Slide Session III: Clinical / Electrophysiology

Psychiatric and Behavioral Side Effects of Anti-epileptic Drugs in Adults with Epilepsy

Presenting Author: Cel Ezeani

Collaborating Authors: Baibing Chen, Asif Javed, Kamil Detyniecki, Richard Buchsbaum, Hyunmi Choi, Stanley R. Resor, Lawrence J. Hirsch

Objective: Psychiatric/behavioral side effects (PSE) are commonly associated with the use of antiepileptic drugs (AED). An earlier study looked at PSEs of the newer AEDs alone. The purpose of this study was to examine the PSE profiles of a broad list of AEDs in addition to using a larger cohort of adult patients. PSE profiles were also compared to that of lamotrigine.

Methods: We reviewed the charts of 2246 adult patients (age ≥ 16) seen at both Columbia and Yale Comprehensive Epilepsy Centers and who were newly started on one of the AEDs between January 1 2000 and October 1 2012. PSE attributed to each AED was compared to PSE rates of the other AEDs using a logistic regression model. AEDs were also compared to lamotrigine because of its relatively safe PSE profile and wide therapeutic use in epilepsy therapy. Rates of toxicity - defined as PSE causing a change in drug dose or discontinuation - were also compared between AEDs.

Results: Overall, 178 of 2246 (8%) patients newly started on one of these AEDs experienced PSEs. The average rate of AED-related PSEs for a single AED was 3.8%. We found that levetiracetam (OR 4.91, p<0.0001) and tiagabine (OR 3.87, p=0.005) had significantly greater rates of PSE compared to other AEDs. PSEs attributed to either of these two AEDs were also more likely to be toxic (levetiracetam: OR 3.95, p<0.0001; tiagabine: OR 5.97, p=<0.0001). On the other hand, lamotrigine, (OR 0.33, p<0.0001), gabapentin (OR 0.33, p=0.016), carbamazepine (OR 0.38, p=0.020), pregabalin (OR 0.39, p=0.039), and valproate (OR 0.26, p=0.008) had significantly lower PSE rates. However, only lamotrigine (OR 0.27, p<0.0001) and valproate (OR 0.30, p=0.037) were significantly less likely than other AEDs to be toxic (See table). Levetiracetam and tiagabine were eight and ten times respectively more likely to cause PSEs (p<0.001) when compared to lamotrigine. They were also significantly more likely to cause discontinuation (<0.001). A history of any psychiatric condition was found to significantly influence AED-related PSEs (OR 1.79, p<0.001). When we controlled for psychiatric history levetiracetam (OR 6.44, p<0.0001) and tiagabine (OR 6.68, p=0.002) had a seemingly stronger influence on causing PSEs. Both also had a greater likelihood of causing toxicity than other AEDs (p<0.0001).

Conclusions: Significant differences exist amongst several AEDs with regard to their PSE profiles. Patients taking levetiracetam and tiagabine experienced significantly more PSEs than other AEDs and were more likely to be toxic. A history of a psychiatric condition significantly predicts PSEs in adult patients with epilepsy. Knowledge of the PSE profile of AEDs will aid clinicians in providing more effective treatment and possibly improve patient compliance.



PSE attributed (%)

OR (95% CI)


PSE causing toxicity (%)

OR (95% CI)





0.38 (0.17, 0.86)



0.38 (0.14, 1.04)





0.28 (0.04, 2.04)



0.43 (0.06, 3.13)





0.33 (0.14, 0.81)



0.41 (0.15, 1.11)





1.15 (0.53, 2.48)



1.24 (0.50, 3.06)





4.91 (3.72, 6.48)



3.95 (2.81, 5.54)





0.33 (0.21, 0.53)



0.27 (0.15, 0.51)





0.64 (0.31, 1.31)



0.48 (0.18, 1.30)





0.33 (0.05, 2.42)



0.51 (0.07, 3.71)





0.39 (0.16, 0.95)



0.48 (0.18, 1.30)





1.11(0.61, 2.01)



1.74 (0.95, 3.19)





3.87 (1.49, 10.03)



5.97 (2.29, 15.57)





0.70 (0.37, 1.34)



0.98 (0.49, 1.94)





1.08 (0.26, 4.49)








0.26 (0.10, 0.70)



0.30 (0.09, 0.93)





1.20(0.75, 1.90)



1.21 (0.69, 2.11)


Recording Gamma Frequencies Using Scalp Electrodes: Comparison of Simultaneous Scalp and Subdural EEG

Presenting Author: Ognen Petroff

Collaborating Authors: Irina Goncharova, Hitten Zaveri

The first studies recording both scalp and cortical EEG noted a significant loss of EEG power and loss of fidelity in the scalp recordings. Most of the subsequent work focused on the loss of fidelity of epileptiform activity and evoked potentials (30-200 ms transients). A special interest for gamma frequency range events emerged as they correlated with cognitive functions. Intracranial EEG reliably records gamma (20-80 Hz) and omega (60-120 Hz) frequencies and, with specialized amplifiers and electrodes, ripples, fast ripples and 600 Hz bursts.

Twenty patients (age 18–55) from the Yale University Epilepsy Surgery Program undergoing intracranial EEG monitoring for surgical evaluation were recruited. Scalp EEG recordings were made using Grass platinum needle electrodes located over the C3, C4, O1, and O2 sites. Simultaneous recordings were obtained from subjacent subdural strip electrodes (4mm diameter platinum disks with 2.3mm exposed surface diameter) as previously reported (Zaveri et al Clin Neurophysiol 2010;121:311–317). Offline analysis was performed with custom software written in a mixture of high level languages and MATLAB. Scalp and intracranial EEG epochs, 1 hour in duration, at least 6 hours removed from a seizure, 3 or 4 days after electrode implantation surgery, during wakefulness between 9–10 AM were selected for analysis. Scalp and subdural recordings were examined for artifacts (EMG, movements, ECG, 60 Hz interference, etc) and segmented with 1-second resolution. EEG power of artifact-free EEG segments was obtained for each electrode contact studied and averaged over the epoch. Average power (millivolts2) was calculated as the signal power between 0.5 and 128 Hz in 0.5 Hz increments. The ratio of scalp EEG electrode to subjacent subdural EEG electrode signal power was calculated between 0.5 and 128 Hz in 0.5 Hz increments.

Overall intracranial and scalp EEG power decreased from 1 to 60 Hz by three orders of magnitude. The ratio of scalp to intracranial power in the beta (13-25 Hz) frequencies ranged from 0.05 to 0.06, whereas the gamma (35-55 Hz) from 0.06 to 0.10. The scalp to intracranial EEG power ratio increased linearly with frequency from 40 to 60 Hz, which suggests a linear increase in scalp power at higher frequencies. Extracranial signals likely account for this significant increase in scalp EEG power and would confound measurement of the gamma frequencies using scalp EEG.

Lateralized Rhythmic Delta Activity in Critically Ill Patients

Presenting Author: Nicolas Gaspard

Rationale: To describe a new variant of rhythmic delta activity encountered in the critically ill, its clinical correlates and compare it with focal polymorphic slowing and lateralized periodic discharges (LPDs or PLEDs).

Methods: Retrospective review of urgent EEG and continuous EEG reports 05/01/2011 and 04/30/2012 to identify patients with lateralized rhythmic delta activity (LRDA). Recordings, medical reports and imaging studies were reviewed. Statistical comparisons were performed with the Chi2-test or the Fisher Exact t-test.

Results: We identified 27 patients (4.7% of all the) with lateralized rhythmic delta activity (LRDA). Mean age was 52+/-3.5years (4mo-87years) and 11 were female. Twenty patients were admitted to an ICU and 17 were stuporous or comatose. Nineteen had a lateralizing neurological examination and the side of LRDA always agreed with the examination.

Twenty-four had a single unilateral focus, 2 had bilateral independent RDA (BIRDA) and 1 had two ipsilateral independent foci; 30 different patterns were identified. Eighteen were located in the frontal region, eight in the temporal region and four in the posterior region; this distribution was similar for LPDs. Most commonly, runs of LRDA consisted in 1-2c/s, 50-200µV, sinusoidal or saw-tooth monomorphic non-evolving waves and lasted less than 10s, although faster frequencies (2-3/s) and longer duration (up to 1 min) were encountered.

All patients except one had an acute or remote focal brain lesion. The most common etiologies were intracranial hemorrhage, ischemic stroke, subarachnoid hemorrhage, subdural hematoma, traumatic brain injury and CNS inflammatory/infectious disorder. This was overall similar to LPDs and focal polymorphic slow activity (PSA) except that acute intracranial hemorrhage was more frequent in the LRDA group than in the PSA group (p < 0.01). Systemic factors (infection, metabolic imbalance) were present in half of the cases.

Twenty-six patients had brain imaging and it was found to be abnormal in 24. Lesions involved the cortex or subcortical white matter, the deep white matter or deep grey structures in 18, 9 and 5 cases respectively. In case of a single focal lesion, it co-localized in the same region as LRDA in all but two cases (17/19) who had a thalamic (same side) and an extensive brainstem hemorrhage (including the upper midbrain and thalamus).

Focal periodic (LPDs or BIPDs/BIPLEDs) and sporadic epileptiform discharges were found in 12/27 and 7/27 patients respectively and always co-localized with LRDA. Seventeen patients (63%) had acute clinical and/or electrographic seizures during their stay; this proportion was similar to the one observed with LPDs/BIPDs (28/49 or 57%) but significantly higher than the one observed with PSA (14/105 or 13%; p=0.001). All were focal seizures and co-localized with LRDA.

Conclusions: We describe a new focal variant of RDA encountered in critically ill patients with an acute or remote brain lesion that bears the same clinical significance as LPDs/PLEDs, as more than half of the patients had acute seizures.

Consciousness during Seizures Determines the Accuracy of Patient Seizure Descriptions

Presenting Author: Kamil Detyniecki

Collaborating Authors: Cel Ezeani, Andrew Bauerschmidt, Robert B. Duckrow, F. Scott Winstanley, Lawrence J. Hirsch, Hal Blumenfeld

The goal of this study was to assess patients' reporting of their seizures. We also looked at predictors that could influence the accuracy of their report. Further, we examined the impact that deficits in memory and consciousness could have on patients' seizure descriptions.

We recruited patients undergoing VEEG evaluation in our institution. Daily, a member of the research team asked them if they had any seizure in the previous 24 hours. We compared their subjective response with a more objective measure by analyzing the video recordings of their seizures. Seizure type, localization, and presence of impaired consciousness were among the characteristics documented. We determined statistical significance using Chi-square or Fisher's exact tests where appropriate to evaluate relationships between patients' characteristics and seizure reporting. To determine which factors impacted seizure reporting, a multivariate binary logistic regression was performed.

115 patients were recruited, of which 65 met all criteria for analysis (confirmed epilepsy, completed questionnaires, seizures captured during VEEG monitoring).

Overall, we captured a total of 295 partial and 42 sGTC seizures. 30% of all seizures were not reported while 70% were reported. We found that SPS were more often reported than both CPS and sGTC in pair-wise comparisons (p=0.008 and <0.001 respectively). This was consistent with our finding on patient level, where patients with only SPS more often reported their seizures. We also found that loss of consciousness (OR 0.25; 95% CI 0.07 – 0.92; p=0.037) and temporal lobe onset (OR 0.17; 95% CI 0.08 – 0.37; p<0.001) had a negative influence on seizure reporting. Although patients who were awake at seizure onset reported more of their seizures (Chi-square 17.34; p<0.0001), sleep state was not a significant independent predictor of seizure reporting. We didn't observe an effect of seizure lateralization on seizure reporting.

This study demonstrates that seizures with impaired consciousness were more likely to go unreported suggesting that consciousness may be a factor influencing the ability of patients to recognize and accurately report their seizures. In addition, temporal lobe onset but not lateralization significantly influenced if seizures were reported. Further study is needed to investigate the impact of altered memory function and language impairment on seizure reporting.

Data Contradicting the New ILAE Classification of "Focal" Seizures

Presenting Author: Courtney Cunningham

Collaborating Authors: Andrew Shorten, Bridget Kiely, Michael McClurkin, Tenzin Choezom, Andrew Bauerschmidt, Hal Blumenfeld

Impaired consciousness in epilepsy has a major negative impact on patient quality of life. Our lab recently developed a standardized testing battery, the Responsiveness in Epilepsy Scale (RES), designed to assess altered behavior during seizures in an objective, prospective fashion. 92 patients undergoing continuous video/EEG monitoring (VEEG) for seizure evaluation at Yale New Haven Hospital were recruited for this study. Patients were tested with RES or RES-II during the ictal and immediate postictal periods. Ictal testing was performed for a total of 80 seizures (31 patients). Scores from RES-I testing were later converted from a 5-point to a 4-point scale, corresponding with RES-II, in order to simplify analysis.

Performance on the initial cycle of questions asked upon onset of partial seizures showed a bimodal distribution. On these initial questions, patients tended to be either entirely unimpaired or completely unresponsive in most seizures. For instance, 28 of 80 seizures scored the maximal score of "4" on the first question asked, while 44 of 80 seizures received the minimal score of "0". Only 8 seizures received intermediate scores on this initial question. This distribution of impairment was typically sustained until seizure termination.

These distinct patterns of impairment correlate with those seen in the traditional categories of "simple partial" and "complex partial" seizures. Our results demonstrate that focal seizures can often be cleanly separated into those with or without impaired responsiveness. This distinction is of critical importance in clinical decision making regarding a patient's ability to work, drive a car, or participate in other daily activities. The latest revision of the ILAE classification guidelines eliminates the distinction of simple partial vs. complex partial seizures; however our findings support the validity of this traditional classification.

Functional Connectivity in the Seizure Onset and Peri-Seizure Onset Area

Presenting Author: Hitten P. Zaveri

Collaborating Authors: Steven M. Pincus, Irina I. Goncharova, Robert B. Duckrow, Lawrence J. Hirsch, Dennis D. Spencer

Rationale: To test if functional connectivity measured from the background intracranial EEG (icEEG) varies with distance to the seizure onset area and with frequency of icEEG activity.

Methods: This study was conducted on 12 unselected adult patients with localization related epilepsy undergoing icEEG monitoring for possible surgery. Intracranial EEG electrode contacts were located from post-implantation CT and MR images and registered to the MRI of a standard brain to allow interpretation of results in the same space. A 1 hr icEEG epoch, recorded during wake and removed from seizures, was studied.

Coherence was estimated for all pairs of electrode contacts ipsilateral to the seizure onset area in 6 frequency bands: delta, theta, alpha, beta, gamma and a high-frequency band. The connectivity of each electrode contact was estimated as the average coherence between it and all electrode contacts within a specified distance.

Results: A graded relationship was observed between connectivity and distance to the seizure onset area such that electrodes with the greatest connectivity were closest to the seizure onset area and electrodes with the lowest connectivity were at a distance of several cm from the seizure onset area. The relationship between distance and connectivity was present primarily in the theta, alpha, beta and gamma frequency bands. Similar relationships were not observed for spike counts or spectral power in different frequency bands.

Conclusions: Patients who suffer from localization-related epilepsy have altered functional connectivity in the seizure onset and peri-seizure onset areas.