NMDAR N-methyl-D-aspartate receptor, VGKC voltage-gated potassium channel, LGI1 leucine-rich glioma inactivated 1, CASPR2 contactin-associated protein-like 2, GAD glutamic acid decarboxylase, GABA gamma-aminobutyric acid, AChR anti-acetylcholine receptor Bold represents predominant subtype cases with undetermined VGKC aAbs. The meta-analysis of anti-LGI1-Ab and anti-VGKC-Ab-mediated encephalitis identified detection sensitivities of 87% (79-92%), I 2 of 0% (p = 0.89) and 86% (63-95%), I 2 of 0% (p < 0.85) respectively (supplemental Fig. 4).

NMDAR (N-methyl-D-aspartate receptor) aAbs
NMDAR encephalitis preferentially affects young women and is frequently associated with teratomas. These aAbs specifically bind to the cell surface and are often associated with autoimmune encephalitis resulting in a significant number of cases with brain 18 F-FDG PET patterns in the literature (n = 66, Table 2). The clinical presentation usually progresses in four stages. The first is a prodromal phase with unspecific viral-like syndromes. This is followed by a typically psychotic phase. These two phases are usually followed by a mutic phase and finally a hyperkinetic phase with dysautonomia [10,18,20,21,[50][51][52][53]. The brain 18 F-FDG PET patterns are clearly dependent on the phase of the disease with a mix of described hypermetabolism and hypometabolism but with a typical anteroposterior gradient, as detailed in Table 2. In most of the cases, a mixed pattern of basal ganglia hypermetabolism and diffuse cortical hypometabolism is reported. The high frequency of hypometabolism reported in this entity is due to the delayed diagnosis of this encephalitis, the prodromal phase being unspecific. The typical hypometabolism of associative posterior areas, mainly occipital, is nevertheless described as an early biomarker for discriminating NMDAR aAb encephalitis from other autoimmune encephalitis [54]. However, different patterns of hypermetabolism affecting cortical areas other than the basal ganglia and the mesial temporal lobes are also described including a significant proportion of hypometabolic regions in the cerebellum and other cortical areas, which reflects the vast difference in clinical expression of this encephalitis. The NMDAR aAb meta-analysis revealed a detection sensitivity of 88% (74-95%), I 2 of 0% (p = 0.53) (supplemental Fig. 4).

GAD (glutamic acid decarboxylase) aAbs
Encephalitis associated with anti-GAD Abs is a relatively new entity, and thus less well described in literature. This encephalitis is caused by aAbs directed against synaptic antigens and is principally characterized clinically by Stiff person syndrome and cerebellar ataxia [9,10,18,51] but also more recently by refractory, mainly temporal, seizures [55]. Consistently with this latter clinical presentation, the 29 cases reported in Table 2 exhibit brain 18 F-FDG PET patterns mainly involving hypometabolism, associated with delayed diagnoses and typically affecting the mesial temporal lobes. The GAD aAb meta-analysis reported a detection sensitivity of 74% (51-89%), I 2 of 29% (p = 0.27) (supplemental Fig. 4).

GABA (gamma-aminobutyric acid) aAbs
Similarly to GAD-associated encephalitis, anti-GABA-Abs also bind to cell membrane antigens. As reported in Table 2, 3 cases with brain 18 F-FDG PET are reported in the literature. In contrast to GAD-associated encephalitis, those associated with anti-GABA-Abs lead to more obvious clinical symptoms, including status epilepticus and refractory epilepsy, cognitive deficits, psychiatric symptoms with depression, confusion, and mutism [10,18,20,21,50,52]. Once again, the observed brain 18 F-FDG PET patterns are related to the clinical characteristics with a majority of hypermetabolism, suggesting obvious early phase symptoms, predominantly involving the mesial temporal lobes. No data are currently available to determine the detection sensitivity of brain 18 F-FDG PET associated with these aAbs.

Onconeuronal aAbs
Onconeuronal aAbs are found in paraneoplastic encephalitis. Approximately two-third of cases involve anti-neuronal-Abs, with neurological symptoms preceding the diagnosis of a tumor by up to 4 years [37]. It is particularly useful to perform whole-body 18 F-FDG PET in these entities since the search for a primitive neoplastic tumor as well as dissemination of the primary lesion can be performed at the same time. A typical clinical evolution for this encephalitis subtype remains to be described [18,50]. Brain 18 F-FDG PET pattern data in the literature are scarce and is based on a total of 11 cases when all anti-neuronal-Abs are combined (Table 2). For anti-Hu-Abs, a diffuse cortical hypometabolism is preferentially reported, even though some hypermetabolism of mesial temporal lobes is also reported, whereas hypometabolism in the mesial temporal lobes is described for the anti-Ma ½ Abs. The onconeuronal aAb meta-analysis revealed a detection sensitivity of 73% (42-91%), I 2 of 0% (p = 0.46) (Supplemental Fig. 4).

Rasmussen's encephalitis
Rasmussen's encephalitis presumably involves an immunemediated mechanism even though the pathophysiology of this progressive disease remains unknown. Although recent observations do not exclusively relate to a childhood pathology, a childhood pathology and a progressive epileptic disorder due to chronic unilateral encephalitis are the two core characteristics [56]. Brain 18 F-FDG PET is a useful imaging tool in this setting since a typical pattern of hemispheric hypometabolism, exceeding the atrophy visualized in MRI, is observed and can help to diagnose the disease.
Only 1 publication involving at least 10 patients is available for Rasmussen's encephalitis and yields a sensitivity of 100% [32].
We also report in Table 2 the number of encephalitic cases with no specific aAbs (n = 19). Due to the probable heterogeneity of entities in this subgroup, the related brain 18 F-FDG PET patterns are also diverse with a predominance of hypometabolism. Figure 3 illustrates the typical brain 18 F-FDG PET patterns for the main aAbs entities of autoimmune encephalitis.

Discussion
Results from our systematic review confirm the importance of brain 18 F-FDG PET in the diagnosis of autoimmune encephalitis and measure an overall detection sensitivity performance of 87% (80-92%). These diagnostic performances seem to be consistent across the main aAbs subtypes (VGKC including LGI1, NMDAR, GAD, and onconeuronal aAbs, supplemental Fig. 4). The diagnostic sensitivity of 18 F-FDG PET from our meta-analysis is clearly higher than that typically reported for brain MRIs (56% [46-66%] in our current study and 25 to 50% in the literature [50,57]), which underscores the importance of systematically deploying this imaging modality in the initial diagnosis of suspected encephalitis cases. This detection sensitivity for assessing autoimmune encephalitis is very helpful, since diagnosis of the disease is currently delayed due to non-specific clinical symptoms and moderate performances of biological and imaging biomarkers [1]. Neurologists, radiologists, and nuclear physicians should also be cognizant of the benefits of brain 18 F-FDG PET in the initial diagnostic assessment of autoimmune encephalitis.
The different brain 18 F-FDG PET pattern results from our systematic review underlines a similarity observed among the neurodegenerative diseases [58]; 18 F-FDG PET abnormalities are strongly related to clinical symptoms. This allows typical brain 18 F-FDG PET patterns to be defined. The mesial temporal lobe involvement is observed in autoimmune encephalitis associated with limbic encephalitis (VGKC, GAD, GABA aAbs). Autoimmune encephalitis associated with a variety of clinical symptoms yields mixed hyper and  18 F-FDG-PET shows hypermetabolism in bilateral mesial temporal lobes. (F) Female patient diagnosed with autoimmune encephalitis related to anti-Hu antibodies. Typical brain 18 F-FDG-PET pattern showed increased metabolism in mesial temporal lobes. (G) Female patient, presenting a Rasmussen encephalitis. 18 F-FDG-PET of the brain showed asymmetric metabolism with hypometabolism of the right hemisphere and right basal ganglia hypometabolic patterns (NMDAR aAbs). Hypermetabolic patterns are associated with more obvious clinical symptoms (GABA aAbs) than hypometabolic ones (GAD aAbs), but the delay between the onset of symptoms and 18 F-FDG PET needs to be taken into account, as hypermetabolism is also observed in the acute phase of disease. In certain specific cases, the brain 18 F-FDG PET pattern is quasi pathognomonic of an autoimmune encephalitis entity (Rasmussen's encephalitis) with more extensive metabolic alteration than morphologic anomalies. Moreover, to reveal additional specific signs of encephalitis in paraneoplastic syndromes associated with autoimmune encephalitis, 18 F-FDG PET is able to provide an extensive assessment of neoplasia by performing a whole-body 18 F-FDG PET simultaneously to the brain 18 F-FDG PET [59,60].
As limitations of the current systematic review, it would have been interesting to report the specificity and accuracy of brain 18 F-FDG PET imaging in autoimmune encephalitis. Indeed, some brain 18 F-FDG PET patterns, especially diffuse hypometabolism, are not specific and are also observed in neurodegenerative diseases [58]. Unfortunately, false positive results or true negative cases can only be identified from wellconducted prospective studies which are rare in the current literature [28,30]. Although the potential bias of publication was taken into account, there are currently no ongoing prospective studies, limiting the possibility of including potentially negative PET results. However, in order to overcome this effect, all case reports were clearly excluded from the analysis and only high-quality studies involving at least 10 patients were finally included. Moreover, the presentation of the funnel plots in the supplemental Fig. 2 shows that the heterogeneity observed in our main analysis is mainly related to the different occurrences of the antibody types. The performance and 18 F-FDG PET patterns observed in our meta-analysis are mainly influenced by the time from the onset of symptoms, which is not always clearly reported in studies, thereby mistaking patterns of hyper-and hypometabolic areas, which may be related to the course of the disease [15,61]. It should nevertheless be noted that such different metabolic patterns could also be related to other factors. It has previously been suggested, in line with our presented results by aAbs, that metabolic patterns are associated with the type of aAbs, depending on whether it is against intracellular antigens (specifically involving mesio-temporal lobes) or against surface antigens (more likely to be hypometabolic patterns) [35]. Moreover, the status epilepticus should also be taken into account as it can lead to potential increases in 18 F-FDG PET metabolism in case of ictal phases related to the encephalitis [62].
Overall, our current systematic and exhaustive literature search and meta-analysis focus on sensitivity diagnostic performances as well as specific brain 18 F-FDG PET patterns in autoimmune encephalitis. We report the convincing performance of brain 18 F-FDG PET in the diagnosis of autoimmune encephalitis, which provides a helpful diagnostic imaging tool to overcome the challenges of diagnosing this entity and underscores the importance of including diagnostic 18 F-FDG PET in any future recommendations. Specific metabolic patterns corresponding to the main autoimmune encephalitis Ab subtypes value add to the diagnostic assessment. Further prospective studies are nevertheless required to further define performances in terms of specificity and accuracy.
Availability of data and material The data that support the findings of this study are available on request from the corresponding author (AV).
Code availability Not applicable.
Author contribution All authors contributed significantly to the analysis and interpretation of the data (MB, MD, MBC, AV), to the writing of the manuscript (EM, AV) and to the revision of the manuscript (EG, AK, LT, AV).

Declarations
Ethics approval and consent to participate This article does not contain any studies with human participants or animals performed by any of the authors.
Consent for publication The consent has been obtained for each patient for whom their FDG PET images are included in the manuscript.

Conflicts of interest
The authors declare no competing interests.