1000/1000
Hot
Most Recent
Immune Epilepsy can be defined as acute symptomatic seizures secondary to autoimmune encephalitis” or“autoimmune-associated epilepsy. The former consider cases in which immune encephalitis with positive neural surface antibodies presents with symptomatic seizures but reaches long-term seizure freedom with immune-targeted therapy; the latter would account for those with antibodies targeting intracellular antigens such as glutamic acid decarboxylase (GAD) and onconeural protein antibodies, in which immunotherapy is frequently ineffective as a result of neural death and permanent brain damage
Epileptic seizures are one of the most frequent clinical presentations of autoimmune encephalitis [1], and in some cases, it may be the only symptom [2]. Immune epilepsy was firstly included in the ILAE Seizure Classification in 2017 [3], defined as patients with evidence of immune-mediated inflammatory disease in the central nervous system. This evidence is based on MRI and cerebrospinal fluid (CSF) findings, clinical response to immunotherapy, and the presence of neural autoantibodies in serum or CSF [4].
In recent publications, many experts have considered that not all seizures triggered by immune damage to the brain should be considered as immune epilepsy [2] since epilepsy is understood as a chronic process [5]. With this in mind, they suggest two different concepts [6]: “acute symptomatic seizures secondary to autoimmune encephalitis” and “autoimmune-associated epilepsy”. The former consider cases in which immune encephalitis with positive neural surface antibodies presents with symptomatic seizures but reaches long-term seizure freedom with immune-targeted therapy [7]; the latter would account for those with antibodies targeting intracellular antigens such as glutamic acid decarboxylase (GAD) and onconeural protein antibodies, in which immunotherapy is frequently ineffective as a result of neural death and permanent brain damage [8]. In any case, being able to differentiate this in clinical practice is sometimes difficult, since some cases of autoimmune encephalitis associated with neural surface autoantibodies present as a “forme fruste” with drug-resistant seizures being their only manifestation, or at least the only apparent manifestation, being similar to the chronic course of “autoimmune-associated epilepsy” [6]. Also, patients harboring neural autoantibodies can present as new-onset epilepsy lacking other features of autoimmune encephalitis and reaching a favorable outcome without the need for immunotherapy [9].
Nevertheless, hypothetical autoimmune pathogenesis may exist in patients diagnosed with epilepsy of unknown etiology and could imply a new therapeutic approach with immune-targeted therapies, for example, in patients with drug-resistant epilepsy. Taking this into account, many studies have attempted to appraise the existence of different neural autoantibodies in series of patients with epilepsy of unknown etiology [10][11][12][13], others have also tried to characterize the features of those patients with positive antibodies [14][15][16][17], and even some of them have attempted to validate predictive models for neural-specific autoantibody status [18][19][20].
Our results show a pooled prevalence of 7.6% (IC95, 4.6–11.2%) in patients with epilepsy of unknown etiology. This suggests a possible immune-mediated mechanism in a low but not negligible proportion of patients with epilepsy of supposed unknown etiology and therefore the opportunity for a targeted treatment. It should be noted that this differs from the results of a different clinical scenario, which is the frequency of epileptic seizures in autoimmune encephalitis, where the prevalence is high [21].
The prevalence between studies varied from 1.2% in Tecellioglu et al. [14] and 14.28% in Iorio et al. [11] and Li et al. [20]. Interestingly, these three studies were performed in drug-resistant patients, a characteristic favoring a possible autoimmune origin [18][19]. The relatively low prevalence in Tecellioglu et al. was probable due to the methodology. In this study, the authors determined GAD65 by IRMA, being the only study included that used this method. We cannot say at the moment that immunoradiometric assay (IRMA) is less sensitive than RIA, but studies in other ambits found differences in the results of both methods [22]. Nevertheless, the comparatively low prevalence in this study can be explained because only testing for onconeuronals and anti-GAD and VGKC neural autoantibodies was performed. Some patients in this study were positive for VGKC but LGI1/CASPR2 were not tested, and hence we did not include them as a positive result. Determination of specific neural surface autoantibodies in this study could have increased the prevalence of positive results. Differences in prevalence were also found between the rest of the studies included in the meta-analysis, and the different methodologies and criteria for patient selection could also be a reason for this. For example, one of the studies with lower prevalence was the one from Bruijn et al. [19], where the included patients could not be suspicious of autoimmune etiology. In order to homogenize the sample for the analysis, we excluded studies with retrospective recruitment, where suspicion of autoimmune etiology is high.
Determination of specific neural surface autoantibodies in this work could have increased the prevalence of positive results. Differences in prevalence were also found between the rest of the studies included in the meta-analysis, and the different methodologies and criteria for patient selection could also be a reason for this. For example, one of the studies with lower prevalence was the one from Bruijn et al. [19], where the included patients could not be suspicious of autoimmune etiology. In order to homogenize the sample for the analysis, we excluded studies with retrospective recruitment, where suspicion of autoimmune etiology is high.
We also found high variability in the methodology for detecting neural autoantibodies among the included studies as well as which autoantibodies were considered for testing. We observed a great difference regarding the techniques used in the studies included to analyze GAD antibodies, and also in the titers that were considered high or very high. Concerning neural surface autoantibodies, although the majority of the studies tested them by IF, some performed IHC first and then confirmed by CBA while others only performed CBA. In addition, the interpretation of what is a positive result by IF was variable between studies. For example, only some studies required confirmation by a second technique for certain autoantibodies [11][13][19][20], others excluded weak positive results and used higher dilution ratio in order to consider them positive [12][15][16], whereas Ansari et al. [10], which was excluded for meta-analysis by sensitivity analysis, considered weak positive results as positive, even with low dilution ratios, which could explain why it is the study with the highest pooled prevalence. In order to avoid pitfalls in the interpretation of neural antibodies results, it is advisable to perform panel testing of multiple autoantibodies rather than single antibody testing, screen both CSF and serum, and discuss cases with reference centers if antibodies directed to surface antigens are found at low titers in serum only, or if results do not match with the clinical presentation [23]. In recent studies in patients with epilepsy of suspected autoimmune etiology [24][25] where CSF analysis was systematically performed, only one patient had a positive result in CSF and was negative in serum [24]. Instead, eight patients had a positive result in serum but were not positive in CSF. Despite this, testing both CSF and serum in patients with suspected autoimmune encephalitis is essential [26], and hence should be also performed in clinical practice and future studiesof patients with epilepsy of suspected autoimmune etiology. Also, caution should be exercised when interpreting the results of commercial kits for onconeural antibodies because high frequency of false positive results and it is advisable to confirm positivity with at least two distinct techniques [27]. In future studies regarding neural autoantibody determination, it would be advisable to apply standardized methods [28][29][30] in order to achieve more homogenous results between studies.
None of the controls presented a positive result, regardless of whether some had other immune or neurological diseases. Despite most of the studies not having a control group, the above suggests that neural autoantibodies may play a role in the epilepsy of these patients. However, neural autoantibodies can also be found, albeit in a smaller proportions, in patients with epilepsy of other defined etiologies [12]. Additionally, some seropositive patients with epilepsy achieve a long-term outcome without the need of immunotherapy [9][31], so the mere presence of neural autoantibodies would not be enough to confirm a direct immune-mediated mechanism.
Regarding the pooled prevalence for each individual neural autoantibody, those with the highest pooled prevalence were GlyR, followed by GAD, NMDAR, and LGI1, with a high heterogeneity between studies. The prevalence for CASPR2 and onconeuronal autoantibodies was homogenous but with a lower frequency. Only one patient had a positive result for AMPAR, and none had a positive result for GABAB, suggesting that determining these antibodies in patients with epilepsy without other data suggestive of encephalitis is improbable, and that if detected, it is unlikely to have relevant significance. Instead, in view of the results, GlyR antibodies should be taken into account in future studies regarding immune epilepsies. These antibodies have been associated exclusively with long-term seizures without any other neurological symptom [32], with a better response to antiepileptic drugs, compared with other immune-based epilepsies [33], and with a response to immunotherapy in drug-resistant cases [34]. However, the high heterogeneity of their prevalence between studies makes it advisable to confirm results in future studies, with the drawback that testing for these antibodies can only be performed by a few laboratories. After GlyR, GAD was the most prevalent autoantibody in our analysis. This is not strange since, contrary to most neural autoantibodies, GAD antibodies have been associated for many years with drug-resistant epilepsies without meeting criteria of limbic encephalitis [4][6][35], which is congruent with the selection criteria of our study. The relatively high prevalence of NMDAR antibodies among other neural surface autoantibodies could be explained by the fact that NMDAR encephalitis is the most frequent autoimmune encephalitis [36]. Epileptic seizures can be the first symptom of many in NMDAR encephalitis [37], and in a few patients, it may be its only clinical expression [36][37]. However, in addition to the heterogeneity in the results, it should be taken into account that false positives and negatives of this antibody are not uncommon when performed in serum [38]. Our data also show that 1% of patients with epilepsy of unknown etiology can suffer from LGI1 encephalitis. Although this entity usually courses with typical features [39], with some even considered pathognomonic [40], epilepsy can be the presenting symptom, and memory and behavior alterations may be subtle or can be attributed to other causes [19]. Related to the previous point, is important to underlie that cognitive impairment is one of the classic manifestations of limbic encephalitis and sometimes is the predominant symptom (autoantibody-associated cognitive impairment) [41]. In patients with epilepsy of unknown etiology, a systematic cognitive exam revealing not only impaired working memory or short-term memory but also long-term memory formation may give us reason to suspect that there is an underlying autoimmune mechanism [42]. This could compensate for the lack of sensitivity in detecting subtle autoimmune encephalitis with actual criteria [26].