📝 Epilepsy and Its Shadow — Research Briefing

How "epilepsy" became a network of distinct disorders

When the bundle's earliest references — Sarah Blake's 2001 piece, the older Medscape chapters — were written, "epilepsy" was treated as a discrete neurological diagnosis with subtypes. The modern position, formalised by the ILAE in 2014 and updated in 2017 and again in 2025, is that epilepsy is an enduring predisposition to generate epileptic seizures and the neurobiologic, cognitive, psychological, and social consequences of this condition. The shift is small in wording and large in implication. A diagnosis of epilepsy now triggers a wider clinical task than seizure control alone — comorbidity screening, social-context assessment, mortality counselling, transition-of-care planning. Most clinics still operate at the seizure-control level; the field's frontier is operationalising the wider mandate.

The 2025 Beniczky et al. classification update (the latest ILAE position paper, published in Epilepsia) keeps the four onset categories (focal / generalised / unknown / unclassified) introduced in 2017, but formally defines consciousness as a classifier — operationally, awareness plus responsiveness — and adds negative myoclonus as a generalised seizure type. Practical consequence: the descriptors a clinician writes in the chart in 2026 differ subtly from those used even three years ago, and the older bundle PDFs (which still use "complex partial seizure," "simple partial," "secondarily generalised") will read as dated to anyone trained in the last decade. Translation: complex partial → focal impaired awareness; simple partial → focal aware; secondarily generalised → focal-to-bilateral tonic-clonic.

The shadow: psychiatric comorbidity

The single biggest revision to the public picture of epilepsy in the last decade is how comprehensively the psychiatric comorbidity has been quantified. The Kwon et al. 2024 JAMA Neurology meta-analysis — 27 studies, 565,000+ epilepsy patients, 13.4m+ controls — found significantly elevated odds for almost every disorder examined. Numbers worth memorising: depression OR 2.5 (13.7% vs 9.3%); anxiety OR 2.1 (8.4% vs 5.7%); any psychotic disorder OR 3.98 (14.8% vs 6.3%); schizophrenia OR 3.7; OCD OR 2.7; PTSD OR 1.8; suicidal ideation OR 2.3 (20.0% vs 11.7%); alcohol misuse OR 3.6; alcohol dependence OR 4.9; ASD OR 10.7; ADHD OR 3.9.

A few of those numbers deserve sitting with. One in five epilepsy patients reports suicidal ideation. One in seven has had a psychotic disorder at some point — a rate that puts epilepsy clinics in the same psychiatric-load territory as community mental health services. The autism OR of nearly 11 strongly suggests shared developmental neurobiology rather than either disorder causing the other, and is consistent with the genetic literature on shared variant risk for early-onset epilepsies and ASD.

The relationship runs in both directions. Patients with primary depression have a higher subsequent risk of being diagnosed with epilepsy than the general population; patients with primary epilepsy have a higher risk of being diagnosed with depression. The cleanest model is shared substrates: HPA-axis dysregulation, network-level structural differences in limbic regions (hippocampus, amygdala, anterior cingulate, insula), serotonergic and GABAergic imbalances. The implication for care is concrete: validated screening (NDDI-E for depression, GAD-7, the Columbia Suicide Severity Rating Scale) belongs at every epilepsy follow-up. Most clinics still don't do this.

Geschwind syndrome — the syndrome that won't quite resolve

The "Geschwind syndrome" — proposed by Norman Geschwind in the 1970s, and named for him in clinical literature thereafter — is the cluster of interictal personality traits sometimes seen in temporal lobe epilepsy patients: hypergraphia (compulsive writing), hyperreligiosity (intensified religious or philosophical preoccupation), altered sexuality (usually decreased), viscosity (verbose, sticky, circumstantial communication, difficulty disengaging from conversations), and intensified mental life (deepened cognitive and emotional responses to ordinarily neutral stimuli). Waxman and Geschwind's 1974 Neurology paper on hypergraphia in TLE is the foundational reference.

The clinical observation is recognised by experienced epileptologists. The diagnostic-entity status is contested. Benson's 1991 review put it bluntly: "more fire than substance" so far. Subsequent literature has not resolved this. The 2021 PMC review "Rethinking Geschwind Syndrome Beyond Temporal Lobe Epilepsy" notes that the same cluster appears in frontotemporal lobar degeneration (Veronelli et al.) and at least one case associated with glioblastoma — suggesting if there's a real syndrome it tracks limbic-network dysfunction rather than seizures specifically. Modern critics describe the hyperreligiosity link in particular as "not terribly compelling" at population level; much of the older case-series literature is selection-biased, with surgical-centre TLE cohorts being assessed by neurologists already looking for the cluster.

What the honest summary looks like for a clinician or family member: Geschwind features can occur in some long-standing TLE patients (they are not invented). There is no specific brain-imaging marker, no validated diagnostic instrument, no controlled-trial evidence for any treatment. It is more useful as a clinical shorthand for recognising a constellation than as a label that warrants pharmacological action. Where the cluster appears, the underlying epilepsy and its psychiatric comorbidities are the things to treat.

Psychogenic non-epileptic seizures (PNES)

PNES — also called dissociative seizures, functional seizures, or non-epileptic attack disorder — are paroxysmal episodes that look like seizures but are not driven by abnormal cortical electrical discharge. DSM-5 places them under "functional neurological symptom disorder, with attacks or seizures." They sit firmly within the Functional Neurological Disorder framework — same predictive-processing mechanism, same multidisciplinary treatment model, same field of clinical research.

The diagnostic problem is that PNES look like seizures. Trained nurses, paramedics, and even neurologists routinely cannot tell the difference at the bedside. The gold-standard test is video-EEG monitoring in an inpatient epilepsy monitoring unit: capture a typical event, observe no epileptiform activity before, during, or after. Outpatient ambulatory EEG misses most PNES because the events don't occur on cue. Average diagnostic delay is about seven years. Roughly 20–30% of patients referred to EMUs for "drug-resistant epilepsy" turn out to have PNES, not epilepsy. Roughly 10–15% have both — concurrent epilepsy and PNES — which makes clean separation impossible without video-EEG capture of the specific event types in question.

Treatment that has evidence: cognitive behavioural therapy (the CODES trial, Goldstein 2020 in Lancet Psychiatry, n=368, missed its primary endpoint of seizure-frequency reduction at 12 months but improved quality of life, distress, and psychosocial function on multiple secondary endpoints — a nuanced positive); rapid taper off antiseizure medications once the diagnosis is solid (delayed taper is associated with worse outcomes); prolonged-exposure therapy for the substantial PTSD subgroup. Counselling that does not target the underlying emotional content tends to fail. Sarah Blake's 2001 lived-experience account makes this point twenty years before the RCTs caught up: "Counselling is only helpful if it addresses the cause of the problem."

The stigma around PNES remains the field's worst-kept secret. Blake's account describes ER staff shouting at "Melanie" for "faking," psychiatric attendants leaving fingernail marks during restraint, and therapists refusing to take seizure-disorder cases. Current 2024–25 clinical reviews explicitly flag the same patterns as ongoing — the harm done by dismissive "good news, there's nothing wrong with you" framing is recognised in modern literature as a documented perpetuating factor, the same mechanism described in the FND briefing. The treatment-relevant fact is that PNES are not malingering and not under conscious control; the patient cannot stop the events on demand any more than an epilepsy patient can.

Frontal-lobe and reflex seizures — the easily-missed presentations

Frontal-lobe seizures are short (often under a minute), often nocturnal, and often hyperkinetic — bicycling movements, pelvic thrusting, screaming, fearful vocalisations, sudden bizarre posturing. They look so different from textbook seizures that they routinely get diagnosed as parasomnias, panic attacks, behavioural outbursts, or paediatric oppositional behaviour for years. The bundle's two Frontal Lobe Syndromes PDFs are present for good reason: clinicians who only know temporal-lobe presentations miss frontal-lobe ones.

Reflex seizures are seizures reliably triggered by a specific stimulus. Photosensitive epilepsy (light/pattern triggers) is the best-known but represents a small minority. Other triggers documented in case series: reading, hot-water immersion (especially in South-Asian populations), specific musical pieces, eating, brushing teeth, startle. Recognising the trigger pattern is diagnostic and changes counselling immediately — a patient who only seizes in front of strobing screens lives a different life than one with unprovoked focal-onset seizures.

Epilepsia partialis continua is a continuous focal motor seizure (often hand or face) that can last hours, days, weeks, or longer; it is rare but the bundle includes a PDF on it because it represents one of the chronic, drug-refractory presentations that anchor any serious epilepsy reading list. Post-traumatic epilepsy — onset following moderate-to-severe TBI — is a distinct epidemiological category with its own latency window and its own preventive-treatment literature (mostly negative for prophylactic ASMs beyond the first week).

Antiseizure medications: the side-effect problem the literature undersells

Older patients on antiseizure medications (ASMs) routinely report that the side effects of their drugs bother them more than the seizures. The literature has belatedly caught up. The 2025 Frontiers review on ASM adverse effects, the 2023 ASM-induced alopecia literature review, and the cosmetic-side-effects systematic review collectively describe a side-effect burden that explains a great deal of "non-compliance."

Alopecia. Valproate is the heavyweight: ~11% incidence in pooled prospective trials; OR 6.29 vs lamotrigine. Lamotrigine and carbamazepine are also implicated, with valproate-induced hair loss being typically diffuse, non-scarring, dose-related, and reversible after discontinuation. Cosmetic side effects have a "high intolerance rate leading to poor therapeutic compliance" — meaning patients quietly stop the drug rather than report the cosmetic concern. The clinical implication is to ask directly.

Cognitive effects. Topiramate is the most notorious ("dopamax" — slowed processing, word-finding difficulty, attention problems); zonisamide, phenobarbital, and high-dose valproate also impair cognition. Lamotrigine, levetiracetam, and oxcarbazepine are cleaner cognitively. Valproate exposure during pregnancy carries cognitive consequences for the offspring (the IQ deficit data is now multi-decade and consistent — the reason valproate is the most-restricted ASM in women of childbearing age in most regulatory regimes).

Behavioural effects. Levetiracetam is cognitively clean but causes irritability, aggression, and depression in roughly 5–15% of users — "Keppra rage" is real and the most common reason for switching off the drug. Perampanel carries a black-box warning for psychiatric side effects (homicidal/suicidal ideation, aggression). Phenobarbital remains in widespread global use partly because it is cheap, despite its sedation, cognitive, and behavioural burden.

SJS/TEN risk. The aromatic anticonvulsants — carbamazepine, lamotrigine, phenytoin, oxcarbazepine, phenobarbital — carry the Stevens-Johnson syndrome / toxic epidermal necrolysis risk that the library's Tegretol/Epitec SJS briefing covers in depth. The slow lamotrigine titration protocol exists specifically to prevent most reactions. Levetiracetam is the standard non-aromatic alternative when SJS risk is a primary concern.

Newer agents. Brivaracetam (cleaner than levetiracetam behaviourally), lacosamide, perampanel, and cenobamate represent the post-2010 generation. Cenobamate is currently the most efficacious novel ASM for focal seizures but requires very slow titration to avoid DRESS syndrome.

When medication isn't enough: surgery for hippocampal sclerosis

The single most important piece of under-applied epilepsy knowledge is that mesial temporal lobe epilepsy with hippocampal sclerosis is highly surgically tractable. Anterior temporal lobectomy delivers Engel-class-I seizure-freedom (the highest tier — completely seizure-free or only auras) in ~73–84% of well-selected patients at long-term follow-up. The 2025 ScienceDirect 10-year cohort reports 73.6%; the Frontiers 621-patient study reports 83.7% at 8.7-year average follow-up. The Wiebe RCT (NEJM 2001) — one of the cleanest pieces of evidence in epileptology — randomised refractory TLE patients to surgery vs continued medical management and found 58% vs 8% seizure-free at one year.

Despite this, most surgical candidates are referred late or never. The widely-cited threshold is "two failed adequately-trialled ASMs" — at that point, drug-refractory epilepsy is statistically much more likely to remain drug-refractory than to respond to a third or fourth medication, and surgical evaluation is appropriate. The threshold is widely under-applied; patients spend years on medication regimens with diminishing returns when surgical referral could have delivered seizure-freedom much earlier. Newer options — selective amygdalohippocampectomy, MRI-guided laser interstitial thermal therapy (LITT), responsive neurostimulation (RNS), deep brain stimulation of the anterior nucleus of the thalamus — give more graduated choices but ATL remains the highest-evidence intervention for the right anatomy.

SUDEP — rare, real, and now thought to be primarily respiratory

Sudden Unexpected Death in Epilepsy is the leading cause of epilepsy-specific mortality. Incidence is approximately 1 per 1,000 patient-years overall, rising to 9 per 1,000 in drug-resistant epilepsy. The proximal trigger is almost always a generalised tonic-clonic seizure. The mechanism is now believed to be primarily respiratory rather than cardiac: post-ictal central + obstructive apnea, autonomic instability, brainstem dysfunction (the pre-Bötzinger complex and nucleus tractus solitarius are repeatedly implicated), leading to respiratory arrest and secondary cardiac arrest. The 2024 PMC review puts it squarely: "increasing support for respiratory depression and decreasing evidence for tachyarrhythmia-induced SUDEP." Genetic substrates (SCN1A, KCNQ1) and neurotransmitter imbalances (serotonin, GABA) modulate risk.

Modifiable risk factors: prone position during sleep (the most replicated), frequency of generalised tonic-clonic seizures (the strongest), non-adherence to medication, alcohol use, lack of supervision at night. AAN/AES practice guidance (2017, still current) is that all epilepsy patients should be told about SUDEP — most clinicians don't. The professional society position is that hiding the risk to spare patient anxiety is not justified by the evidence; informed patients tend to engage more with the modifiable risk factors, not less.

The autonomic dimension

The bundle includes a PDF specifically on epilepsy and the autonomic nervous system because autonomic involvement in seizures is broader than non-specialists realise. The 2017 ILAE classification recognises autonomic seizures as a focal subtype: ictal tachycardia, bradycardia, asystole, piloerection, gastric rising, flushing, sweating, pallor, urinary urgency. Some autonomic seizures (especially Panayiotopoulos syndrome in children — early-onset benign occipital epilepsy with autonomic features) are mistaken for syncope, gastroenteritis, or behavioural episodes for years.

Autonomic involvement also matters peri-ictally and post-ictally — heart-rate variability changes are detectable in the periods around seizures and have been investigated as biomarkers for seizure prediction and SUDEP risk stratification. The autonomic imbalance characteristic of chronic uncontrolled epilepsy (sympathetic predominance, parasympathetic withdrawal) is one of the mechanisms invoked to explain elevated cardiovascular morbidity and SUDEP risk in this population.

Lived experience: Sarah J. Blake, 2001

The bundle anchors in a 2001 piece by Sarah J. Blake titled "Living with Pseudoseizures." The disclaimer at the top notes that the term "pseudoseizure" was used because it was prevalent in medical literature and is often the diagnostic term given to people; the author and much of the community now prefer "psychogenic non-epileptic seizures." (DSM-5 went further and located the condition within functional neurological symptom disorder.)

The piece describes Blake's friend "Melanie" (a pseudonym), her experience of repeated PNES events, the stigma she encountered in emergency departments and psychiatric wards (being shouted at as a malingerer, restrained until she was bruised, told by therapists they did not treat seizure disorders), and the slow process of identifying triggers (rage, fear, panic) and learning to recognise auras (a specific headache, a particular mood-shift). It describes deep breathing as occasionally helpful, hypnosis as suggested but not tried, and counselling that addresses the underlying emotional content as the actual treatment. It anticipates almost everything the modern PNES literature now formalises — including the central insight that "calling the seizures what they are is not enough" if no work is done on the substrate.

The piece holds up almost perfectly against 2024–25 clinical reviews. The vocabulary has changed. The mechanism description has gained the predictive-processing / active-inference framework. The treatment evidence has accumulated (CODES, the current FND multidisciplinary model). The patient experience — the stigma, the dismissal, the years of misdiagnosis, the slow grinding work of identifying triggers and learning to feel emotions safely — is remarkably unchanged. That is a finding too: the literature has caught up with what patients were saying twenty-five years ago, and the bedside has caught up much less than the literature has.

Sources

• ILAE 2025 — Updated classification of epileptic seizures (Beniczky et al.)
• ILAE 2017 — Operational classification of seizure types (Fisher et al.)
• Kwon et al. 2024 — Epilepsy and psychiatric comorbidities meta-analysis (PubMed)
• Medscape — Epilepsy Tied to Increased Risk for Psychiatric Comorbidities (Dec 2024)
• Cleveland Clinic Journal — Psychogenic nonepileptic seizure: an empathetic, practical approach (2022)
• StatPearls — Psychogenic Nonepileptic Seizures
• Goldstein 2020 — CODES trial (Lancet Psychiatry)
• PNES with concurrent epilepsy — Acta Epileptologica review
• Benson 1991 — The Geschwind syndrome (PubMed)
• Rethinking Geschwind Syndrome Beyond Temporal Lobe Epilepsy (2021)
• Geschwind Syndrome in Frontotemporal Lobar Degeneration
• Waxman & Geschwind 1974 — Hypergraphia in TLE (Neurology)
• Antiseizure Medication-Induced Alopecia: A Literature Review (2023)
• Adverse effects of ASMs — pharmacogenetics & interactions (Frontiers 2025)
• Long-term seizure outcomes of mesial TLE surgery — 10-year follow-up (2025)
• Wiebe 2001 — Effectiveness and Efficiency of Surgery for TLE (NEJM)
• The Crucial Interplay Between Lungs, Brain, and Heart in SUDEP (Brain Sci 2025)
• SUDEP: respiratory vs cardiac contributions (PMC 2024)
• Sarah J. Blake — Living with Pseudoseizures (2001), bundle source PDF