Abstract
Doran E, Kelly A, Stanila R, Healy L, Doherty CP. Epilepsia. 2025 Aug;66(8):2743-2753. Objective: Following emerging evidence of autoimmune-associated seizures in medication-refractory epilepsy, we began offering a trial of immunotherapy to selected patients. Here, we review this approach's treatment response, predictive clinical features, and utility. Methods: This was a retrospective single-center cohort study (2018–2022) of empiric, palliative immunotherapy in 31 adults with highly refractory, highly active epilepsy. Since 2018, in line with the International League Against Epilepsy's addition of “immune” as an etiology in the Classification of Epilepsy, we initiated immunotherapy after comprehensive antiseizure medication failures while at the same time screening for an autoimmune origin. The workup included assessing clinical features, serum autoantibody testing, cerebrospinal fluid testing (where feasible), magnetic resonance imaging (MRI), and electroencephalography. All patients received intravenous methylprednisolone or IV immunoglobulin according to previously published protocols, and follow-up was for at least 12 months. Results: Nine patients (29%) in this highly refractory cohort demonstrated a sustained treatment response, measured as a greater than 50% improvement in seizure frequency for at least 12 months. Three patients (10%) became seizure-free. Six patients (20%) were classified as partial responders and experienced an initial response that was not sustained. Apart from a trend toward a diagnosis of focal epilepsy, we did not identify any specific serological, clinical, electrodiagnostic, or imaging features with statistical significance that were predictive of treatment response. Significance: This patient group demonstrated a reasonable response rate to an immunotherapy trial. These findings are surprising but support the consideration of an immunotherapy trial in patients with refractory epilepsy. Requirements for repeated courses of immunotherapy differed significantly between patients, and this is an area of interest for further research. The basis for response in this cohort remains unclear; in some cases, antiseizure medication changes may have contributed; however, without any apparent autoimmune features, we consider potential blood–brain barrier repair or a placebo effect as hypothetical alternative mechanisms of action for the response to immunotherapy.
Commentary
Over the last two decades, there has been a dramatic expansion of knowledge on seizures arising from autoimmune phenomena. Thirty years ago, a sudden, severe encephalopathy with unexplained seizures might have been attributed to an “undiscovered virus.” Today, a neurologist would almost certainly implicate the immune system first. As a medical student in 2008, I recall a pediatric patient who presented with confusion, dyskinesias, and seizures and remained a medical mystery for months, only to later be recognized as having N-methyl-D-aspartate (NMDA)-receptor encephalitis. Today, a first-year neurology resident would be expected to raise that diagnosis in the emergency room.
This expansion of knowledge was driven at first by the identification and study of autoimmune encephalitis syndromes such as NMDA-receptor encephalitis, anti-LGI1 antibody-mediated encephalitis, anti-GFAP encephalitis, or various forms of limbic encephalitis. Patients often present with recognizable clinical patterns, associated with specific antibodies.
Naturally, the discovery of these syndromes gave a more prominent place to immune-mediated mechanisms in epilepsy. The 2017 ILAE classification made etiology a central pillar of epilepsy, with “immune” as one of the six basic categories 1 ; the word “immune” did not even appear in the 1989 classification. 2
With increased knowledge, however, the boundaries of the unknown also expanded. The hope for better treatments has led neurologists to seek autoimmune causes for patients with pharmacoresistant epilepsy that do not neatly fit well-defined syndromes. Here, the waters become murkier. Antibodies to glutamic acid decarboxylase (GAD) are found in different forms of epilepsy, notably in temporal-lobe epilepsy that does not always differ in obvious ways from patients without these antibodies, and often with limited responsiveness to immune therapy. 3 For other antibodies sometimes associated with epilepsy, phenotypes can be broad, pathogenicity may be debated, and treatment responsiveness is less clear. Other patients may have a dramatic onset of focal epilepsy without suspicious antibodies, prompting speculation about “seronegative” autoimmune epilepsy—an echo, in some ways, of the “undiscovered virus” era.
In parallel, research has brought attention to the role of inflammation in the pathogenesis and perpetuation of epilepsy, even in clearly nonimmune causes. 4 The result is that clinicians are increasingly asked to consider immunotherapy across a wider range of epilepsy patients, and some argue that it should be considered in all patients with pharmacoresistant epilepsy.
In their study, Doran et al 5 describe a protocol created in 2018 and retrospectively report the results. It targets a specific niche of highly refractory, highly active focal or generalized epilepsy: frequent seizures (>1/week for ≥6 months) despite comprehensive nonsurgical approaches, often while awaiting surgical evaluation or deemed not to be surgical candidates. An extensive workup was performed, including serum antibody testing and cerebrospinal fluid (CSF) when feasible, but treatment was offered regardless of results. Patients received steroids or, when contraindicated, intravenous immunoglobulin (IVIG), with steroid-sparing agents in selected cases; practice evolved toward intravenous methylprednisolone after side effects with oral prednisolone. Antiseizure medications were adjusted as clinically indicated. Seizure frequency was assessed at around 8 weeks and again at 12 months through standard clinical follow-up and patient report.
In total, 31 patients received immunotherapy and were retrospectively studied. Overall, 29% were labeled “responders” with a ≥50% seizure reduction at 12 months, and 20% were “partial responders,” with ≥50% reduction at 8 weeks but not sustained over the full year. A small number of patients achieved seizure freedom. Adverse effects were common: over a third reported side effects, and a few discontinued therapy because of serious complications, including steroid-induced psychosis and IVIG-associated aseptic meningitis.
Looking at the group as a whole, these patients have few features strongly suggestive of classic autoimmune epilepsy. Epilepsy duration was long; none of those tested had positive CSF autoantibodies; and four had positive serum antibodies (3 GAD and 1 weak voltage-gated potassium channel reactivity with a competing epileptogenic explanation). Only one of these four responded to immunotherapy. Furthermore, although a substantial proportion had no clear epileptogenic lesion on magnetic resonance imaging, several, including some who improved, had a clear structural basis for epilepsy, including a patient with Sturge–Weber syndrome.
The finding that such a large proportion had sustained improvement is promising, but interpretation is challenging for several reasons. The retrospective design leaves the cohort vulnerable to selection bias: it is not clear from what denominator these 31 patients were chosen, nor which factors influenced the decision to offer immunotherapy to them rather than to others who are not analyzed. Major confounders remain, particularly concurrent antiseizure medication changes, which the authors acknowledge could plausibly account for improvement in some cases. Finally, seizure quantification is necessarily “real-world”: the primary outcome hinges on a percentage change in seizure frequency, yet follow-up counts are based largely on patient-reported seizure burden, with inevitable imprecision (this challenge plagues much of epilepsy research). Taken together, these factors make it difficult to distinguish a true immunotherapy signal from natural fluctuation, regression to the mean, placebo effects, or the impact of evolving antiseizure regimens.
The real challenge, therefore, is not simply whether to determine whether immunotherapy can work, but to identify who is most likely to benefit and at what cost. The authors helpfully provide patient-level detail for all 31 cases. Certain observations stand out. None of the patients with generalized epilepsy responded, and all sustained responders had focal-onset seizures. When one examines the patient-level data, epilepsy duration appears shorter among responders: based on the durations listed in the case tables, the median duration is 6 years in responders and 10 years in nonresponders. Looking specifically at seizure-free cases, epilepsy duration was no more than 3 years. The patient with Sturge–Weber syndrome and negative antibodies achieved sustained seizure freedom after a relatively short course of steroid therapy, despite a substantial reduction in antiseizure medications over the ensuing year; an outcome that is striking and deserves further study, but remains difficult to interpret.
Ultimately, it will not be possible to draw firm conclusions from these data. Nonetheless, by formalizing and reporting an empiric immunotherapy pathway in a group that would not usually receive it, the authors add legitimacy and a degree of structure to a practice that is sometimes attempted warily in desperate circumstances. These data are best interpreted as an empirical immunotherapy experience in highly active refractory epilepsy, without a demonstration of immune etiology and, in most cases, without any clear suggestion of it.
The next step will be to determine, ideally in prospective studies with tighter control of antiseizure medication changes and more standardized seizure ascertainment, which patients are most likely to benefit, and how to balance that benefit against the nontrivial risks of treatment. Those with severe focal or multifocal epilepsy of relatively short duration would seem a good place to start.
