Sudden Unexpected Death in Epilepsy A Personalized Prediction Tool
7 May 2021
This international study co-led by UCL and published online on April 28th 2021 in Neurology, the medical journal of the American Academy of Neurology, presents a model based on routine clinical data that can predict an individual's risk of Sudden Unexpected Death in Epilepsy.
Although most individuals with epilepsy recover completely from a seizure, some tragically die for unclear reasons. There are some factors that are known to increase the average risk of Sudden Unexpected Death in Epilepsy (SUDEP), such as an increased frequency of convulsions, but predicting an individual’s risk of SUDEP remains challenging.
The study investigators, including Dr Ashwani Jha, Professor Ley Sander, and Dr Beate Diehl from UCL Queen Square Institute of Neurology, combined observational data from 1273 individuals with epilepsy across 4 studies conducted in the US, England and Wales, Sweden, and Scotland to develop a predictive model that forecasts an individual's risk of SUDEP. The model used twenty-two routinely acquired clinical variables and cutting-edge probabalistic algorithms to make the predictions..
The primary outcome was the performance of the model at differentiating SUDEP cases from non-SUDEP cases on unseen data samples as compared to two more commonly used estimates based on the population average risk or convulsion frequency only. The researchers found their model had greater predictive accuracy.
Important predictors of SUDEP in the novel model included Generalised Tonic Clonic seizure and focal-onset seizure frequency, excessive alcohol consumption, younger age at epilepsy onset, and a family history of epilepsy. Adherence to antiseizure medication predicted lower SUDEP risk.
As well as providing an individualised risk the model also communicates how certain it is, which is critical in clinical scenarios where the aim is often to provide reassurance or motivation to change.
The study’s limitations include an inability to draw causal links between the aforementioned risk factors and SUDEP, and the need to test the model in an external validation dataset.
"This study provides the first tool to assess a patient’s risk of dying from SUDEP, and can allow them to see, for example, how improved medication adherence can translate into a longer life". Prof Orrin Devinsky (Comprehensive Epilepsy Center, New York University Langone Medical Center, New York, NY, USA)
"SUDEP is the most tragic outcome of epilepsy and this tool will allow the individualized prediction of its risk and therefore decrease the number of these untimely deaths". Prof Ley Sander (NIHR University College London Hospitals Biomedical Research Centre, UCL Queen Square Institute of Neurology, London, UK & Stichting Epilepsie Instellingen Nederland (SEIN), Heemstede, Netherlands)
- Ashwani Jha, et al. Sudden Unexpected Death in Epilepsy: A PersonaliZed Prediction Tool. Neurology Apr 2021, 10.1212/WNL.0000000000011849;
- Dr Ashwani Jha's academic profile
- Professor Ley Sander's academic profile
- Dr Beate Diehl's academic profile
Short-form article: Ashwani Jha, et al. Sudden Unexpected Death in Epilepsy: A Personalized Prediction Tool. Neurology 2021;00:1. doi:10.1212/WNL.0000000000011849
Individualised model predictions of SUDEP risk. Predicted SUDEP risk for individuals A–J (mean [circle] and 80% [red line] and 95% [black line] certainty range). H had fewer GTCS than A, but higher risk and suffered SUDEP. Dashed line = average population risk.