Authors :
Presenting Author: Vladimir Vashin, BS – UTHealth Houston
Clarissa Hoffman, BS – McGovern Medical School
Erika Davila, BS – UTHealth Houston
Sandipan Pati, MD – University of Minnesota
Nuria Lacuey, MD, PhD – UTHealth Houston
Yuri Dabaghian, PhD – UTHealth Houston
Samden Lhatoo, MD – UTHealth Houston
Rationale:
Sudden Unexpected Death in Epilepsy (SUDEP) is the leading direct cause of epilepsy-related mortality and no reliable premortem biomarkers exist. The emergence of ultra-long intracranial recordings from responsive neurostimulation (RNS) devices offers a new opportunity to investigate premortem neural dynamics. The RNS monitors continuous intracranial EEG data and captures the hourly counts of detected interictal epileptiform activity (IEA). We hypothesized that long-term changes in the structure of IEA accumulation may identify patients who are at higher risk of SUDEP. Using a novel analytical framework, we aimed to identify biomarkers of SUDEP from RNS IEA alone. Methods:
We analyzed an average of 1,066 days (range: 229–3,119) of IEA detection recordings from 16 patients with medically intractable epilepsy with implanted RNS device, who died with the device still on. Six patients were excluded due to having less than 9 months of data (five) or corrupted files (one). A total of 10 individuals were further analyzed and were divided in two cohorts. Six were classified as a SUDEP cohort (SC) and four as non SUDEP cohort (nSC). To assess the structure of IEA accumulation, we used Vladimir Arnold’s β-score—a scale-invariant metric of “evenness” in numerical patterns. It distinguishes randomly scattered values from clustered or unevenly spaced ones, yielding a dimensionless score analogous to a z-score. IEA accumulation evenness was quantified in sliding windows matched to each patient’s dominant cycle, identified via a cyclicity analysis. To test whether disorder increased over time, we compared β-scores across (1) three equal-duration epochs, and (2) the final month versus the remainder of the recording. Statistical significance was assessed using the Wilcoxon rank-sum test with Bonferroni correction.
Results:
Cyclicity analysis revealed a dominant circadian peak in all patients, supporting the use of 24-hour windows for β-score evaluation. Most IEA accumulation traces exhibited mundane disorder, with β-scores largely within the expected stochastic range (1.4 ≤ β ≤ 3.6). Three out of four nSC patients showed a decrease in β-score in the final third of their recordings, and all four had a significant decrease during the last month. In contrast, five out of six SC patients exhibited a significant increase in β-score during both the final third and the final month of data.Conclusions:
We conclude that a progressive increase in IEA accumulation disorder during treatment, mainly during the last month preceding the fatal event, may be a biomarker for SUDEP. Further studies are needed to confirm our observations.Funding: None