This Week in Neurology — Aug 19, 2026

Welcome to This Week in Neurology. This week we're covering 10 notable papers spanning multiple sclerosis diagnosis and disease-modifying therapy safety, epilepsy pharmacology and access to surgery, and a set of antibody-driven diagnoses that are easy to miss. Let's dive in.

We'll start with multiple sclerosis, where the theme this week is that both diagnosis and drug selection are getting more granular. In Neurology, Solomon and colleagues offer a practical review of the 2024 McDonald criteria and what their global implementation will actually require [1]. The substantive changes are worth committing to memory: the anatomical locations that count for dissemination in space have expanded, new paraclinical findings can substitute for dissemination in time, and for the first time the criteria formally incorporate the central vein sign, paramagnetic rim lesions, cerebrospinal fluid kappa free light chains, and optical coherence tomography. Perhaps the biggest conceptual shift is a new diagnostic pathway for people who are asymptomatic or have nonspecific presentations but typical multiple sclerosis findings on MRI, alongside the unification of the attack-onset and progressive-onset pathways. The authors are candid that this cuts both ways. Earlier diagnosis means earlier treatment, but prior revisions were repeatedly misunderstood and misapplied, and misapplication is a documented driver of misdiagnosis. Hence the specific guidance on paediatric patients, older patients, and those with vascular comorbidity, where incidental white matter change is the trap. And they flag a real equity problem: criteria that depend on optical coherence tomography and advanced MRI sequences are not neutral across health systems, particularly in low and lower-middle income countries.

Once the diagnosis is made, the question becomes which therapy, and here two papers in Multiple Sclerosis Journal converge on infection risk. Smoot and colleagues used United States claims data to propensity-match just under two thousand people with multiple sclerosis who stayed on either an oral fumarate or an anti-CD20 monoclonal antibody for at least two years [7]. By year two the fumarate group already had a modestly lower infection rate, but the gap widened with time: from year four onward, fumarate-treated patients had roughly a thirty percent lower infection rate. The pattern matters more than the headline number. Infection rates in the fumarate group stayed essentially flat, while in the anti-CD20 group they climbed progressively year on year, which is what you would expect from cumulative hypogammaglobulinaemia. This is claims data, so residual confounding and unmeasured disease severity apply, but it argues for actively revisiting long-term anti-CD20 exposure rather than treating it as set and forget. In the same journal, Wend-Lamita Konseiga and colleagues report a sobering counterpoint from a different mechanism [9]: a 61-year-old man with stable multiple sclerosis on natalizumab who developed acute hemiparesis, had multiple ring-enhancing lesions on MRI, and was found on biopsy and polymerase chain reaction to have reactivated cerebral toxoplasmosis. He progressed to coma and died despite antiparasitic therapy. He also carried a low-risk monoclonal gammopathy of undetermined significance. The lesson is that natalizumab's compartmentalised blockade of central nervous system immune surveillance is not only about progressive multifocal leukoencephalopathy, and that in older patients with haematological comorbidity, a ring-enhancing lesion deserves a broad infectious differential before it is called a tumefactive relapse.

Turning to epilepsy, which dominates the week, the most immediately actionable paper is in JAMA Neurology. Schubert and colleagues asked a question most of us answer by habit: which antiseizure medication is safe in a patient with epilepsy who also needs a direct oral anticoagulant [2]. Using a target trial emulation across an international federated electronic health record network, they assembled just over nine and a half thousand adults on antiseizure monotherapy plus a direct oral anticoagulant, and compared levetiracetam, valproate, and moderate and strong enzyme-inducing drugs against a pooled lamotrigine or lacosamide reference. Strong enzyme inducers behaved as pharmacology predicts: roughly a fifty percent higher risk of thromboembolic events, presumably through accelerated anticoagulant clearance, though with less major bleeding. The surprise is levetiracetam, which is the drug most of us reach for precisely because we think it is interaction-free. Levetiracetam was associated with roughly double the thromboembolic risk and about a sixty percent higher all-cause mortality, with no offsetting reduction in bleeding. Valproate carried higher mortality and roughly triple the rate of intracranial major bleeding. The authors estimate that around twenty-eight percent of thromboembolic events in this population might have been avoided under a lamotrigine or lacosamide strategy. Importantly, in patients on vitamin K antagonists the thromboembolic signal was essentially absent, which supports a direct oral anticoagulant-specific interaction rather than pure confounding. This is observational and channelling bias cannot be excluded, but the mechanistic coherence is hard to dismiss, and for a patient on apixaban or rivaroxaban it makes lamotrigine or lacosamide the more defensible default.

Three further epilepsy papers deal with timing and access. In Brain, Millevert and colleagues assembled an international cohort of 282 individuals with pathogenic KCNQ2 variants and asked whether the timing of sodium channel blockers matters [5]. Carbamazepine and oxcarbazepine were the most effective drugs in both loss-of-function groups, and in loss-of-function developmental and epileptic encephalopathy, starting a sodium channel blocker within the first month of life was associated with earlier seizure offset and with higher rates of attaining major motor milestones. That developmental association held after adjusting for seizure control at one month and for total medication burden, which suggests the benefit is not purely a function of stopping seizures. The authors are appropriately restrained: some children did poorly despite early, correct treatment, so variant severity and other modifiers matter. Still, this is a strong argument for rapid genetic testing in neonatal-onset epilepsy, because the result changes the drug you choose that week, not that year. Also in Neurology, Feng and colleagues addressed the contentious question of paternal valproate [6]. Using Taiwan's national birth cohort of over two and a half million children, with more than seventeen hundred exposed to paternal valproate during spermatogenesis and 548 exposure-discordant sibling sets, they found no increased risk of autism, attention-deficit hyperactivity disorder, intellectual disability, tic disorder, or congenital malformation. The null result persisted when restricted to fathers with epilepsy and in the sibling comparison. Given the regulatory restrictions now applied to men of reproductive age in several countries, this is reassuring evidence from a non-Nordic population, though the authors stop short of calling for a change in practice.

Rounding out epilepsy, Epilepsia carries two contributions on the treatment gap. Stern and colleagues pooled two studies of diazepam nasal spray used out of hospital and identified 402 episodes meeting the International League Against Epilepsy definition of tonic-clonic status epilepticus, meaning seizures continuing beyond five minutes [8]. About sixty-two percent terminated by the thirty-minute mark, close to half of episodes terminated within twenty minutes of the dose, fewer than five percent recurred within twenty-four hours, and a second dose was needed in only five percent. Serious treatment-emergent status events occurred in about seven percent of patients. This supports equipping families with an immediate-use rescue medication rather than waiting for emergency services. And Englot offers a critical review of epilepsy surgery utilisation twenty-five years after the landmark randomised trial [4]. Fewer than one percent of potentially eligible patients are referred, and those who are referred arrive at a centre on average two decades after seizure onset. Stereo-electroencephalography, laser interstitial thermal therapy, and thalamic neuromodulation have widened candidacy, but Englot warns against letting minimally invasive options eclipse resective and disconnective surgery, which remain the only potentially curative procedures validated by randomised evidence.

Finally, two papers on antibodies that masquerade as something else. In JAMA Neurology, Pluvinage and colleagues used proteome-wide phage display in patients with idiopathic myelopathy and identified autoantibodies against CD320, the transcobalamin receptor that ferries vitamin B12 into cells [3]. Just over half of the discovery cohort was positive, with about forty-five percent positivity in two independent validation cohorts. Antibody-positive patients had lower bioactive B12 in cerebrospinal fluid despite this being a central nervous system-restricted deficiency, and a recognisable phenotype: subacute onset, a normal cerebrospinal fluid profile, and dorsolateral cord signal change on MRI. Five patients received B12 supplementation with or without immunosuppression and four improved. That is a tiny treatment series, but for a category of disease where twelve to eighteen percent of myelopathies never get a label, a testable and potentially treatable mechanism deserves attention. Alongside it, Panda and colleagues in Epilepsia describe two patients with anti-LGI1 encephalitis who first presented with sick sinus syndrome and were managed as cardiac patients until the encephalitis declared itself [10]. Anti-LGI1 is largely reversible with early immunotherapy and can cause permanent damage or death if missed, so unexplained bradyarrhythmia with any cognitive or behavioural change should prompt an antibody panel rather than only a pacemaker.

If you only have time for one paper this week, make it the JAMA Neurology target trial emulation of antiseizure medications during direct oral anticoagulant therapy [2]. It challenges the widespread assumption that levetiracetam is the interaction-free safe harbour, and it applies to a large, easily identified group of patients sitting in your clinic today.

Here are the key takeaways from this week in Neurology. When a patient with epilepsy needs a direct oral anticoagulant, lamotrigine or lacosamide now look like the safer default, and levetiracetam should no longer be assumed innocent. The 2024 McDonald criteria enable earlier multiple sclerosis diagnosis but demand disciplined application, particularly in older patients and those with vascular white matter change. Infection risk on anti-CD20 therapy rises progressively with years of exposure, while natalizumab-associated opportunistic infection extends beyond progressive multifocal leukoencephalopathy in older patients. In neonatal-onset epilepsy, fast genetic diagnosis matters, because early carbamazepine or oxcarbazepine in KCNQ2 loss-of-function disease tracked with better seizure and developmental outcomes. And keep unexplained myelopathy and unexplained bradyarrhythmia on your autoimmune radar, because anti-CD320 and anti-LGI1 are both potentially treatable.

That's your roundup for This Week in Neurology. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And a quick rating on Apple Podcasts or Spotify helps other physicians discover the show.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.