Antibody Investigations in 2,750 Children With Suspected Autoimmune Encephalitis

Discussion

In this study of pediatric patients with suspected AE other than ADEM, one-fifth had serum or CSF immunoreactivity with rat brain tissue, among which more than 90% had antibodies against neuronal cell surface proteins. The 2 most frequent neuronal or glial targets were NMDAR (76%) and MOG (5%), followed by GAD65 (2%), GABA_AR (2%), and several other cell surface or intracellular antigens (altogether comprising 6%).

We used a diagnostic approach that combines clinical information and brain tissue assays adapted to recognize most neural antibodies except for D2R, GlyR, and a subset of MOG antibodies.^9,10 Thus, excluding these 3 antibodies which were not investigated in all patients, we are able to provide the relative frequency of most antibodies.

In children, MOG antibodies usually associate with demyelinating disorders, such as ADEM, which were not part of this study.⁸ However, even with the exclusion of ADEM and considering that before 2014, MOG antibodies were not routinely examined in children with suspected AE, these antibodies were the second most common after NMDAR antibodies. Most patients with MOG antibodies had cortical encephalitis, highlighting the importance of MOG antibody testing in this clinical context.⁸

The small number of cases with antibodies other than NMDAR and MOG is noteworthy: they represented 10% of all positive cases and 2% of the entire series (2,750 patients). Two of 8 (25%) patients with isolated GABA_AR antibodies had predominant cerebellar symptoms, which is unusual in adult patients.¹¹ Antibodies against intracellular or onconeuronal antigens were extremely uncommon; the most frequent were directed against GAD65 (usually nonparaneoplastic) and Hu, which included 3 opsoclonus-myoclonus-ataxia associated with neuroblastoma and 1 brainstem encephalitis without tumor. Given that in the context of opsoclonus-myoclonus-ataxia with neuroblastoma, the Hu antibodies form part of the antitumor response rather than the neurologic syndrome,³ only 4 patients (one each: Hu, Yo, Ma2, and KLHL11) among 2,750 (0.1%) were diagnostically assisted by onconeural antibodies. These findings warn against the excessive use of paraneoplastic panels in children with potential AE.

We identified 53 (9%) patients with antibodies against unknown antigens that deserve future investigations to characterize novel clinical-immunological phenotypes.

A limitation of our study is that until 2014, not all patients were routinely tested for MOG and GlyR antibodies; therefore, some patients with these antibodies could have been missed. A similar problem applies to D2R antibodies, but their association with AE is extremely rare.¹² By contrast, the number of patients with GABA_AR antibodies may reflect a referral bias due to the absence of commercial testing until recently.

These limitations, however, do not affect the implications of our findings, which suggest that the current approach to neural antibody testing in children is inadequate. In these patients, testing should be based on the clinical picture focusing mainly on NMDAR and MOG antibodies that by far are the most frequently involved (81%), followed by GAD65 and possibly GABA_AR antibodies, which are much less common (4%). In adults, commercial antibody testing frequently associates with false-positive or misinterpretation of results.^13,-,15 Therefore, the implementation of the same tests to children, who have a different repertoire of diseases with more restricted antibody associations, may amplify the number of diagnostic errors. In children, many antibodies are so rare that their detection outside the expected syndrome requires confirmation with other tests before characterizing the disorder as antibody-associated or escalating treatment with immunotherapy. Thus, in children with suspected AE, up-front testing for many of these antibodies seems unneeded.