Cartography of Collapse: Mapping the Architecture of SUDEP

Abstract

Seizure-Related Biomarkers of Sudden Unexpected Death in Epilepsy (SUDEP) in Drug-Resistant Focal Epilepsy (REPO2MSE): A Prospective, Multicentre Case-Control Study

Ryvlin P, Huot M, Valton L, Maillard L, Bartolomei F, Derambure P, Hirsch E, Michel V, Chassoux F, Petit J, Crespel A, Biraben A, Navarro V, Kahane P, De Toffol B, Thomas P, Rosenberg S, Bernini A, Charlois AL, Craciun L, Chorfa F, Ducouret P, Ferreira A, Leclercq M, Marty M, Mercedes Alvarez B, Sampaio M, Spahr A, Timestit-Kurland N, Touya M, Roy P, Rheims S; REPO2MSE study group. Lancet Neurol. 2026;25(1):50-60. doi: 10.1016/S1474-4422(25)00379-5. Epub 2025 Nov 21. PMID: 41285145.

Background: Novel biomarkers of the risk of sudden unexpected death in epilepsy (SUDEP) are needed to better inform people with epilepsy of their individual risk and identify those at a high risk. The aim of this study was to identify such biomarkers, particularly with the exploration of seizures characteristics that have not been previously investigated, including peri-ictal peripheral oxygen saturation (SpO₂) and site of seizure onset. Methods: We conducted a nested case–control study of SUDEP within a dedicated nationwide prospective cohort. Eligible participants were adults with drug-resistant focal epilepsy undergoing in-hospital seizure monitoring at 16 epilepsy monitoring units in France. Clinical data, results from presurgical investigations, and raw recordings from video electroencephalogram, electrocardiogram, and SpO₂ were collected until the end of the recruitment period. The French National Directory of Natural Persons Identification was queried annually to identify deaths. The SUDEP cases were adjudicated on the basis of medical records and interviews documenting the circumstances of death. Each SUDEP case was matched to 4 controls on the basis of study center and date of inclusion. The SUDEP risk factors were identified using LASSO-penalized conditional logistic regression. Findings: From May 18, 2010, to August 23, 2015, we enrolled a total of 1074 participants and followed their vital status until the end of 2018, yielding a total of 6828 patient-years of follow-up. A total of 42 participants died during follow-up, including 18 cases of definite or probable SUDEP, resulting in a SUDEP rate of 2.64/1000 patient-years (95% CI 1.36-3.92). Four risk factors were significantly associated with the risk of SUDEP: an extratemporal epileptogenic zone (OR 37.8, 95% CI 3.21-446.2, P = .0039), a BMI of 30 or higher (26.0, 2.0-339.6, P = .013), male sex (12.6, 1.5-106.8, P = .0201), and predominantly nocturnal seizures (6.0, 1.2-28.7, P = .026). In contrast, the presence of peri-ictal SpO₂ of less than 80% during focal seizures, the frequency of focal-to-bilateral tonic-clonic seizures, heart rate variability, age at epilepsy onset, number of antiseizure medications, and history of depression were not significantly associated with SUDEP. Interpretation: Extratemporal epilepsies involving the perisylvian region or frontal lobe appear to be associated with an increased risk of SUDEP. This finding warrants confirmation in larger cohorts and underscores the need to improve the diagnosis and surgical management of extratemporal epilepsies, which might contribute to improved SUDEP risk stratification and prevention.

Commentary

At the turn of the twentieth century, Gowers described sudden death in epilepsy as one of the enduring enigmas of clinical neurology. ¹ More than a century later, sudden unexpected death in epilepsy (SUDEP) remains precisely that: statistically predictable yet individually unfathomable. We have grown fluent in reciting its risk factors: convulsive seizures, frequency, and medication nonadherence. We speak of safety devices. We counsel. We count. Yet when families ask, “Why?”, the arithmetic feels hollow.

The REPO2MSE study by Ryvlin and colleagues reframes that question with understated audacity. ² In this nationwide, prospective, meticulously phenotyped study of drug-resistant focal epilepsy (DRE) across 16 French centers, investigators identified 18 SUDEP cases. ² The most striking independent association with SUDEP was not seizure frequency but the presence of an extratemporal epileptogenic zone, particularly perisylvian and frontal localizations. ² Extratemporal onset conferred ∼38-fold higher odds of SUDEP, occurring in 83% of cases versus 33% of controls, ² a striking and largely uncharted signal, previously suggested in only one smaller study. ³

For decades, we have conceptualized SUDEP through the lens of seizure burden and severity. Our narrative has been largely convulsion-centric: generalized tonic-clonic seizures lead to hypoxemia, apnea, and ultimately cardiac arrest. ⁴ This model gained empirical traction from landmark work such as MORTEMUS, which documented terminal apnea preceding asystole in SUDEP. ⁴ Yet this study complicates this script. Strikingly, focal-to-bilateral tonic-clonic seizure frequency, long considered the cornerstone of SUDEP risk, did not retain independent significance in the final model. ² This does not negate the well-established relationship between convulsive seizures and SUDEP. ⁵ More likely, it reflects the limited power of the current study. ²

Likewise, oxygen desaturation below 80%, long suspected as a key discriminator, did not distinguish SUDEP from controls. ² This finding aligns with the contemporaneous US prospective cohort by Ochoa-Urrea, which similarly failed to confirm hypoxemia as a stand-alone predictor, though it implicated prolonged ictal and postictal central apnea as a SUDEP risk. ⁶ The REPO2MSE lacked respiratory belt data and could not directly interrogate central apnea. ² Still, the convergence is provocative. If oxygen desaturation is not decisive, the fatal cascade may lie in failed recovery, an orchestrated process governed by cortical and autonomic networks rather than oxygen levels alone.

Frontal and perisylvian regions sit at the crossroads of these networks. The insula, operculum, frontocentral cortex, and the cingulum, regions disproportionately overrepresented in the SUDEP group, ² are deeply embedded in the central autonomic network. ⁷ Extratemporal epilepsies more often generalize, and frontal seizures are more often nocturnal. ² But seizures arising from these territories may do more than generalize; they may transiently destabilize the circuitry required for postictal recovery. In this context, extratemporal epilepsy emerges as a marker of network fragility.

This provocative reframing shifts SUDEP from a mere complication of “bad seizures” to a manifestation of vulnerable architecture. Two patients may experience convulsive seizures of similar frequency, yet the one whose seizures originate from insulo-opercular networks may confront a different physiological aftermath.

The implications extend beyond pathophysiology into clinical practice. The epileptogenic zone has traditionally informed surgical candidacy and predicted seizure freedom. Mortality counseling, by contrast, has remained anchored to frequency and nocturnal occurrence. REPO2MSE suggests that these domains cannot remain separate. If localization independently contributes to SUDEP risk, it must be included in the conversation with patients.

Localization is also potentially modifiable. Resective surgery and neuromodulation can reshape network topology. Yet extratemporal epilepsies, particularly those involving eloquent cortex or posterior insula, are often surgically complex and sometimes deferred. If these epilepsies carry heightened mortality risk, the ethical calculus of “watchful waiting” becomes more difficult to justify when the outcome at stake is not only seizure freedom but also survival.

Importantly, REPO2MSE also reaffirms established SUDEP risk factors while repositioning them. As in prior studies, ⁵ male sex, BMI, and predominantly nocturnal seizures remain independently associated with SUDEP. The confirmation of obesity as a risk factor echoes prior population-based analyses ⁸ and invites mechanistic speculation: impaired repositioning from prone posture, exacerbated sleep-disordered breathing, and reduced cardiorespiratory reserve. Yet the Sleep Apnea Scale did not independently predict SUDEP in REPO2MSE, underscoring the limitations of surrogate measures. ² Nocturnal seizures occur in a state of diminished supervision and altered arousal thresholds. The association between SUDEP and male sex is similarly well-established. ⁵ Sex differences in autonomic tone, sleep architecture, and hormonal modulation of cortical excitability likely intersect with seizure networks, amplifying vulnerability. ⁹ When these factors converge with autonomic network fragility, risk may become multiplicative.

Equally revealing are variables that did not survive multivariable scrutiny. Interictal heart rate variability, long proposed as a marker of autonomic instability, was not associated with SUDEP, echoing previously conflicting data.^2,10 Neither age at epilepsy onset, number of antiseizure medications, nor history of depression emerged as SUDEP predictors. These findings suggest that static markers of disease severity or treatment burden may be less informative than dynamic network properties. Resting autonomic tone may not capture ictal vulnerability. Depression, though clinically consequential, may not independently drive the fatal cascade.

Of course, caution is warranted. Eighteen SUDEP cases cannot settle the matter. Wide confidence intervals around the odds ratios signal statistical imprecision. Extratemporal epilepsies may correlate with unmeasured mediators of SUDEP risk, such as sleep architecture and unrecognized seizure clustering. The cohort was restricted to adults with focal DRE. Generalizability to generalized epilepsies or pediatric populations remains uncertain. Replication in larger, international cohorts is imperative.

But conceptual progress does not require statistical perfection. The coherence between localization and fatal collapse warrants attention. Identifying the epileptogenic zone as a potentially modifiable biomarker introduces a structural dimension to SUDEP risk stratification. It compels us to ask whether extratemporal epilepsies merit expedited evaluation. It raises the possibility that neuromodulation might confer differential benefits in anatomically high-risk networks.

Perhaps the deeper lesson is epistemological. Our traditional SUDEP risk framework is built on measurable, yet incomplete variables. Frequency, sex, and nocturnality are accessible, but they may obscure subtler dimensions of vulnerability embedded within neural circuits. Localization offers a structural lens through which to reinterpret risk in ways previously underappreciated.

For clinicians, this raises uncomfortable but necessary questions. Should patients with extratemporal epilepsies receive different counseling? Should presurgical evaluation be expedited in this subgroup? Should localization be incorporated into emerging SUDEP risk calculators? And if future studies confirm these findings, will we recalibrate our intervention thresholds?

Gowers grappled with the mystery of sudden death in epilepsy. ¹ The REPO2MSE does not resolve this mystery but shifts the vantage point. It suggests that within the cortex's convolutions, certain territories carry dire consequences. SUDEP risk may be encoded not only in seizure frequency but also in the cortical cartography. For years, we have asked how often seizures occur. We may now need to ask, with equal seriousness and candor, where they begin.