Authors :
Presenting Author: Saboo Krishnakant, PhD – University of California, San Francisco
Joline Fan, MD, MS – Uiversity of California, San Francisco
Ehsan Tadayon, MD – University of California, San Francisco
Edward Chang, MD – University of California, San Francisco
Vikram Rao, MD – Department of Neurology and Weill Institute for Neurosciences, University of California, San Francisco
Ankit Khambhati, PhD – University of California, San Francisco
Rationale:
State-dependent effects of responsive neurostimulation (RNS) on brain network dynamics in epilepsy and on therapeutic efficacy are not well understood. Discerning these can help optimize device parameters. Circadian changes in the aperiodic component of electrocorticography (ECoG) are a prognostic indicator of response to RNS [1]. ECoG functional connectivity (FC) forecasts seizure risk over a 24hr period [2] and has been implicated in the therapeutic responsiveness of RNS [3]. Moreover, effectiveness of stimulation depends on the seizure risk state [4]. Therefore, we investigated whether circadian dynamics of networks, and their modulation through RNS, predict ongoing seizure control.Methods:
We studied participants with drug-resistant epilepsy who were treated with RNS through bilateral hippocampal leads. Patient-reported seizure counts were collected longitudinally, averaged, and compared to baseline to quantify clinical outcome. We investigated chronic interictal ECoG intermittently recorded by the RNS device. FC within the left and right hippocampus was extracted for theta, alpha, beta, and gamma bands. We separately tracked the circadian cycles of band-specific FC (FC-CC) for each hippocampus. Circadian variation in the FC (amplitude) and the hour at which FC was strongest (peak hour) were measured using Lomb-Scargle analysis for irregularly sampled timeseries. For each band, we computed: (i) Total amplitude: the total amount of cycling in the network; (ii) Peak sync: the extent to which the network was cycling in synchrony. Lastly, we tracked the hourly counts of closed-loop stimulation events triggered by the device in each person.
Results:
We analyzed 100,137 ECoG clips collected over 64,820 days from 38 people who reported their seizure counts a total of 145 times. We first tested whether network-level synchronization of circadian cycles is related to seizure reduction. We found that high FC-CC peak sync in alpha was associated with a better outcome (r=0.43, P=0.007) (Figure 1a) and predicted time-matched longitudinal outcomes (r=0.42, P< 0.001). We next tested whether the distribution of stimulation over the course of the day was predictive of outcome. Stimulation delivered outside 7AM – 11AM was most predictive of the outcome (r=−0.50, P=0.001) (Figure 1b), suggesting time-of-day effects of stimulation. A multivariate model including network cycling features and stimulation predicted time-matched longitudinal outcomes (r=0.63, P=0.001). Conclusions:
Strength and alignment of FC circadian cycles across bilateral hippocampi predicted better outcomes as early as 6 months post-implant. Stimulation at certain times of the day modulated circadian network dynamics in treatment responders. These biomarkers can be clinically useful to predict and track seizure reduction in RNS.
References:
[1] Charlebois, et al., Epilepsia 65.5 (2024): 1360-1373.
[2] Khambhati, et al., Nat Med 30.10 (2024): 2787-2790.
[3] Khambhati, et al., Sci Transl Med 13.608 (2021): eabf6588.
[4] Chiang, et al., Brain Stimul 14.2 (2021): 366-375.
Funding:
Schmidt Science Fellows