19

Evaluating the clinical utility of the Refractory Epilepsy Screening Tool for Lennox-Gastaut Syndrome (REST-LGS) in a specialised adult epilepsy and intellectual disability clinic

Hannah Humble¹, Vivienne Evans², Dr Rhys Thomas^3,4

¹Newcastle University, ²Clinical Research Department (Neurology), Royal Victoria Infirmary, ³Translational and Clinical Research Institute, Royal Victoria Infirmary, ⁴Neurosciences, Royal Victoria Infirmary

Background: Lennox-Gastaut Syndrome (LGS) is an electroclinical syndrome with diagnostic criteria ratified by the ILAE. However, many adults are not diagnosed due to a lack of recognition, or uncertainty in adult teams. The Refractory Epilepsy Screening Tool (REST-LGS) was created to help with identification. The prevalence of LGS worldwide is suggested to be at least one million people. Screening tools are a cornerstone in preventative medicine, when used in the correct setting and their caveats understood. The aim of this study is to evaluate the clinical utility of REST-LGS and make recommendations for improving its’ specificity. 

Methods: We retrospectively collected data from electronic patient records who attended a specialised adult clinic (March 2023 to 2026, to account for follow up). REST-LGS has 4 major criteria: ≥2 seizure types, onset before age 12, generalised slow spike-and-wave discharges on EEG and cognitive impairment, and 4 minor criteria: seizure persistence, helmet use, history of VNS, ketogenic diet or surgery and other specific EEG findings. The outcome is ‘likely’ if three of the major and at least two minor criteria are met or ‘unlikely’. Gold-standard diagnoses were based on ILAE criteria and where this was not possible, clinical experience with EEG support.

Results: 495 individuals were identified for screening, 122 were excluded due to limited information or failed clinic attendance. Of the 373 included, 56 patients had a gold-standard diagnosis of LGS, and 16 patients had Dravet syndrome. REST-LGS suggested an outcome of ‘likely’ for 47.45% (n=177) and ‘unlikely’ for 52.55% (n=196). For those with a known LGS diagnosis the tool suggested 98.21% (n=55) were ‘likely’ and 1.79% (n=1) ‘unlikely’. All known LGS patients had ≥2 seizure types, onset before age 12 and persistent seizures. The sensitivity of REST-LGS in this population was 98.21%, however the specificity was 61.51%; positive predictive value of 31.1%. This is shown as those with confirmed Dravet syndrome (n=16), all identified as ‘likely-LGS’. 

Conclusion: We must review the 122 ‘likely’ patients to identify any missed cases, or if the tool is over-identifying. However, an ‘unlikely’ result is very helpful with a negative predictive value of 99.5%. REST-LGS could be used to prompt referral to a specialist epileptologist. However, the lack of specificity limits its use as a standalone tool. Solutions targeting improved specificity could include inclusion of cognitive specific criteria highlighting regression or using negative marking if another diagnosis is known or equally as likely.

20

Why the way we measure deprivation in studies of people with epilepsy matters

Kathryn Bush¹, Dr Rhys Thomas¹, Andrew Kingston¹, Sheena E Ramsay¹, Owen Pickrell²

¹Newcastle University, ²Swansea University

Introduction: Epilepsy is associated with socioeconomic deprivation in the United Kingdom, with the highest rates observed in the most deprived populations. In large population studies area-level measures of deprivation are frequently available and widely used. This study aims to understand how different measures of socioeconomic deprivation influence population-level risk estimates for incident epilepsy.

Methods: We created a Welsh Cohort study of 2.4 million individuals with linked medical, death, area-level demographic and census data, in the Secure Anonymised Information Linkage (SAIL) databank. Individuals were followed up between 2011-2024 and there were 17,455 all-age incident epilepsy cases observed. Measures of deprivation were recorded at study baseline, prior to epilepsy diagnosis. We examined the correlation between area-level and household-level measures of deprivation at a population-level. We used Cox-proportional hazard ratios to compare the risk of epilepsy in the most deprived versus least deprived groups, using: i) census household-deprivation scores, ii) individual-level factors (education and occupation), and iii) Welsh Index of multiple deprivation (WIMD) area-level measures. 

Results: All methods used showed graded associations between socioeconomic deprivation and incident epilepsy in Wales, with the highest rates of incident epilepsy in the most deprived populations. Area-level and household-levels of deprivation were highly correlated (p <0.001), but area-level measures will mis-classify some individuals. Models were adjusted for age, sex and rural-urban classification. The all-age adjusted hazard ratio (AHR) of epilepsy in the most deprived areas versus least was AHR:1.66 (1.55-1.77). In the most deprived households versus least all-age AHR:3.20 (2.50-4.10). Adults (16-64 years) with no educational qualifications versus level-4 (degree) AHR 2.23 (2.01-2.47). Adults in occupational social-class 8 (most deprived) had an AHR 4.23 (3.55-5.04) compared to social-class 1. 

Conclusions: In this population-level study, individual-level and household-level deprivation measures produce higher incident epilepsy risk estimates, compared to area-level measures. Studies using area-level measures of deprivation may under-estimate deprivation associated risk. Studies seeking to understand the influence of socioeconomic factors on epilepsy incidence or outcomes, should consider using household-level or individual-level measures (available in census-data), in preference to area-level data.

21

Verbal memory recall and retrieval in Temporal Lobe Epilepsy

Amrita Gupta^1,2, Carissa Kelley^1,2, Marzena Arridge^2,3, Sherry Liu^1,2, Hannah Bergman³, Andrea Hill³, Christopher Chua Yuan Kit⁴, John Sidney Duncan^1,2,3, Sallie Baxendale^1,2, Meneka Kaur Sidhu^1,2,3

¹UCL Queen Square Institute of Neurology, ²Department of Epilepsy, National Hospital for Neurology and Neurosurgery, ³MRI Unit, Chalfont Centre for Epilepsy, ⁴Division of Neurology, University Medical Centre, National University Hospital

Introduction: Temporal lobe epilepsy (TLE) affects cognitive networks and is associated with episodic memory impairment. We previously described reorganization of memory encoding networks in TLE using functional magnetic resonance imaging (fMRI). The impact of TLE on verbal memory retrieval networks are less well understood. We investigated verbal memory retrieval networks in left (LTLE) and right TLE (RTLE) compared to healthy controls (HC). 

Methodology: 26LTLE (median age 36.5 (32.0–48.2), 29RTLE (37(30–49)) and 25HC (40(35–44)) underwent verbal memory fMRI encoding, recall and recognition paradigms with a baseline of odd/even number judgement. BIRT Memory and Information Processing Battery (BMIPB-II) was used to assess verbal learning (VL) and delayed recall (DR) with Mann-WhitneyU tests on age-normed z-scores for group comparisons. 

SPM 12 was used for one-sample t-tests to assess within group activations; two-sample t-tests for group differences. Extra-temporal activations are reported at p<0.01 uncorrected (specific subtraction contrasts) with small-volume correction for medial temporal lobe (MTL) (FWE, p<0.05). 

Results: Neuropsychology: LTLE performed significantly worse than RTLE on both VL and DR; VL (median z = −1.18 vs −0.67; U = 226.5, p = .011) and DR (median z = −1.44 vs −0.42; U = 201.0, p = .003). Both patient groups performed worse than controls. 

Encoding: Controls showed L>R parahippocampal gyrus (PHG) activations. Compared to controls, both LTLE and RTLE had reduced neocortical activations bilaterally (LTLE: left middle temporal, middle frontal, bilateral supramarginal, RTLE: right lingual, cuneus). 

Free Recall: Controls showed right PHG, entorhinal, left insula, left thalamus, right middle temporal and inferior frontal gyrus activations. LTLE showed greater bilateral fusiform, left insula, superior temporal activations compared to controls. 

Recognition: Controls showed right hippocampus, PHG, right middle occipital activation. LTLE showed increased activation in the right hippocampus and PHG, left insula, left superior temporal and right fusiform gyrus compared to controls. 

RTLE showed reduced right hippocampus and PHG gyrus activations compared to controls for both recognition and recall. 

Conclusions: In LTLE, both recall and recognition engaged greater left insula and right-MTL network (including the fusiform gyrus) more than HC. For recall this was more extra-temporal reorganisation whilst recognition engaged the contralateral MTL. In RTLE, right MTL activation was reduced for both recognition and recall compared to controls but relatively preserved at encoding. These findings suggest recall and recognition rely on distinct networks that are altered in both LTLE and RTLE, with corresponding deficits in neuropsychological memory scores seen in both groups.

22

Both Face Encoding and Recognition Network Disruption is Seen in Left and Right Temporal Lobe Epilepsy

Carissa Kelley^1,2, Amrita Gupta^1,2, Marzena Arridge^2,3, Sherry Liu^1,2, Hannah Bergmann³, Andrea Hill³, Christopher Chua Yuan Kit⁴, John S. Duncan^1,2,3, Sallie Baxendale^1,2, Meneka K. Sidhu^1,2,3

¹UCL Queen Square Institute of Neurology, ²Department of Epilepsy, National Hospital for Neurology and Neurosurgery, ³MRI Unit, Chalfont Centre for Epilepsy, ⁴Division of Neurology, University Medical Centre, National University Hospital

Purpose: Visual episodic memory difficulties occur in primarily non-dominant temporal lobe epilepsy (TLE). Functional magnetic resonance imaging (fMRI) has demonstrated visual memory encoding network reorganization in TLE, but retrieval networks remain poorly understood. We investigated face encoding and recognition network reorganization in left TLE (LTLE) and right TLE (RTLE).

Methods: 25 controls and 55 TLE patients (26 left) performed a neutral and fearful face encoding and recognition task with a baseline task of odd/even number judgement. SPM12 was used to investigate within-group activations (one-sample t-test) during encoding and recognition and between-group differences (two-sample t-tests). Group differences in BIRT Memory and Information Processing Battery (BMIPB II) design learning (DL) and delayed recall (DR) were assessed using Mann-Whitney U tests on age-normed z-scores.

Mesial temporal activations are reported using small volume correction (6mm) at FWE p<0.05 and neocortical activations at p<0.01, as stringent subtraction contrasts were used.

Results: There was no significant difference between LTLE and RTLE for DL or DR; LTLE (median: -0.89, IQR: -1.32 to -0.51) and RTLE (-0.73, -1.21 to 0.17) for DL, LTLE (0, -0.95 to 0.48) and RTLE (-0.57, -1.49 to 0.65) for DR.

Encoding: Both patient groups showed fewer mesial temporal lobe (MTL) activations (anterior parahippocampal gyrus/amygdala) than controls. RTLE showed increased right posterior hippocampal and increased left neocortical activations, and LTLE showed reduced left neocortical and increased right neocortical activations compared to controls.

Recognition: Both patient groups showed greater fusiform and anterior cerebellum activations than controls, and RTLE showed bilateral posterior cerebellar activations exceeding both other groups. RTLE showed increased bilateral activations in the MTL with additional bilateral neocortical activations. LTLE showed reduced left inferior parietal lobule/supramarginal gyrus activation and increased right posterior hippocampal activation and additional predominantly right neocortical activations. 

Key default mode network hubs (posterior cingulate cortex/precuneus) were more active in RTLE compared to controls or LTLE during both encoding and recognition.

Recognition vs Encoding: RTLE had greater cerebellar activation during recognition than encoding, while LTLE showed increased left posterior hippocampal activation with no cerebellar increase.

Conclusions: These findings suggest face encoding and recognition networks are disrupted in TLE. LTLE showed reduced left and increased right hemisphere activations during encoding and recognition, compared to controls. RTLE showed more extensive bilateral reorganization, including default mode and cerebellar regions, with cerebellar activation greatest during recognition. As the cerebellum contributes to emotion and visual recognition, this pattern may reflect compensatory efforts for network dysfunction.

23

Mortality and comorbidities in people with Functional/Dissociative Seizures in Wales

Russell Khan¹, Beata Fonferko-shadrach¹, Huw Strafford¹, Kathryn Bush⁴, Owen Bodger¹, Kelly Price², Kathryn Cullen¹, Inder Sawhney¹, Ronan Lyons¹, Arron Lacey¹, Ashley Akbari¹, Joe Anderson, Phil Smith⁵, Teja Zeribi⁷, Bridget Mildon⁷, Rob Powell^1,2, Markus Reuber³, Owen Pickrell^1,2

¹Swansea University Medical School, ²Neurology Department, Morriston Hosptial, Swansea Bay University Health Board, ³School of Medicine and Population Health, Academic Neurology Unit, University of Sheffield, ⁴Population Health Sciences Institute, Newcastle University, ⁵Neurology Department, University Hospital Wales Cardiff, Cardiff and Vale University Health Board, ⁶Neurology Department, Royal Gwent Hospital Newport, Aneurin Bevan University Health Board, Newport., 7FND Hope International

Purpose: There is increasing evidence that people with Functional/dissociative seizures (pwFDS) have increased mortality and comorbidity risks. We aimed to characterise these using population-level, routinely collected data in Wales.

Method: Using a previously validated algorithm (sensitivity=47.7%, specificity=99.8%), within the SAIL Databank, we identified people with FDS (pwFDS) in Wales diagnosed between 2003 and 2023. We matched (deprivation, age and sex) each pwFDS to three cases: with epilepsy and without FDS (epilepsy cohort); and without epilepsy or FDS (non-seizure comparators). We compared socioeconomic measures, mortality, comorbidities and healthcare utilisation in pwFDS, people with Epilepsy (pwE), and non-seizure comparators. 

Results: There were 2857 pwFDS (8571 people in both epilepsy and non-seizure comparator cohorts). 75% of pwFDS were female, 29% had epilepsy, and 27% were from the most deprived area (quintile). For the FDS, epilepsy, and non-seizure comparator cohorts: 76%, 53% and 28% had a common mental illness; 10%, 7% and 1.2% had a severe mental illness; 28%, 26%, and 57% were currently working; 35%, 23%, and 7% had long-term sickness.

11%, 18%, and 6% of the FDS, epilepsy and non-seizure comparator cohorts died during the study. Corresponding mortality hazard ratios, for the FDS and epilepsy cohorts (when compared to non-seizure comparators) were: 2.0 (95%CI=1.7–2.4) and 2.4 (95%CI=2.2–2.7). When adjusted for common and serious mental health disorders the trend for higher mortality hazard ratios remained: HR= 2.0 (95%CI=1.7–2.3) and 3.5 (95%CI=3.1–3.8). pwFDS and pwE were more likely to die of: external, neurological or mental health causes when compared to people without seizure disorders. 

pwFDS, pwE and people with both epilepsy and FDS were estimated to have 3.8, 2.8 and 6.0 the costs of non-seizure comparators in terms of hospital admissions, emergency department attendance and outpatient appointments.

Significance: People with FDS have significant higher rates of psychiatric comorbidities, unemployment, mortality and healthcare utilisation when compared to age, sex and deprivation matched comparators. Mortality hazard ratios remained elevated for people with FDS even when adjusting for psychiatric comorbidities. It is important for healthcare professionals and service planners to recognise the burden of FDS. Further work is necessary to understand the causes of increased mortality and to prioritise treatments.

24

Vulnerability factors for non-epileptic seizures in patients with epilepsy – a retrospective cross-sectional study

Benjamin Sacks, Bonnie Chow, Akihiro Koreki, Rohan Kandasamy, Mahinda Yogarajah

National Hospital For Neurology & Neurosurgery

Background: Functional seizures occur in approximately 5–20% of people with epilepsy, a prevalence substantially higher than the general population. The reasons for this co-occurrence remain unclear and may relate to psychiatric comorbidity, medication exposure, or disease-specific vulnerability factors. 

Methods: We compared patients with a confirmed dual diagnosis of epilepsy and NES (n=131) with a control group of patients with epilepsy alone (n=131) treated at the National Hospital for Neurology & Neurosurgery and the Chalfont Centre for Epilepsy. Demographic and clinical characteristics were described, and differences in anti-seizure, anti-psychotic and anti-depressant medication exposure were assessed using propensity score matching and binary logistic regression. 

Results: There were no significant differences in seizure frequency, age of epilepsy onset and age at first anti-epileptic medication between groups. Abnormal MRI scans (p=<0.001) and epileptiform EEGs (p=0.007) were significantly correlated with the epilepsy only group. Females were significantly over-represented in the dual diagnosis group (73% vs 54%, p=0.001). The dual diagnosis group had significantly higher rates of psychological trauma (15% vs 5%, p=0.012) and mood disorders (76% vs 55%, p=<0.001). No significant anti-seizure medication associations were identified in either group. Antipsychotic prescriptions were significantly lower in the dual diagnosis group, compared to the epilepsy only group with an odds ratio of 0.21 [0.05 - 0.89] (p=0.034) after correction for medication indication & psychiatric comorbidities. 

Discussion: In this analysis, dual diagnosis was associated with female sex & history of mood or anxiety disorders. They appeared to have a lower incidence of focal epilepsy & abnormal, however, we hypothesise that this represents selection bias. It is reassuring to note that no anti-seizure medications are associated with the development of functional seizures in epileptic patients. In this study, antipsychotics were a protective factor against the development of functional seizures in patients with epilepsy. In a predictive processing model of functional neurological disorders, dopamine codes the misattribution of salience to otherwise benign stimuli. Our results in this study, where patients with epilepsy who received dopamine blocking agents had a lower incidence of functional seizures, are consistent with this model.

25

Impaired cerebrovascular reactivity to carbon dioxide indicates cerebral vasomotor dysfunction in patients with late onset unprovoked seizures and epilepsy

Josephine Mayer^1,2,3, Rehan Junejo^2,4, Helen Shantsila^1,2, Laura Bonnett¹, Gregory Lip^1,2, Tony Marson^1,3

¹University Of Liverpool, ²Liverpool Centre for Cardiovascular Sciences, ³The Walton Centre, ⁴Manchester Metropolitan University

Background: Late-onset unprovoked seizures and epilepsy (LOUE) is associated with an increased risk of stroke. Cerebrovascular reactivity to carbon dioxide (CVR CO₂) is used to index cerebrovascular health and cerebral vasomotor function. Impaired CVR CO₂ suggests a brain vulnerability to ischaemic events. We sought to investigate cerebrovascular health in LOUE; we hypothesised that LOUE have lower CVR CO₂ compared to controls, indicating reduced cerebral vasodilatory reserve and vasomotor dysfunction to explain stroke risk. Methods: Thirty LOUE participants without an identified cause were recruited from a tertiary Neurology centre (age = 60 ± 9 [mean ± standard deviation] years) and were compared to twenty-one hypertensive (HTN; 62 ± 8 years) and twenty-one healthy (59 ± 9 years) controls. Participants with prior stroke were excluded. Transcranial doppler recorded middle cerebral artery blood velocity (MCAVm) at rest (five minutes) and during step increase (4 minutes) and decrease (2 minutes) in end tidal CO₂(PETCO₂) with 5% CO₂ administration and hyperventilation. Cerebrovascular conductance index (CVCi) was defined as ratio of MCAVm and mean arterial pressure (MAP). CVR CO₂ was calculated as the linear slope for MCAVm and CVCi versus PETCO₂. QRisk3 score (a validated score for 10-year risk of myocardial infarction and stroke) was calculated for each participant. 

Results: Baseline MCAVm and CVCi were lower in LOUE (59.69 ± 12.89 cm.s−1; 0.55 ± 0.13 cm.s−1.mmHg−1) and HTN (59.22 ± 8.70 cm.s−1, 0.54 ± 0.10 cm.s−1.mmHg−1 ) versus healthy (64.64 ± 9.53 cm.s−1, 0.622 ± 0.14 cm.s−1.mmHg−1 ) but were not significantly different. MCAVm and CVCi CVR CO₂ were lower in LOUE (1.37 ± 0.43 cm.s−1, 0.012 ± 0.004 cm.s−1.mmHg−1) than HTN (1.49 ± 0.33 cm.s−1, 0.013 ±0.004 cm.s−1.mmHg−1) and healthy (1.68 ± 0.3 cm.s−1 [p = 0.015; one-way ANOVA], 0.015 ± 0.005 cm.s−1.mmHg−1 [p = 0.02]). Post-hoc analysis revealed significant differences between LOUE and healthy controls (p < 0.015). Stepwise multiple regression identified LOUE, increasing age and chronic kidney disease as statistically significant predictors of reduced MCAVm CVR CO₂ and CVCi CVR CO₂ . QRisk3 score was higher (indicating increased risk of vascular events) in LOUE and HTN compared to healthy controls (median (IQR) 14.2 (11.1), 15.4 (10.6), 9.0 (8.6); p=0.031 respectively).

Conclusion: LOUE was associated with impaired CVR CO₂ and elevated QRisk3 score compared to healthy controls. Vascular risk factors and cerebral vasomotor dysfunction may contribute to stroke risk in this group and may be a treatment target to reduce stroke risk.

26

Network structure of epilepsy and SUDEP-related risk in a real-world digital self-assessment cohort

Jake Ahern^1,2, Sammy Ashby³, Ben Donovan³, Peter Hannon⁴, Lance Watkins^2,5,6, Richard Laugharne^2,10, Brendan Mclean^2,7,8, Rohit Shankar^2,8, John Terry^2,9

¹University Of Birmingham, ²CIDER, University of Plymouth, ³SUDEP Action, ⁴Suvo Limited, ⁵University of South Wales, ⁶Mental Health and Learning Disabilities Service Group, Swansea Bay University Health Board, ⁷Cornwall Partnership NHS Foundation Trust, ⁸Neurology Department, Royal Cornwall Hospitals NHS Trust, ⁹Neuronostics Ltd., ¹⁰Department of Intellectual Disability Neuropsychiatry, Research Team, Cornwall Partnership NHS Foundation Trust

Background: Sudden unexpected death in epilepsy (SUDEP) and epilepsy related mortality risk is multifactorial, dynamic, and spans seizure control, mental health, substance use, and medication and healthcare access. Here, the structure of epilepsy and SUDEP-related risk factors captured by EpSMon (an epilepsy self-assessment risk app) is characterized using network analysis. We determine whether the digital self-assessment data contains clinically meaningful risk structure.

Methods: For each risk factor captured by EpSMon, multivariable mixed-effects logistic regression models estimated pairwise associations with all other variables. A directed, weighted network was constructed from associations after false discovery rate correction and backbone extraction. Network community structure, node centrality, bridging factors, and the directed subnetwork of pathways leading to worsening generalised tonic-clonic seizures (GTCS) were analysed.

Results: The final cohort comprised 337 users contributing 1569 self-assessments (median 4 assessments; mean age 37 years; 68% female). Network analysis revealed five stable risk communities corresponding to seizure severity, mental health, substance use, healthcare utilisation, and comorbid medical conditions. Anxiety and depression were the most broadly influential predictors of other risk factors. Substance use variables were the most strongly predicted nodes in the network. Nocturnal seizures were a prominent predictor of worsening GTCS after multivariable adjustment; other seizure-related factors contributed through indirect pathways.

Conclusion: Epilepsy and SUDEP-related risk factors captured by routine digital self-assessment form a structured system with clinically interpretable communities, distinct functional roles, and indirect pathways to convulsive seizure deterioration. These findings are hypothesis-generating and support the prospective evaluation of more dynamic approaches to digital risk assessment in epilepsy. Validation in independent cohorts and co-design with patients and clinicians are required before these structural findings can inform clinical tools.

27

Characteristics and themes of pregnancy-related SUDEP: a registry-based case series using the Epilepsy Deaths Register

Jiwon Kim¹, Alex Grundmann^2,3, Ben Donovan⁴, Dr Rhys Thomas^2,3

¹University Hospital Of North Tees, ²Translational and Clinical Research institute, Newcastle University, ³Royal Victoria Infirmary, ⁴SUDEP Action

Background: Pregnancy is a high-risk period for women with epilepsy, associated with adverse outcomes, including sudden unexpected death in epilepsy (SUDEP). National guidelines emphasise effective risk communication, specialist-led care, and optimisation of anti-seizure medication (ASM). However, real-world adherence to these standards remains unclear.

Aim: To identify recurring themes in pregnancy-related SUDEP and assess adherence to guideline-recommended care.

Methods: We conducted a retrospective analysis of pregnancy-related SUDEP cases from the SUDEP Action Epilepsy Deaths Register (2013–2025), which contains data derived from third-party reports provided by bereaved contacts. Quantitative analysis assessed adherence to guideline-recommended care domains where sufficient information was available, and qualitative analysis identified recurring themes in clinical management and risk communication.

Results: Thirteen pregnancy-related deaths were identified, all consistent with definite or probable SUDEP. Median age at death was 25 (range 17-37). The majority were cohabiting (85%), antepartum (85%), found at home (92%) and found prone (54%). Adherence to epilepsy care standards in pregnancy was variable, particularly in the domain of risk communication and awareness. Among 5 assessable cases, specifically referring to SUDEP risk communication, 4 cases did not meet the guideline recommendation and 1 case partially met it. The most recurrent theme from the narrative data, emerging in 9 out of 13 cases, was limited communication and awareness of SUDEP and pregnancy-related risks. Challenges surrounding ASM management were identified, including medication changes in the month prior to death (8/11), concerns about ASM (4/13) and increasing seizure frequency during pregnancy or following ASM changes (5/13). Most individuals were under a form of specialist care (11/12), had seen a specialist in the last 12 months (9/11) and prescribed ASM (9/11).

Conclusion: Despite the limitation of third-party derived data being susceptible to recall or reporting bias, these pregnancy-related SUDEP cases highlight potential areas for improvement in risk communication and medication management. Most individuals were under specialist care, prescribed ASM and recently reviewed, suggesting pregnancy-related SUDEP may occur even among women engaged with specialist services.

28

A System-Wide Approach to SUDEP Prevention: Implementation of a Safety Checklist and Workforce Training Programme in South Yorkshire

Anita Winter¹, Ben Donovan², Jane Hanna²

¹NHS South Yorkshire ICB, 2SUDEP Action

Background: Over 1,000 epilepsy-related deaths occur annually in the UK, with mortality rates around three times higher in people with epilepsy than the general population. Sudden Unexpected Death in Epilepsy (SUDEP) is a significant and potentially preventable cause of premature mortality, particularly among people with a learning disability and autistic people. Learning from the South Yorkshire LeDeR Programme and the NHS England STEP TOGETHER – Mind the Gap epilepsy benchmarking report identified variation in SUDEP awareness, inconsistent risk discussions and limited use of structured safety planning. 

Aim: To improve SUDEP awareness, standardise risk assessment, and support consistent seizure safety planning through workforce training and implementation of the SUDEP & Seizure Safety Checklist.

Methods: South Yorkshire ICB commissioned SUDEP Action to deliver a SUDEP & Seizure Safety Programme comprising multi-sector SUDEP awareness training, promotion of the adult and children’s SUDEP & Seizure Safety Checklists, and development of an accessible electronic SUDEP leaflet. Nine 90-minute training sessions were delivered between June 2025 and March 2026 using the ECHO model, incorporating clinical expertise and lived experience. Evaluation used a mixed-methods approach, combining quantitative pre- and post-training surveys with qualitative feedback to assess knowledge, confidence and intended practice change. 

Results: The programme engaged 1,135 participants across health, social care, education and VCSE sectors, demonstrating strong system-wide engagement and attracting interest beyond South Yorkshire. 

Evaluation demonstrated measurable improvements in self-rated knowledge and confidence. On a 1–7 scale, where 1 represented the highest level of knowledge/confidence, mean scores improved across all measures: knowledge of SUDEP (4.4 to 2.9), awareness of modifiable risk factors (4.5 to 3.0), awareness of good practice resources (4.6 to 2.9), and knowledge of good practice following a sudden epilepsy-related death (4.6 to 3.0). The proportion of respondents rating themselves within the top three categories increased from 28% pre-training to 71% post-training.

Qualitative feedback highlighted increased confidence in discussing SUDEP, awareness of risk-reduction strategies, and improved understanding of how to use the checklist to support person-centred conversations, shared decision-making and safety planning. SUDEP Action also reported increased clinician and organisational registrations for the SUDEP & Seizure Safety Checklists, providing evidence of engagement with structured risk-reduction tools.

Conclusion: This programme demonstrates an evidence-informed, system-wide approach to reducing SUDEP risk. Combining ECHO-based training with practical risk-reduction tools improved workforce knowledge, confidence and engagement with seizure safety planning. Early findings suggest a scalable model for embedding consistent, person-centred epilepsy safety practices across health and care systems.

29

Learning from Prevention of Future Death Reports in Epilepsy 

Julia Stirling, Ben Donovan, Jane Hanna

SUDEP Action

Introduction: At least 21 epilepsy-related deaths occur each week in the UK and young adults have the highest relative risk of epilepsy-related death, particularly from SUDEP. Up to 80% of deaths in young adults are thought to be preventable through improved epilepsy care. Prevention of Future Deaths (PFD) reports offer clear insights into failures in epilepsy care that increase risk but represent only the tip of the iceberg. Nevertheless, they are a red-flag for the true scale of preventable harm. 

Coroners have a duty to report where deaths could be prevented, but the process is shaped by structural inequalities. NHS Trusts are routinely legally represented, while, due to cost, bereaved families rarely are. This imbalance affects the evidence heard and learning captured. SUDEP Action’s free, family centred casework service helps address this gap, ensuring over 70 families a year are supported through complex investigations and inquests.

Methodology: A retrospective qualitative analysis was conducted with the following complimentary sources of information around learning from epilepsy-related deaths:

- Notes from SUDEP Action’s casework service

Six PFD reports since 2024 - Charlie Marriage, Andrew Ewin Ripp, Amber Walker, David Compton, Paul Nash, and John Fisher,

- Clive Treacey’s Independent Review

Key themes were drawn out independently by two researchers prior to further analysis.

Results: A cross-cutting theme across all deaths was a systemic failure to communicate with patients about their person-centred risk including risk of SUDEP. Furthermore, analysis highlighted consistent failures in epilepsy care including:

- delays in accessing regular and emergency medications

- appointment delays

- poor discharge information

- failures in primary care and pharmacy systems.

Conclusions: Findings of this study echo concerns that have been being voiced for over two decades. Key issues around person-centred communication of risk, access to medications and epilepsy support, and systematic learning from deaths, can be traced back to the Prevent21 “Time to Listen” consensus. Despite evidence from past leading to present, many issues continue to be left unaddressed, causing preventable risks to persist and putting people with epilepsy at risk of harm and potential premature mortality.

Furthermore, learning from deaths is still hampered by inconsistent use of Royal College of Pathologists guidelines and misattributed causes of death. The Hillsborough inspired Public Authority Accountability Bill would transform the PFD system by introducing a statutory duty of candour for public bodies and funded legal support for families at inquest.

30

Antiseizure Medication Monotherapy and Long-Term Mortality in Epilepsy 

Gashirai Mbizvo¹, Glen P. Martin², Matthew Sperrin², Josemir W. Sander³, Iain E. Buchan¹, Gregory Y.H. Lip¹, Anthony G. Marson¹

¹The University Of Liverpool, ²University of Manchester, ³UCL Queen Square Institute of Neurology

Background: People with epilepsy have an increased risk of premature death. Whether long-term mortality differs by initial antiseizure medication (ASM) selection is unclear, as prospective trials are typically underpowered to assess this. We therefore emulated a clinical trial of ASM monotherapy and mortality in newly diagnosed epilepsy using observational data. 

Methods: We emulated a target trial using international health data from TriNetX. We assessed patients with newly diagnosed generalised or unclassifiable epilepsy initiating lamotrigine, levetiracetam or valproate, and patients with focal epilepsy initiating lamotrigine, levetiracetam, carbamazepine or lacosamide. Pairwise cohorts were propensity score matched on baseline demographic, clinical, laboratory, medication and healthcare-utilisation covariates. Outcomes were ten-year all-cause death and unexplained death (ICD-10-CM R99). Cox models estimated hazard ratios (HRs) with 95% confidence intervals (CIs). Ten-year absolute risks and numbers needed to treat (NNTs) were derived from Kaplan–Meier survival probabilities.

Results: Among 144,055 patients with generalised or unclassifiable epilepsy, 98,735 initiated levetiracetam, 25,672 valproate and 19,648 lamotrigine. Compared with levetiracetam, lamotrigine was associated with lower hazards of all-cause death (HR 0.642 [CI 0.583-0.707]) and unexplained death (0.519 [0.426-0.632]). Ten-year all-cause mortality was 8.96% with lamotrigine and 13.2% with levetiracetam, corresponding to an NNT of 24; that is, for every 24 patients treated with lamotrigine rather than levetiracetam, one death was prevented over 10 years. Lamotrigine was also associated with lower mortality than valproate, and valproate associated with lower mortality than levetiracetam. For every 38 patients treated with valproate instead of levetiracetam, one death was prevented. Among 41,230 patients with focal epilepsy, 26,713 initiated levetiracetam, 7,819 lamotrigine, 5,118 carbamazepine and 1,580 lacosamide. Lamotrigine and carbamazepine did not differ. Compared with levetiracetam or lacosamide, lamotrigine and carbamazepine were associated with 40.0-49.5% lower hazards of all-cause death and 44.7-68.4% lower hazards of unexplained death. For every 18 patients treated with lamotrigine instead of levetiracetam, one death was prevented. The corresponding figure was 22 for carbamazepine against levetiracetam.

Conclusions: Initial ASM choice may have important implications for long-term survival in epilepsy. Lamotrigine had the lowest mortality hazards and should remain a preferred first-line treatment. By contrast, levetiracetam was the most frequently initiated ASM, yet was associated with higher mortality hazards than nearly all comparators. This is concerning given the marked global shift towards levetiracetam use. Clinicians should be cautious about defaulting to levetiracetam as initial monotherapy when appropriate alternatives are available. Lacosamide may also be a less favourable initial treatment when long-term mortality is considered.

31

Recognising Ictal Asystole: New Seizure semiology in established epilepsy

Naseem Al Salihi¹, Khalid Hamandi²

¹Department of Neurology, University Hospital of Wales, ²The Welsh Epilepsy Centre, Department of Neurology, University Hospital of Wales

Background: Ictal asystole (IA) is a recognised manifestation of epileptic seizures, most commonly seen in temporal or insula epilepsies. It may present as a new seizure semiology associated with sudden collapse and severe injuries, and may contribute to sudden unexpected death in epilepsy (SUDEP). Therefore, early recognition is essential.

Case presentation: We present four cases of ictal asystole captured during pre-surgical video EEG telemetry and routine follow-up.

Case 1: A 58-year-old woman with right temporal lobe epilepsy for 13 years presented with recurrent collapses occurring approximately every two months. Pre-surgical video EEG telemetry admission demonstrated ictal asystole lasting 8 seconds.

Case 2: A 38-year-old man with known right parietal dysembryoplastic neuroepithelial tumour (DNET) subtotal resection at the age of 7 and drug-resistant epilepsy presented with intermittent syncopal episodes. Pre-surgical video EEG detected bradycardia and syncope during one of the episodes. An implantable loop recorder confirmed ictal asystole lasting 19 seconds. He underwent resective epilepsy surgery with dramatic improvement in seizure frequency from more than 20 seizures per day to fewer than one seizure per week.

Case 3: A 73-year-old woman with right temporal lobe epilepsy and hippocampal sclerosis since age of 18 years developed a change in seizure semiology, notably recurrent collapses resulting in bone fractures. An Implantable loop recorder identified ictal bradyarrhythmia/asystole.

Case 4: A 41-year-old man with left temporal lobe epilepsy presented after an unexplained fall and loss of consciousness with no recollection of events. Pre-surgical video EEG telemetry captured ictal asystole lasting 20 seconds.

Results: All patients were treated with permanent pacemakers due to the length of asystole and risk of future injury or fatal asystole. Case 4 had a leadless pacemaker due to his relatively young age. All patients no longer had episodes of collapse and continued their usual epilepsy follow-up. 

Conclusion: A recent history of change in seizure type, with initial focal onset, followed by sudden collapse without bilateral tonic-clonic movements, was in keeping with the subsequently recorded asystole, and such a seizure pattern should prompt consideration of this phenomenon and investigation with telemetry or long-term loop recorder. Management of ictal asystole with permanent pacemakers is balanced against the risk-benefit of long-term pacemaker insertion and future lead complications.