Tracking seizure cycles beats a prospective moving average

Tracking seizure cycles beats a prospective moving average | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Tracking seizure cycles beats a prospective moving average Rachel E. Stirling , View ORCID Profile Benjamin H. Brinkmann , View ORCID Profile Dean R. Freestone , View ORCID Profile Philippa J. Karoly doi: https://doi.org/10.1101/2025.11.03.25338700 Rachel E. Stirling 1 Graeme Clark Institute and Biomedical Engineering, The University of Melbourne , Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Benjamin H. Brinkmann 2 Department of Neurology , Mayo Clinic, Rochester MN, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Benjamin H. Brinkmann Dean R. Freestone 1 Graeme Clark Institute and Biomedical Engineering, The University of Melbourne , Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Dean R. Freestone Philippa J. Karoly 1 Graeme Clark Institute and Biomedical Engineering, The University of Melbourne , Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Philippa J. Karoly For correspondence: karolyp{at}unimelb.edu.au Abstract Full Text Info/History Metrics Data/Code Preview PDF Abstract This commentary addresses the debate regarding the predictive value of multiday seizure cycles versus simple statistical baselines. Multidien seizure cyclicity is a prevalent, patient-specific phenomenon with promise for epilepsy management. We challenge the assertion that cycle tracking is no better than a 90-day moving average, which is an inherently retrospective model that lags changes in seizure likelihood. We compared a causal cyclic forecast to a prospectively applied moving average across a large seizure diary cohort (n=768) and two gold-standard chronic EEG cohorts (n=24). At the group level for the EEG and diary cohorts, cycle tracking demonstrated significantly superior accuracy to the moving average for both hourly and daily forecasts (p < 0.0001). These results confirm that event-based cyclical models offer more accurate, simulated real-world forecasts. We conclude that robust forecasting tools must prioritize the detection and modeling of seizure cycles to move beyond simple baseline performance and provide actionable clinical utility. Multidien cycles of seizures have emerged as a prevalent phenomenon in epilepsy, alongside well documented circadian patterns. These longer rhythms have been observed at individual-specific weekly, monthly, seasonal, or other multiday periods and occur across sex, medications, age, and epilepsy type ( Karoly et al., 2021 ). Detecting and tracking individuals’ seizure cycles holds promise for epilepsy management, and significant advances have been made in the reliable extraction of seizure cycles from different data sources. Projecting cycles forward in time to forecast future periods of seizure risk has been demonstrated successfully, although several challenges remain. Chang and colleagues ( Chang et al., 2025 ) contend that the forward projection of multiday seizure cycles is no better than using the historic (90-day) long-term average of seizure rates to estimate the chance of a seizure the next day. On an individual level, this will be true for some people and not others. It is likely that the 90-day historic average provides an approximation of longer (i.e. monthly) rhythms and clustered seizures, although it is likely less precise for people with shorter (i.e. weekly) rhythms and certainly contains no information about circadian trends. While benchmarking seizure forecasts against a historic moving average is a reasonable baseline, suggesting that tracking seizure cycles performs no better than tracking a moving average rate in the general population is not currently supported by data. We were unable to replicate the results reported for the 90-day moving average method. We compared a prospectively applied 90-day moving average forecast to a causal cyclic forecast (Nasseri et al., 2025) in a large seizure diary cohort (n=768) ( Stirling et al., 2024 ) and two small, gold-standard benchmarked chronic EEG cohorts: the My Seizure Gauge (MSG) cohort, consisting of 11 subjects who had a chronic EEG monitoring device (UNEEG SubQ, NeuroPace RNS, or Medtronic Summit RC□+□S) for 6-24 months (Nasseri et al., 2025); and the NeuroVista (NV) cohort, consisting of 13 subjects who were implanted with the NeuroVista device for 6-24 months ( Cook et al., 2013 ). In the chronic EEG cohort, cycle tracking was more accurate than a moving average in all but one of the MSG and all but one of the NV cohorts, for both hourly and daily pseudo-prospective forecasts (with forecast horizons extending to the next seizure event). Cycles tracking performed significantly better than a moving average at the group level comparing both hourly and daily AUC (p < 0.0001, paired t-test, combined MSG and NV cohorts). In the seizure diary cohort, cycle tracking was more accurate than 86% and 63% of moving average forecasts at the hourly and daily timescales, respectively. These proportions are similar to the number of people with significant circadian and multiday cycles (∼90% and ∼60%) observed in longitudinal seizure diary cohorts ( Karoly et al., 2018 ; Leguia et al., 2021 ). Hourly and daily group level performance confirmed cycles tracking was significantly better than moving average (p < 0.0001, paired t-test). Results are shown in Table 1 and both chronic EEG datasets are publicly available (MSG Cohort: https://www.epilepsyecosystem.org/my-seizure-gauge-1 and NV Cohort: https://doi.org/10.26188/5b6a999fa2316 ). These comparisons suggest that event-based seizure cycles provide significantly more accurate forecasts than a moving average in a simulated real-world setting, highlighting the importance of prospective implementation during testing. Moving averages, though simple, are inherently retrospective, as they estimate risk from recent seizure counts within a fixed window. This approach, which may effectively track long cycles retrospectively, lags impending changes in seizure probability, depends heavily on window length, and is unable to project an underlying cycle forward in time. As a result, it performs poorly at forecasting lead seizures following quiescent periods and is less reliable than causal event-based cyclical models. View this table: View inline View popup Table 1. Subject demographic characteristics, clinical characteristics and pseudo-prospective forecasting results. SG = Seizure Gauge cohort, NV = NeuroVista cohort (participant IDs deidentified). FIAS = focal impaired awareness seizures, FAS = focal awareness seizures, TCS = tonic clonic seizures, FBTCS = focal to bilateral tonic clonic seizures Averages and standard deviations are reported for age, seizure frequency, and all pseudo-prospective results. Algorithms with superior performance (based on AUC) are bolded, per subject, at both hourly and daily timescales. The results shown in Table 1 show that event-based cyclical models tend to perform better compared to moving average when using EEG-confirmed seizures compared to self-reports. It is certainly true that any forecast derived from an inaccurate record of sself-reported events would not capture the underlying neurophysiological cycles and would perform poorly in a prospective or pseudoprospective setting. Given that half or more of epilepsy patients have been shown to keep inaccurate diaries, the low accuracy demonstrated by Chang et al and others ( Elger & Hoppe, 2018 ; Goldenholz et al., 2024 ; Proix et al., 2021 ; Schulze-Bonhage et al., 2023 ; Wong et al., 2025 ) is not surprising. However, this performance reflects a fundamental limitation of the data rather than any forecasting method. It is well established that self-reported seizure burden is not closely correlated with individuals’ electrographic seizure rates ( Cook et al., 2013 ; Halliday et al., 2025 ; Wong et al., 2025 ). Nevertheless, for statistical reasons, a repeating pattern (i.e.a cycle) can be more reliably estimated from background noise; meaning that even unreliable self-reported seizures show similar periodic patterns as epileptic activity captured from EEG ( Karoly et al., 2020 ; Leguia et al., 2021 ; Reynolds et al., 2025 ). This suggests that seizure cycles still provide a better retrospective biomarker of underlying changes in epileptic activity than self-reported seizure rates. As more continuous data and reliable seizure detectors become available (either from wearable or implanted devices), the application of cycle-tracking becomes substantially more powerful, and it is unlikely that event-based moving average models will be able to match the performance of continuous cyclic forecasts. In any case, to suggest that a model is clinically useless because it can be computed ‘on the back of a napkin’ equates complexity with utility. We believe this is an error for medical applications, where ease of use, interpretability, and application should be prioritized. A forecasting method with many false alarms may cause alarm fatigue if sent to caregivers but may be helpful if integrated into a neuromodulation device. Further, most patients do not have an intuitive, everyday awareness of the current long-term trend in their seizure rates (the moving average model). Seizures are often rare, and humans are not naturally good at detecting historic health trends over months or years ( Schmier & Halpern, 2004 ). Providing insight into past seizure patterns, whether a long-term trend or multiday cycle, is desirable to people with epilepsy ( Grzeskowiak & Dumanis, 2021 ; Janse et al., 2019 ; Stirling et al., 2024 ; Goldenholz et al., 2024 ; Schmier & Halpern, 2004 ) and can drive improved quality of life ( Stirling et al., 2024 ). It remains to be tested whether some people derive more benefit from a cycle vs a moving average model, and this likely depends on their individual statistical patterns and seizure frequency. Clearly, both models must be assessed, and the time has now arrived for these clinical trials. Data Availability The data that support the findings of this study are available in Figshare (NV cohort seizure times) and Github (MSG cohort seizure times). Data from the seizure diary cohort can be provided upon reasonable request to the corresponding author. Code to compute the prospective moving average benchmark is available from Github at https://github.com/RiPLresearch/forecast-benchmarks/tree/main and code to compute prospective seizure cycles can be shared upon reasonable request to the corresponding author. https://www.epilepsyecosystem.org/my-seizure-gauge-1 https://doi.org/10.26188/5b6a999fa2316 References ↵ Chang , C.-Y. , Moss , R. , Westover , M. B. , & Goldenholz , D. M. ( 2025 ). Rigorous evaluation of five models for e-diary-only seizure forecasting: Retrospective and prospective datasets do not outperform the Napkin method . Epilepsia , n/a(n/a). doi: 10.1111/epi.18677 OpenUrl CrossRef ↵ Cook , M. J. , O’Brien , T. J. , Berkovic , S. F. , Murphy , M. , Morokoff , A. , Fabinyi , G. , D’Souza , W. , Yerra , R. , Archer , J. , & Litewka , L. ( 2013 ). Prediction of seizure likelihood with a long-term, implanted seizure advisory system in patients with drug-resistant epilepsy: A first-in-man study . The Lancet Neurology , 12 ( 6 ), 563 – 571 . OpenUrl PubMed ↵ Elger , C. E. , & Hoppe , C. ( 2018 ). Diagnostic challenges in epilepsy: Seizure underreporting and seizure detection . The Lancet Neurology , 17 ( 3 ), 279 – 288 . doi: 10.1016/S1474-4422(18)30038-3 OpenUrl CrossRef PubMed ↵ Goldenholz , D. M. , Eccleston , C. , Moss , R. , & Westover , M. B. ( 2024 ). Prospective validation of a seizure diary forecasting falls short . Epilepsia , 65 ( 6 ), 1730 – 1736 . doi: 10.1111/epi.17984 OpenUrl CrossRef PubMed ↵ Grzeskowiak , C. L. , & Dumanis , S. B. ( 2021 ). Seizure forecasting: Patient and caregiver perspectives . Frontiers in Neurology , 12 , 717428 . OpenUrl PubMed ↵ Halliday , A. J. , Gillinder , L. , Lai , A. , Seneviratne , U. , Fontenot , H. , Cameron , T. , McLean , K. , Niemiec , A. , Raghupathi , R. , Ganguly , T. M. , Ellis , C. , Conrad , E. C. , Briggs , R. , Bulluss , K. , Kwan , P. , Perucca , P. , O’Brien , T. J. , McGonigal , A. , Gutman , M. , … Cook , M. J. ( 2025 ). The UMPIRE study: A first-in-human multicenter trial of bilateral subscalp monitoring for epileptic seizure detection . Epilepsia , 66 ( 9 ), 3426 – 3439 . doi: 10.1111/epi.18458 OpenUrl CrossRef PubMed ↵ Janse , S. A. , Dumanis , S. B. , Huwig , T. , Hyman , S. , Fureman , B. E. , & Bridges , J. F. P. ( 2019 ). Patient and caregiver preferences for the potential benefits and risks of a seizure forecasting device: A best–worst scaling . Epilepsy & Behavior , 96 , 183 – 191 . doi: 10.1016/j.yebeh.2019.04.018 OpenUrl CrossRef PubMed ↵ Karoly , P. J. , Cook , M. J. , Maturana , M. , Nurse , E. S. , Payne , D. , Brinkmann , B. H. , Grayden , D. B. , Dumanis , S. B. , Richardson , M. P. , Worrell , G. A. , Schulze-Bonhage , A. , Kuhlmann , L. , & Freestone , D. R. ( 2020 ). Forecasting cycles of seizure likelihood . Epilepsia . doi: 10.1111/epi.16485 OpenUrl CrossRef ↵ Karoly , P. J. , Goldenholz , D. M. , Freestone , D. R. , Moss , R. E. , Grayden , D. B. , Theodore , W. H. , & Cook , M. J. ( 2018 ). Circadian and circaseptan rhythms in human epilepsy: A retrospective cohort study . The Lancet Neurology , 17 ( 11 ), 977 – 985 . doi: 10.1016/S1474-4422(18)30274-6 OpenUrl CrossRef PubMed ↵ Karoly , P. J. , Rao , V. R. , Gregg , N. M. , Worrell , G. A. , Bernard , C. , Cook , M. J. , & Baud , M. O. ( 2021 ). Cycles in Epilepsy . Nature Reviews Neurology . ↵ Leguia , M. G. , Andrzejak , R. G. , Rummel , C. , Fan , J. M. , Mirro , E. A. , Tcheng , T. K. , Rao , V. R. , & Baud , M. O. ( 2021 ). Seizure cycles in focal epilepsy . JAMA Neurology . ↵ Proix , T. , Truccolo , W. , Leguia , M. G. , Tcheng , T. K. , King-Stephens , D. , Rao , V. R. , & Baud , M. O. ( 2021 ). Forecasting seizure risk in adults with focal epilepsy: A development and validation study . The Lancet Neurology , 20 ( 2 ), 127 – 135 . OpenUrl PubMed ↵ Reynolds , A. , Stirling , R. E. , Håkansson , S. , Karoly , P. , Lai , A. , Grayden , D. B. , Cook , M. J. , Nurse , E. S. , & Peterson , A. ( 2025 ). Evaluating the accuracy of monitoring seizure cycles with seizure diaries . Epilepsia , 66 ( 5 ), 1585 – 1598 . doi: 10.1111/epi.18309 OpenUrl CrossRef ↵ Schmier , J. K. , & Halpern , M. T. ( 2004 ). Patient recall and recall bias of health state and health status . Expert Review of Pharmacoeconomics & Outcomes Research , 4 ( 2 ), 159 – 163 . doi: 10.1586/14737167.4.2.159 OpenUrl CrossRef ↵ Schulze-Bonhage , A. , Richardson , M. P. , Brandt , A. , Zabler , N. , Dümpelmann , M. , & San Antonio-Arce , V. ( 2023 ). Cyclical underreporting of seizures in patient-based seizure documentation . Annals of Clinical and Translational Neurology , 10 ( 10 ), 1863 – 1872 . doi: 10.1002/acn3.51880 OpenUrl CrossRef ↵ Stirling , R. E. , Nurse , E. S. , Payne , D. , Naim-Feil , J. , Coleman , H. , Freestone , D. R. , Richarson , M. P. , Brinkmann , B. H. , D’Souza , W. J. , & Grayden , D. B. ( 2024 ). User experience of a seizure risk forecasting app: A mixed methods investigation . Epilepsy & Behavior , 157 , 109876 . OpenUrl PubMed ↵ Wong , V. , Hannon , T. , Fernandes , K. M. , Cook , M. J. , & Nurse , E. S. ( 2025 ). Unseen yet overcounted: The paradox of seizure frequency reporting . Epilepsy & Behavior , 165 , 110335 . doi: 10.1016/j.yebeh.2025.110335 OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted November 06, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Tracking seizure cycles beats a prospective moving average Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Tracking seizure cycles beats a prospective moving average Rachel E. Stirling , Benjamin H. Brinkmann , Dean R. Freestone , Philippa J. Karoly medRxiv 2025.11.03.25338700; doi: https://doi.org/10.1101/2025.11.03.25338700 Share This Article: Copy Citation Tools Tracking seizure cycles beats a prospective moving average Rachel E. Stirling , Benjamin H. Brinkmann , Dean R. Freestone , Philippa J. Karoly medRxiv 2025.11.03.25338700; doi: https://doi.org/10.1101/2025.11.03.25338700 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Neurology Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (299) Cardiovascular Medicine (4425) Dentistry and Oral Medicine (443) Dermatology (382) Emergency Medicine (607) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1507) Epidemiology (15221) Forensic Medicine (30) Gastroenterology (1123) Genetic and Genomic Medicine (6588) Geriatric Medicine (667) Health Economics (997) Health Informatics (4524) Health Policy (1368) Health Systems and Quality Improvement (1612) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15910) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (145) Nephrology (667) Neurology (6588) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1143) Occupational and Environmental Health (956) Oncology (3331) Ophthalmology (970) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (663) Pediatrics (1690) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5440) Public and Global Health (9220) Radiology and Imaging (2195) Rehabilitation Medicine and Physical Therapy (1369) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (710) Sports Medicine (529) Surgery (710) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ffceb653d9d1640',t:'MTc3OTQ2NDQyOA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();