Authors :
Nina Ghosn, PhD – University of Pennsylvania
Presenting Author: Kerry Nix, MS – University of Pennsylvania
Kevin Xie, PhD – University of Pennsylvania
William Ojemann, BS – University of Pennsylvania
Colin Ellis, MD – University of Pennsylvania
Genna Waldman, MD – Neurology
Eli Cornblath, MD, PhD – University of Pennsylvania
Erin Conrad, MD, MA – University of Pennsylvania
Brian Litt, MD – University of Pennsylvania
Rationale:
In the epilepsy monitoring unit (EMU), anti-seizure medications (ASM) are reduced to induce seizures for diagnostic purposes. However, we lack standardized protocols to taper medications, which may contribute to increased morbidity through longer implantations and dangerous, severe seizures. Understanding how medication taper influences these outcomes can help optimize patient care, reduce risks, and improve resource allocation. We aim to develop guidance for optimal medication taper strategies to maximize diagnostic yield while minimizing risk.
Methods:
We extracted data from the electronic medical record from 639 EMU patient records between 2017 and 2023 using natural language processing. We applied pharmacokinetic models to estimate serum ASM concentrations. We applied regression with LASSO and mixed effects models to identify predictors of length of stay, and mixed effects models to identify predictors risk of convulsion.
Results:
Overall, lower baseline seizure frequency increased convulsion likelihood (OR = 0.000, 95% CI: 0.000–0.238, p = 0.006), while preictal ASM load and history of convulsions were not significantly associated with convulsion likelihood (preictal ASM load: OR = 0.333, 95% CI: 0.000–3.189, p = 0.472; history of convulsions: OR = 2.784, 95% CI: 0.910–9.275, p = 0.076). However, there was an interaction between preictal ASM load and patient risk factors: In patients without a history of convulsions, convulsive seizures occurred at lower ASM loads, and higher preictal ASM load was protective (GLMM, OR = 0.084, 95% CI: 0.018–0.392, p = 0.0017). Preictal ASM load, baseline seizure frequency, and ASM trough were not associated with convulsive likelihood (p > 0.05). In patients with a history of convulsions, ASM loads were not distinguishable between convulsive and non-convulsive seizures (GLMM, OR = 1.16, 95% CI: 0.37–3.62, p = 0.80). Time to first seizure (β = 0.73, 95% CI 0.63–0.83, p < 0.001) and number of ASMs at baseline (β = 11.79, 95% CI 8.53–15.06, p < 0.001) were positively correlated with longer hospital stays.