Annual Meeting Abstracts: View

  • (Abst. 3.092), 2018
  • Electrographic Events Recorded by a Responsive Neurostimulator May Provide an Objective Assessment of Clinical Seizure Burden
  • Authors: David C. Spencer, Oregon Health and Science University; Mark S. Quigg, University of Virginia School of Medicine; Nathan B. Fountain, University of Virginia School of Medicine; Beata Jarosiewicz, NeuroPace, Inc.; Tara L. Skarpaas, NeuroPace, Inc.; and Martha J. Morrell, Stanford University / NeuroPace, Inc.
  • Content:

    Rationale: Studies have demonstrated that interictal epileptiform activity recorded by the RNS® System fluctuates with both circadian [Spencer et al., 2015; Baud et al., 2018] and slower multidien rhythms [Baud et al., 2018]. In addition, Baud et al., found that electrographic seizures (ES) recorded by the neurostimulator preferentially occurred during narrow phases of these rhythms suggesting that it may be possible to anticipate epochs of heightened seizure risk. However, the relationship between the recorded ES and patient-reported clinical seizures (CS) was not assessed. Here we sought to understand the relationship between the ES recorded by the RNS System and the CS reported by patients. Methods: The study focused on electrographic events recorded by the RNS System termed long episodes (LEs), which consist of abnormal electrocorticographic (ECoG) activity detected by the neurostimulator that did not return to baseline within a period of time (e.g. 30 seconds) predefined by the physician. These events often correspond to ES, and the criteria for defining a LE could be adjusted based on the duration of an individual patient’s ES. Hourly counts of LEs are nearly continuously stored by the device and available for analysis after upload to database. We assessed the relationship between daily counts of LEs and CS in a subset of patients for whom LE reliably identified ES [Quigg et al., 2015].  First, a cross-correlation analysis was conducted for patients with complete seizure diary data to assess if recorded LEs truly corresponded in a significant time-linked fashion to reported daily CS. Second, the positive and negative predictive value of LEs was calculated. Additional analyses assessed the impact of clinical characteristics (e.g. region of onset) on the relationship between LE and CS. Results: Fifty-four of the 128 patients in Quigg et al. had LE’s that reliably identified ES over 84 day periods with stable detection settings. In these patients the median number of LE per day was 0.7 (IQR: 0.3 – 23) and the median number of CS per day was 0.2 (IQR: 0.1 – 0.4). Cross-correlation analysis (Figure 1) showed a significant (p<0.0000001) peak in the correlation coefficients around time zero indicating that the two events, LEs and patient-reported CS, tended to co-occur. The negative predictive value of LEs in these patients was 92% (95% CI: 85-96%), indicating that a seizure-free day as determined by the neurostimulator was typically also a seizure-free day according to patient report. In contrast, the positive predictive value was 43% (95% CI: 35-52%), indicating that many of the recorded LE’s were not reported by the patient as CS. Conclusions: LE and CS events tended to co-occur, suggesting that, for some patients, the RNS System may provide an objective assessment of clinical seizure burden. However, in most patients self-reported CS may underestimate the extent to which ES occur. This reinforces the epileptologists’ concern that seizure diaries may under-represent the extent to which seizures occur because ES are subclinical, occur during sleep or are forgotten. Funding: None
  • Figures:
  • Figure 1