AED | Audio Epilepsy Digest

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 24: Before pregnancy: where does epilepsy care fall short? Puck and Olin begin AED's new scripted conversation format with three studies that look at reproductive care in epilepsy from different places: prescription claims, clinical records, and women's reports of the information they received. These studies were published in 2025–2026 and involve different people and endpoints. In this episode:- Harrison's US Medicaid study finds that nearly one in three people with an included susceptible contraceptive method had a sustained recorded prescription overlap with an enzyme-inducing antiseizure medication. The conversation explores which methods appeared most and least often, how long overlap windows lasted, and when each prescription came first.- Tao's single-center review examines what was documented before a first delivered pregnancy, including teratogenicity discussions, folic-acid recommendations, and recorded use. The pediatric-transition group was small and younger than the established adult-care group.- Baker's online survey asks women what reproductive information reached them, when it arrived, and whether it helped with decisions. Different survey questions have different response counts. The hosts put the findings together as a reason to check the specific medication and contraceptive method, ask whether pregnancy information arrived in time, and notice questions a clinical note may leave open. They discuss PROMPT-EPI, the editor's early AI-scribe template project, as a possible way to surface care gaps. Its simulated-visit pilot did not test a live clinic cue or patient outcomes. Reading notes:- Prescription overlap is not a measured contraceptive failure rate or proof of actual use.- A missing chart entry is not proof that a conversation did not happen.- The survey is self-selected; its responses are not a population prevalence estimate.- The three studies do not follow one group through the same care pathway. Sources:- Harrison et al. Neurology 2026. DOI: 10.1212/WNL.0000000000218679. PMID: 42842852.- Tao et al. Epilepsy & Behavior 2025. DOI: 10.1016/j.yebeh.2025.110292. PMID: 39983588.- Baker et al. Epilepsy & Behavior 2025. DOI: 10.1016/j.yebeh.2025.110737. AED Quiz (non-accredited self-assessment):https://audioepilepsydigest.com/episode-024-aed-quiz.html AED website:https://audioepilepsydigest.com/ AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-024-ai-review.html

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 23: Less invasive, but how much benefit? Three recent studies ask what a smaller intervention can promise: treatment-specific benefit, durable seizure control, and a meaningful safety advantage. In this episode:- A small sham-controlled taVNS crossover study and the difference between improvement and demonstrated superiority.- Long-term SEEG-guided radiofrequency thermocoagulation, changing treatment groups and incomplete follow-up.- Volume-based RF-TC for focal cortical dysplasia: a promising MRI safety signal with important limits. Reading notes:- These interventions and populations are distinct, not a head-to-head comparison.- Hong’s pilot did not demonstrate active-treatment superiority to sham; it establishes neither equivalence nor universal ineffectiveness.- Pesente’s one-year 19/135 seizure-free result applies to the selected RF-TC-only group, including repeat procedures. The abstract mixes responder denominators; the disputed responder and five-year percentages are omitted from the episode.- Hayashi’s primary endpoint was MRI-detected deep white matter infarction. Overall neurological deficit rates were not significantly different. Similar observed seizure outcomes do not establish noninferiority or equivalence; favorable ILAE 1–3 is broader than seizure freedom. Additional interpretation notes:- A null active-versus-sham result does not prove absence of physiological effect. Carryover remains possible but was not established in this small study.- The heat-sink explanation is a hypothesis. Lower MRI infarction incidence is not elimination of infarction or all neurological harm. Sources:- Hong et al. Epilepsia Open 2026. DOI: 10.1002/epi4.70354. PMID 42836555.- Spessatto Pesente et al. Epilepsia 2026. DOI: 10.1002/epi.70524. PMID 42836337.- Hayashi et al. Epilepsia 2026. DOI: 10.1002/epi.70454. PMID 42776563. AED Quiz (non-accredited self-assessment):https://audioepilepsydigest.com/episode-023-aed-quiz.html AED website:https://audioepilepsydigest.com/ AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-023-ai-review.html

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 22: The Epilepsy Follow-Up: Continue, Adjust, Switch—or Refer? An adult outpatient special prompted by a nurse practitioner's request: three everyday visits, from breakthrough seizures to treatment burden and repeated medication trials. The August 2025 Xcopri liver-label update provides a concrete example of checking medication-specific safety requirements. In this episode:- Reassessing events, actual medication use and barriers before automatic escalation.- Asking how a patient feels and functions, even when seizures have stopped.- Reviewing adequacy of previous treatments and explaining comprehensive epilepsy referral. Clinical context:- Liver-warning symptoms include unexplained nausea, vomiting, right upper quadrant pain, fatigue, anorexia, jaundice or dark urine. Seek prompt clinical assessment. Early DRESS may occur without an evident rash.- Xcopri labeling specifies baseline ALT, AST and total bilirubin unless available within three months. Follow-up testing is clinically indicated; no universal monthly schedule or liver-failure incidence is supplied.- SANAD II studied newly diagnosed focal epilepsy, not switching stable patients or comparing cenobamate.- Individualize emergency plans within local protocols and prescribing standards. NICE is UK guidance, not US regulation.- Jehi’s surgical-referral consensus recommends offering evaluation up to age 70 and considering older patients without surgical contraindications. Comprehensive care remains individualized. Referral evaluates options and does not commit a patient to surgery.- The label specifies ALT, AST and total bilirubin; other urgent tests are clinician-directed, not universal label requirements. Clinically indicated follow-up is broader than symptoms alone. Sources:- Xcopri US prescribing information, revised August 2025.- NICE NG217, chapters 4, 7 and 10.- Marson et al. SANAD II focal epilepsy. Lancet 2021. DOI: 10.1016/S0140-6736(21)00247-6.- Jehi et al. Timing of referral for epilepsy surgery evaluation. Epilepsia 2022. DOI: 10.1111/epi.17350. AED Quiz (non-accredited self-assessment):https://audioepilepsydigest.com/episode-022-aed-quiz.html AED website:https://audioepilepsydigest.com/ AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-022-ai-review.html

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 21: FCD: What Does the Diagnosis Actually Tell Us? A monthly special on focal cortical dysplasia: what an integrated diagnosis explains, what anatomy and electrophysiology add, and where AI detection and targeted treatment remain uncertain. Key takeaways:- MRI impression, histopathology and molecular findings are connected but distinct. FCD type is not a severity ladder.- Selected bottom-of-sulcus studies do not establish a universal resection rule.- Mechanistic promise and exploratory subgroup signals do not establish a prescribing rule. Editorial clarifications:- MTOR variants were not detected in the available blood or saliva samples in Lim’s study; negative peripheral testing does not exclude brain-restricted mosaicism.- Regional surgical-outcome differences are observational and may reflect surgical constraints; they are not individual predictions.- No washout in the everolimus crossover trial allows potential carryover and limits interpretation. The trial did not demonstrate overall superiority.- Independent AI testing evaluates performance under different conditions; detection is not a resection map. Papers discussed:- Najm et al. (2022). DOI: 10.1111/epi.17301- Wagstyl et al. (2022). DOI: 10.1111/epi.17130; correction retained.- Macdonald-Laurs et al. (2023). DOI: 10.1111/epi.17495- Ripart et al. Corrected MELDGraph study; correction retained.- Kaas et al. (2026). DOI: 10.1016/j.eplepsyres.2026.107778- Lim et al. (2015). DOI: 10.1038/nm.3824- Kim et al. (2025). DOI: 10.1002/epi4.13104 AED website:https://audioepilepsydigest.com/ AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-021-ai-review.html Non-accredited AED Quiz:https://audioepilepsydigest.com/episode-021-aed-quiz.html

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 20: Who Decides on Epilepsy Surgery? From identifying people who have not reached evaluation to center-level recommendations and estimates of seizure benefit and visual harm, five papers examine how the path to epilepsy surgery is shaped—and what these studies cannot decide for an individual. The opening welcomes Timon, the editor’s new Sheltie puppy, before the research discussion. AED Quiz for this episode:https://audioepilepsydigest.com/episode-020-aed-quiz.html Listen and follow Audio Epilepsy Digest:https://audioepilepsydigest.com/ Key takeaways:- Caraway: recommendations varied across 29 pediatric centers after adjustment. A recommendation included invasive monitoring; it did not establish an operation, outcome, or inappropriate decision.- Simmons: an EHR rule flagged possible cases for chart review. Among 84 confirmed cases of drug-resistant epilepsy, 48 had no evaluation initiated; the rule did not itself diagnose drug resistance.- Coordinator tracking was associated with evaluation status. This observational finding establishes neither benefit nor lack of benefit from a coordinator intervention.- Karpychev: MRI network measures showed an exploratory long-term outcome signal, with external AUC 0.65. The model is not a surgical exclusion rule or a resection map.- Wreghitt: probabilistic tractography better delineated optic-radiation anatomy and estimated injury correlated with visual-field loss in 33 adults. It did not establish guaranteed safe margins or prevention of visual deficits.- Englot supplies historical review-based referral context. Referral, recommendation, prediction, treatment, and patient preferences remain distinct decisions. Editorial clarifications:- Current ASM count or EHR medication exposure is not the same as adequately failed trials.- In the MRI model, seizure freedom is the positive class; specificity 80% and NPV 63.9% concern non-seizure-free classification.- The reflective closing question is commentary. The MRI paper is a disclosed one-day exception to the August 13–September 12 literature window. Papers discussed:- Caraway AR et al. Epilepsia (2026). PMID: 42621761.- Englot DJ. Epilepsia (2026). PMID: 42611555.- Simmons GB et al. Epilepsy & Behavior (2026). PMID: 42697047.- Karpychev V et al. Neurology (2026). PMID: 42585607.- Wreghitt S et al. Journal of Neurosurgery (2026). PMID: 42696783. AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-020-ai-review.html

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 19: The Implantation Honeymoon—Who Gets the Credit: Implant, Stimulation, or Time? When seizures change after intracranial electrodes are implanted but before stimulation begins, what can we actually attribute to the implant—and what remains time, sampling, treatment, or expectation? The newest adult SEEG study leads this episode with results first. From there, we move from clinical seizure reports to scheduled RNS ECoG, rare no-stimulation observations, a disclosed single-center abstract from our own group, and an implanted-sham randomized trial. The evidence supports a real postimplant observation problem, not one proven mechanism or a universal treatment effect. AED Quiz for this episode:https://audioepilepsydigest.com/episode-019-aed-quiz.html Listen and follow Audio Epilepsy Digest:https://audioepilepsydigest.com/ Key takeaways:- Roy-Chowdhury's adult SEEG cohort is a set of distinct last-follow-up groups, not one 51-patient longitudinal curve.- The paper reports early responder and seizure-freedom counts, but its six-month denominator differs between the narrative and Table 1.- Lane found a modest delay to first seizure during intracranial monitoring after adjustment for prior-day medication load; daily seizure frequency was not significantly reduced.- Sun documented months of scheduled ECoG signal evolution after RNS implantation, an electrographic endpoint that does not establish clinical seizure benefit or one cause.- Exceptional RNS and DBS observations show that changes can occur without active stimulation, but small samples and concurrent care prevent causal attribution.- The editor's AES abstract suggests a predictive association, not proof that implantation caused later response or a validated biomarker.- NAUTILUS compared active with sham stimulation on a shared implanted background. Its prespecified effectiveness endpoint was negative; later findings are supportive and cannot replace that endpoint. Papers discussed:- Roy-Chowdhury et al. Post-SEEG seizure-frequency study. Epilepsia. 2026. PMID: 42474247.- Lane et al. Implantation effect during intracranial monitoring. Acta Neurol Scand. 2017. PMID: 27531652.- Sun et al. Postimplant ECoG evolution in RNS. Clin Neurophysiol. 2018. PMID: 29233473.- Thuberg et al. DBS insertional effect without stimulation. Neuromodulation. 2021. PMID: 33577139.- Rumschlag et al. Extended RNS implantation-effect case. Epileptic Disord. 2024. PMID: 38787629.- McCarthy and Burdette. RNS programming review. Neurotherapeutics. 2026. PMID: 42302539.- Uysal et al. NAUTILUS randomized trial. Epilepsia. 2026. PMID: 42233958. Editorial reference and disclosure:- Rehim et al. “Does Honeymoon Lead to Happily Ever After? Single Center Experience with Implantation Effect in RNS Therapy.” AES Annual Meeting abstract 1.251. 2025. Erafat Rehim is the presenting author and AED editor. This author-involved, retrospective single-center abstract is treated as an unvalidated hypothesis, not independent confirmation. AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-019-ai-review.html Editorial review disclosure:The final draft cleared two independent source-fidelity and clinical-safety reviews with minor listening notes. Earlier drafts were withheld for causal overstatement, endpoint confusion, unsupported mechanisms, or narrative imbalance. The clinician-editor completed listening QA and authorized release for Tuesday, September 1, 2026 at 7:00 AM EDT.

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 18: How Safe Is Safe? The Long-Term Tradeoffs of Common Anti-Seizure Medications When a patient asks, “You say Keppra is safe long term. Exactly how safe is it?”, the honest answer is not one lifetime percentage. This episode separates years of clinical experience from years of structured observation, then compares levetiracetam, lamotrigine, and lacosamide across common burden, rare serious harm, cumulative toxicity, changing susceptibility, and seizure-control benefit. Long follow-up can be reassuring, but selected continuers are not the same denominator as everyone who starts treatment. Episode infographic AED Quiz for this episode:https://audioepilepsydigest.com/episode-018-aed-quiz.html Listen and follow Audio Epilepsy Digest:https://audioepilepsydigest.com/ Key takeaways:- “Safe long term” means a changing benefit-risk profile, not zero risk or a lifetime guarantee.- Approval history gives decades of signal-detection experience; it does not provide a controlled lifetime denominator.- Maximum structured observation reached beyond seven years for levetiracetam, six years for lamotrigine, and up to eight years for lacosamide in selected continuation cohorts.- Levetiracetam's ongoing reassessment targets include mood or behavioral effects, fatigue, and renal-dose dependence.- Lamotrigine's defining serious-rash risk is front-loaded during initiation and titration, while neurologic effects may persist or recur with dose or interaction changes.- Lacosamide's dizziness and imbalance profile must be considered alongside conduction susceptibility, cardiac disease, and interacting drugs.- Comparative tolerability depends on population, comparator, endpoint, and seizure-control benefit; this is not a universal drug ranking. Papers discussed:- Marson et al. SANAD II focal-epilepsy trial. Lancet. 2021. PMID: 33838757.- Fröling et al. Levetiracetam/brivaracetam adverse-event meta-analysis. 2026. PMID: 41824075.- Depondt et al. Long-term levetiracetam retention. JNNP. 2006. PMID: 16361605.- Bauer et al. Long-term levetiracetam follow-up. 2006. PMID: 16911344.- Faught et al. Six-year lamotrigine continuation. 2004. PMID: 14751204.- Li et al. Three-year real-world LTG-LEV comparison. 2020. PMID: 32679487.- Ben-Menachem et al. Lacosamide versus carbamazepine-CR extension. Epilepsia. 2019. PMID: 31755090.- Rosenfeld et al. Lacosamide extension up to eight years. 2014. PMID: 25461210.- Buoso et al. Antiseizure medications and bone health. 2024. PMID: 40530090. AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-018-ai-review.html Editorial disclosure:The final fifth draft cleared two independent source-fidelity and clinical-safety reviews with minor notes. Earlier expanded drafts were withheld for causal overstatement, an incorrect study attribution, loss of the direct Keppra question, and unsupported conversion of heterogeneous bone-health evidence into individual risk and treatment advice. The clinician-editor completed listening QA and authorized release for Monday at 7:00 AM EDT.

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 17: What Counts as Neuromodulation Success? Should the 50% responder rate remain the primary definition of neuromodulation success? This episode stages the strongest case for a simple common benchmark against the strongest case for a multidimensional clinical verdict. VNS, pediatric thalamic DBS and RNS, long-term RNS, ESTEL, CONNECTiVOS, PuLsE, and two biomarker studies are used to test what the threshold captures—and what it hides. AED Quiz for this episode:https://audioepilepsydigest.com/episode-017-aed-quiz.html Listen and follow Audio Epilepsy Digest:https://audioepilepsydigest.com/ Key takeaways:- Keep the 50% responder rate as a standardized anchor, but never let it stand alone.- Pair any threshold with the underlying reduction distribution, seizure freedom, denominator, and time horizon.- Similar 12-month responder proportions do not prove similar early experience or device superiority.- Quality of life, harms, and treatment burden are core counseling domains rather than decorative secondary outcomes.- ESTEL preserved a negative diary primary comparison alongside a positive secondary ambulatory-EEG responder signal; objective detection and patient importance are not synonyms.- A high AUC in a tiny held-out split or a post-treatment blood-marker association is hypothesis-generating, not a patient-selection tool. Papers discussed:- Zhao et al. Vagus nerve stimulation as an add-on therapy in patients with epilepsy. Acta Epileptologica. 2026. doi:10.1186/s42494-026-00272-4.- Tenhoeve et al. Comparative multicenter evaluation of thalamic neuromodulation for treatment-resistant epilepsy in children. Epilepsia. 2026. doi:10.1002/epi.70402.- Zhou et al. Nonlinear neurodynamics of N2 sleep EEG predict outcomes of ANT-DBS in epilepsy. Journal of Neural Engineering. 2026. doi:10.1088/1741-2552/ae94b2.- Tang et al. Peripheral blood inflammatory biomarkers and response to VNS in pediatric drug-resistant epilepsy. European Journal of Pediatrics. 2026. doi:10.1007/s00431-026-07256-z.- Niazi et al. What defines response to VNS in children with drug-resistant epilepsy? Epilepsia. 2025. doi:10.1111/epi.18520.- Dalic et al. DBS of Thalamic Centromedian Nucleus for Lennox-Gastaut Syndrome (ESTEL Trial). Annals of Neurology. 2022. doi:10.1002/ana.26280.- Nair et al. Nine-year prospective efficacy and safety of brain-responsive neurostimulation for focal epilepsy. Neurology. 2020. doi:10.1212/WNL.0000000000010154.- Ryvlin et al. The long-term effect of VNS on quality of life: The PuLsE trial. Epilepsia. 2014. doi:10.1111/epi.12611. AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-017-ai-review.html Editorial disclosure:Two independent AI-assisted reviews cleared the final candidate with minor notes and no blocking source-fidelity or clinical-safety finding. The first draft was withheld after it confused an electrographic responder proportion with seizure-reduction magnitude and described interval seizure freedom as cure. The corrected final draft preserves endpoint hierarchy, denominators, design limits, and the research-stage status of proposed biomarkers. The clinician-editor authorized shipment for Monday at 7:00 AM EDT.

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 16: Signals, Silence, and Zero Events What can an observational association, the absence of a consistent reported pattern, and zero observed events actually justify in medication counseling? This episode compares three very different medication-safety evidence systems: a nationwide prenatal antiseizure-medication cohort, a global eslicarbazepine pregnancy safety database, and a small selected cohort restarting cenobamate after short interruptions. The practical method is simple: identify the denominator, comparator, missingness, exposure measurement, outcome ascertainment, and selection process before translating any memorable conclusion into care. AED Quiz for this episode:https://audioepilepsydigest.com/episode-016-aed-quiz.html Listen and follow Audio Epilepsy Digest:https://audioepilepsydigest.com/ Key takeaways:- The Korean cohort found adjusted associations with selected neurodevelopmental outcomes; it did not establish a causal medication effect.- Only 48.2% of exposed mothers in that cohort had epilepsy, so indication, disease severity, familial factors, and prescription-based exposure remain central limitations.- The eslicarbazepine database collected reports across miscarriage, congenital anomaly, neonatal adverse effect, and premature birth, but substantial missingness and the lack of a valid denominator or comparator prevent incidence and comparative-risk claims.- Zero attributed hypersensitivity events among 32 previously tolerant cenobamate patients is reassuring within that selected setting, but an approximate upper uncertainty bound remains near 9%.- Residual drug exposure is a proposed pharmacokinetic explanation, not proof of immune tolerance or authorization for a universal maintenance-dose restart protocol. Papers discussed:- Kim et al. Prenatal exposure to Antiseizure Medications for all indications and the risk of neurodevelopmental disorders in offspring. Seizure. 2026. doi:10.1016/j.seizure.2026.05.030.- Lopes et al. Pregnancy outcomes following Eslicarbazepine acetate exposure in women with epilepsy. Seizure. 2026. doi:10.1016/j.seizure.2026.05.027.- Feldman et al. Hypersensitivity reaction risk when restarting cenobamate after discontinuation. Epilepsia Open. 2026. doi:10.1002/epi4.70288. AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-016-ai-review.html Editorial disclosure:Two independent AI-assisted reviews cleared the final candidate with minor notes and no blocking source-fidelity or clinical-safety finding. The remaining notes concerned prior-rash-history wording, complete reporting of pregnancy outcome categories, pronunciation of selected technical terms, separation of two rash observations, and maintaining the restart mechanism as a hypothesis. The clinician-editor listened to and approved the release candidate.

NOTE: This podcast is an evolving collaboration between human and AI. While we strive for accuracy, AI hosts may misinterpret or oversimplify source material. Always refer to the original published articles for clinical decision-making. If you find any claims made by the AI hosts to be inaccurate, please let us know. Your feedback directly improves future episodes. Episode 15: Fellow Bootcamp: OSA Is Not Just a Comorbidity When should obstructive sleep apnea be treated as more than a passive comorbidity in epilepsy care? This special uses the nine Bradford Hill viewpoints to examine whether OSA may modify epilepsy outcomes. The hosts keep three evidence lanes separate: seizure and interictal-discharge burden, incident or late-onset epilepsy, and surrogate SUDEP-risk markers. They also examine what the current CPAP studies can—and cannot—support. AED Quiz for this episode:https://audioepilepsydigest.com/episode-015-aed-quiz.html Listen and follow Audio Epilepsy Digest:https://audioepilepsydigest.com/ Key takeaways:- Bradford Hill viewpoints organize causal reasoning; they are not a checklist, vote, or proof score.- OSA is commonly reported in people with epilepsy, but prevalence estimates are heterogeneous and do not establish causation.- Prospective and claims-based studies support an incident-epilepsy and late-onset-epilepsy hypothesis, with residual confounding and exposure measurement still important.- Associations with revised SUDEP risk scores concern surrogate markers, not observed SUDEP events or evidence of prevention.- Treat confirmed OSA for established sleep-medicine indications and track epilepsy outcomes without promising an epilepsy-specific response. Sources include Rehim, Vendrame, and Devinsky 2026; Lin et al. 2017; Harnod et al. 2017; Carosella et al. 2023; Soontornpun et al. 2024; Pornsriniyom et al. 2014; Malow et al. 2008; and McCarter et al. 2018. AI-assisted editorial/source review:https://audioepilepsydigest.com/episode-015-ai-review.html Editorial disclosure:Two independent AI-assisted reviews cleared the episode with minor notes and no blocking source-fidelity or clinical-safety finding. Minor notes concern a few contextual certainty words, process-like language near the opening, an over-categorical effect-size heuristic, the ARIC sleep-substudy denominator, and author-name pronunciation. The clinician-editor listened to and approved the release candidate.