Advancing SUDEP risk prediction using a multicenter case-control approach

使用多中心病例对照方法推进 SUDEP 风险预测

• 批准号: 10463739
• 负责人: Orrin Devinsky
• 金额: $ 62.82万
• 依托单位: NEW YORK UNIVERSITY SCHOOL OF MEDICINE
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-08-15 至 2026-07-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10463739
• 关键词: Address; Affect; Amygdaloid structure; Bayesian Modeling; Biological Markers; Bradycardia; Brain Stem; Cardiopulmonary; Case-Control Studies; Central Sleep Apnea; Characteristics; Clinic; Clinical; Counseling; Data; Data Collection; Data Set; Development; Disease; Electrocardiogram; Electroencephalography; Electrophysiology (science); Epilepsy; Exhibits; Factor Analysis; Frequencies; Functional disorder; Goals; Heart Arrest; Hippocampus (Brain); Image; Individual; Intervention; Knowledge; Machine Learning; Magnetic Resonance Imaging; Measures; Methods; Modeling; Monitor; Multicenter Studies; National Institute of Neurological Disorders and Stroke; Outcome; Patients; Performance; Persons; Physiological; Posture; Premature Mortality; Prevention strategy; Prospective Studies; Records; Research; Retrospective cohort; Risk; Risk Factors; Sample Size; Scalp structure; Seizures; Selection Bias; Severities; Sudden Death; Supervision; Techniques; Testing; Time; Tonic - clonic seizures; United States National Institutes of Health; awake; base; biomarker performance; brain magnetic resonance imaging; candidate marker; case control; clinical care; clinical practice; clinical risk; cost; data harmonization; deep learning; epidemiology study; follow-up; heart rate variability; high risk; innovation; machine learning method; magnetic resonance imaging biomarker; magnetic resonance imaging/electroencephalography; neuroimaging marker; novel; novel marker; personalized predictions; personalized risk prediction; potential biomarker; predictive modeling; prospective; random forest; research clinical testing; risk prediction; risk prediction model; screening; sex; statistical and machine learning; sudden unexpected death in epilepsy; tool

Project Summary Patients with epilepsy have a 25-fold elevated risk of sudden death. Sudden Unexpected Death in Epilepsy (SUDEP) is the most common disease-related cause of premature mortality in people with seizure disorders, affecting 1 of every 1000 patients annually. Despite this, mechanisms and risk factors of SUDEP remain largely unknown. Having ongoing generalized tonic-clonic seizures (GTCS) and no nocturnal supervision are the only definite risk factors, and reducing seizures is the only currently available preventive strategy. Several other clinical factors and potential biomarkers such as prolonged post-ictal generalized EEG suppression (PGES) that follows GTCS, abnormal inter-ictal ECG, and structural brain MRI abnormalities were associated with increased SUDEP risk, but none were rigorously confirmed in a large case-control study. While limitations of single-center studies in accumulating a sufficient number of cases are well recognized, prospective multicenter studies are also severely limited by the time, expense, and loss of follow-up constraining sample size and power. To sidestep these limitations, we propose a retrospective multisite case-control study that will screen >40,000 patients from 86 epilepsy monitoring centers worldwide, with a conservative expected total of >185 SUDEP cases and 370 age/sex-matched controls. Employing our comprehensive approaches to identify SUDEP cases combined with novel data harmonization techniques will allow us to: 1) provide an unprecedentedly large curated dataset of SUDEP, 2) identify clinical, electrophysiological, and imaging predictors of SUDEP using advanced machine learning methods, and 3) develop an individualized model to predict SUDEP risk that can be used in clinic. The proposed study will test the hypothesis that SUDEP cases exhibit different electroclinical and imaging characteristics that can provide an individualized prediction model. We will identify ictal electroclinical and interictal electrophysiological and neuroimaging biomarkers of SUDEP. We will compare markers of seizure severity between SUDEP cases and age/sex-matched living epilepsy patients, including decerebrate or decorticate posturing during GTCS, PGES duration, postictal bradycardia + asystole and post-convulsive central apnea. Additionally, we will assess putative interictal biomarkers including decreased low frequency power in ECG heart rate variability and decreased MRI-derived volumes in the right hippocampus/amygdala and brainstem. We will also employ machine-learning techniques to uncover novel biomarkers from interictal electrophysiological data. Finally, using a Bayesian framework, we will develop an individualized SUDEP risk prediction tool that combines clinical features with measures derived from routine EEG, ECG, and MRI. Our goal is to create a SUDEP case-control dataset to identify clinical risk factors and biomarkers that will help to create a robust model of an individual’s SUDEP risk based on measures derived from routine clinical care and testing such as interictal ECG, MRI and EEG. Such a tool could transform clinical practice, facilitate trials of SUDEP interventions, and ultimately save lives.

项目概要 癫痫患者猝死的风险增加 25 倍。癫痫猝死 (SUDEP) 是导致癫痫患者过早死亡的最常见疾病相关原因， 每年影响千分之一的患者。尽管如此，SUDEP 的机制和风险因素在很大程度上仍然存在 未知。持续全身强直阵挛发作 (GTCS) 且没有夜间监管是唯一的原因 明确的危险因素，减少癫痫发作是目前唯一可用的预防策略。其他几个 临床因素和潜在的生物标志物，例如长期发作后全身脑电图抑制（PGES）， 遵循 GTCS，异常的发作间期心电图和结构性脑 MRI 异常与增加 SUDEP 风险，但没有在大型病例对照研究中得到严格证实。虽然单中心的局限性 积累足够数量病例的研究已得到广泛认可，前瞻性多中心研究 时间、费用和随访损失也严重限制了样本量和功效。回避 鉴于这些局限性，我们建议进行一项回顾性多中心病例对照研究，该研究将筛选超过 40,000 名来自以下地区的患者： 全球有 86 个癫痫监测中心，保守预计 SUDEP 病例总数 >185 例，SUDEP 病例总数 > 370 例 年龄/性别匹配的对照。采用我们的综合方法来识别 SUDEP 案例，并结合 新颖的数据协调技术将使我们能够：1）提供前所未有的大型精选数据集 SUDEP，2) 使用先进机器识别 SUDEP 的临床、电生理和成像预测因子 学习方法，3) 开发个性化模型来预测可用于临床的 SUDEP 风险。 拟议的研究将检验 SUDEP 病例表现出不同的电临床和成像的假设 可以提供个性化预测模型的特征。我们将识别发作性电临床和 SUDEP 的发作间期电生理学和神经影像生物标志物。我们将比较癫痫发作的标志物 SUDEP 病例与年龄/性别匹配的活体癫痫患者之间的严重程度，包括去大脑或 GTCS、PGES 持续时间、发作后心动过缓 + 心搏停止和惊厥后中枢期间的皮质姿势 呼吸暂停。此外，我们将评估假定的发作间期生物标志物，包括低频功率降低 右海马/杏仁核的心电图心率变异性和 MRI 衍生体积减少 脑干。我们还将利用机器学习技术来发现发作间期的新生物标志物 电生理数据。最后，使用贝叶斯框架，我们将开发个性化的 SUDEP 风险 预测工具，将临床特征与常规脑电图、心电图和 MRI 得出的测量结果相结合。 我们的目标是创建一个 SUDEP 病例对照数据集来识别临床风险因素和生物标志物，这将有助于 根据常规临床护理得出的测量结果，创建个人 SUDEP 风险的稳健模型 测试，如发作间期心电图、MRI 和脑电图。这样的工具可以改变临床实践，促进试验 SUDEP 干预，最终拯救生命。