Accurate detection of spontaneous seizures using a generalized linear model with external validation.

OBJECTIVE: Seizure detection is a major facet of electroencephalography (EEG) analysis in neurocritical care, epilepsy diagnosis and management, and the instantiation of novel therapies such as closed-loop stimulation or optogenetic control of seizures. It is also of increased importance in high-throughput, robust, and reproducible pre-clinical research. However, seizure detectors are not widely relied upon in either clinical or research settings due to limited validation. In this study, we create a high-performance seizure-detection approach, validated in multiple data sets, with the intention that such a system could be available to users for multiple purposes. METHODS: We introduce a generalized linear model trained on 141 EEG signal features for classification of seizures in continuous EEG for two data sets. In the first (Focal Epilepsy) data set consisting of 16 rats with focal epilepsy, we collected 1012 spontaneous seizures over 3 months of 24/7 recording. We trained a generalized linear model on the 141 features representing 20 feature classes, including univariate and multivariate, linear and nonlinear, time, and frequency domains. We tested performance on multiple hold-out test data sets. We then used the trained model in a second (Multifocal Epilepsy) data set consisting of 96 rats with 2883 spontaneous multifocal seizures. RESULTS: From the Focal Epilepsy data set, we built a pooled classifier with an Area Under the Receiver Operating Characteristic (AUROC) of 0.995 and leave-one-out classifiers with an AUROC of 0.962. We validated our method within the independently constructed Multifocal Epilepsy data set, resulting in a pooled AUROC of 0.963. We separately validated a model trained exclusively on the Focal Epilepsy data set and tested on the held-out Multifocal Epilepsy data set with an AUROC of 0.890. Latency to detection was under 5 seconds for over 80% of seizures and under 12 seconds for over 99% of seizures. SIGNIFICANCE: This method achieves the highest performance published for seizure detection on multiple independent data sets. This method of seizure detection can be applied to automated EEG analysis pipelines as well as closed loop interventional approaches, and can be especially useful in the setting of research using animals in which there is an increased need for standardization and high-throughput analysis of large number of seizures.

Evaluation of subchronic administration of antiseizure drugs in spontaneously seizing rats.

OBJECTIVE: Approximately 30% of patients with epilepsy do not experience full seizure control on their antiseizure drug (ASD) regimen. Historically, screening for novel ASDs has relied on evaluating efficacy following a single administration of a test compound in either acute electrical or chemical seizure induction. However, the use of animal models of spontaneous seizures and repeated administration of test compounds may better differentiate novel compounds. Therefore, this approach has been instituted as part of the National Institute of Neurological Disorders and Stroke Epilepsy Therapy Screening Program screening paradigm for pharmacoresistant epilepsy. METHODS: Rats were treated with intraperitoneal kainic acid to induce status epilepticus and subsequent spontaneous recurrent seizures. After 12 weeks, rats were enrolled in drug screening studies. Using a 2-week crossover design, selected ASDs were evaluated for their ability to protect against spontaneous seizures, using a video-electroencephalographic monitoring system and automated seizure detection. Sixteen clinically available compounds were administered at maximally tolerated doses in this model. Dose intervals (1-3 treatments/d) were selected based on known half-lives for each compound. RESULTS: Carbamazepine (90 mg/kg/d), phenobarbital (30 mg/kg/d), and ezogabine (15 mg/kg/d) significantly reduced seizure burden at the doses evaluated. In addition, a dose-response study of topiramate (20-600 mg/kg/d) demonstrated that this compound reduced seizure burden at both therapeutic and supratherapeutic doses. However, none of the 16 ASDs conferred complete seizure freedom during the testing period at the doses tested. SIGNIFICANCE: Despite reductions in seizure burden, the lack of full seizure freedom for any ASD tested suggests that this screening paradigm may be useful for testing novel compounds with potential utility in pharmacoresistant epilepsy.

Evaluation of antiseizure drug efficacy and tolerability in the rat lamotrigine-resistant amygdala kindling model.

OBJECTIVE: The lamotrigine-resistant amygdala kindling model uses repeated administration of a low dose of lamotrigine during the kindling process to produce resistance to lamotrigine, which also extends to some other antiseizure drugs (ASDs). This model of pharmacoresistant epilepsy has been incorporated into the testing scheme utilized by the Epilepsy Therapy Screening Program (ETSP). Although some ASDs have been evaluated in this model, a comprehensive evaluation of ASD prototypes has not been reported. METHODS: Following depth electrode implantation and recovery, rats were exposed to lamotrigine (5 mg/kg, i.p.) prior to each stimulation during the kindling development process (~3 weeks). A test dose of lamotrigine was used to confirm that fully kindled rats were lamotrigine-resistant. Efficacy (unambiguous protection against electrically elicited convulsive seizures) was defined as a Racine score < 3 in the absence of overt compound-induced side effects. Various ASDs, comprising several mechanistic classes, were administered to fully kindled, lamotrigine-resistant rats. Where possible, multiple doses of each drug were administered in order to obtain median effective dose (ED50) values. RESULTS: Five sodium channel blockers tested (eslicarbazepine, lacosamide, lamotrigine, phenytoin, and rufinamide) were either not efficacious or effective only at doses that were not well-tolerated in this model. In contrast, compounds targeting either GABA receptors (clobazam, clonazepam, phenobarbital) or GABA-uptake proteins (tiagabine) produced dose-dependent efficacy against convulsive seizures. Compounds acting to modulate Ca2+ channels show differential activity: Ethosuximide was not effective, whereas gabapentin was moderately efficacious. Ezogabine and valproate were also highly effective, whereas topiramate and levetiracetam were not effective at the doses tested. SIGNIFICANCE: These results strengthen the conclusion that the lamotrigine-resistant amygdala kindling model demonstrates pharmacoresistance to certain ASDs, including, but not limited to, sodium channel blockers, and supports the utility of the model for helping to identify compounds with potential efficacy against pharmacoresistant seizures.

Correction of medication nonadherence results in better seizure outcomes than dose escalation in a novel preclinical epilepsy model of adherence.

OBJECTIVE: Medication nonadherence directly contributes to poor seizure control. A lack of emphasis on correcting poor adherence and failures in patient adherence can result in unwarranted alterations to a patient's drug regimen. We have modeled nonadherent patients in an animal model of epilepsy to study how alterations to pharmacotherapy, made without consideration of a patient's adherence, result in changes to seizure control. METHODS: Newly diagnosed rats with epilepsy were treated with carbamazepine (CBZ) during a 4-week baseline period to establish their baseline seizure rate in the presence of 50% adherence. Next, animals were randomized to one of three treatment interventions and monitored for 6 weeks. Groups included: (1) no change in therapy-rats continued the 50% adherent paradigm; (2) dose escalation-the dose of CBZ was doubled, and the 50% adherent paradigm continued; and (3) nonadherence corrected-rats continued the initial dose of CBZ, but the adherence rate was adjusted to 100% (ie, fully adherent). RESULTS: The rats in the no change in therapy arm displayed a 61% increase in seizure burden over the 6-week intervention phase. Similarly, rats in the dose escalation arm had a 66% worsening of their daily seizure burden. In contrast, rats in the nonadherence corrected arm displayed a 33% reduction in their daily seizure burden; a significant improvement when compared to the normalized seizure burden scores of rats in the other two treatment arms (P < 0.01). SIGNIFICANCE: We found that failure to correct medication nonadherence resulted in an increase in daily seizure burden in rats, even following dose escalation. In the presence of nonadherence, dose escalation worsened seizure control. In contrast, correcting nonadherence alone resulted in improved seizure control. These findings suggest that improving adherence should be prioritized over dose escalation in the clinical management of uncontrolled epilepsy.

Hippocampal TNFα Signaling Contributes to Seizure Generation in an Infection-Induced Mouse Model of Limbic Epilepsy.

Central nervous system infection can induce epilepsy that is often refractory to established antiseizure drugs. Previous studies in the Theiler's murine encephalomyelitis virus (TMEV)-induced mouse model of limbic epilepsy have demonstrated the importance of inflammation, especially that mediated by tumor necrosis factor-α (TNFα), in the development of acute seizures. TNFα modulates glutamate receptor trafficking via TNF receptor 1 (TNFR1) to cause increased excitatory synaptic transmission. Therefore, we hypothesized that an increase in TNFα signaling after TMEV infection might contribute to acute seizures. We found a significant increase in both mRNA and protein levels of TNFα and the protein expression ratio of TNF receptors (TNFR1:TNFR2) in the hippocampus, a brain region most likely involved in seizure initiation, after TMEV infection, which suggests that TNFα signaling, predominantly through TNFR1, may contribute to limbic hyperexcitability. An increase in hippocampal cell-surface glutamate receptor expression was also observed during acute seizures. Although pharmacological inhibition of TNFR1-mediated signaling had no effect on acute seizures, several lines of genetically modified animals deficient in either TNFα or TNFRs had robust changes in seizure incidence and severity after TMEV infection. TNFR2-/- mice were highly susceptible to developing acute seizures, suggesting that TNFR2-mediated signaling may provide beneficial effects during the acute seizure period. Taken together, the present results suggest that inflammation in the hippocampus, caused predominantly by TNFα signaling, contributes to hyperexcitability and acute seizures after TMEV infection. Pharmacotherapies designed to suppress TNFR1-mediated or augment TNFR2-mediated effects of TNFα may provide antiseizure and disease-modifying effects after central nervous system infection.

Development and pharmacologic characterization of the rat 6 Hz model of partial seizures.

OBJECTIVE: The mouse 6 Hz model of psychomotor seizures is a well-established and commonly used preclinical model for antiseizure drug (ASD) discovery. Despite its widespread use both in the identification and differentiation of novel ASDs in mice, a corresponding assay in rats has not been developed. We established a method for 6 Hz seizure induction in rats, with seizure behaviors similar to those observed in mice including head nod, jaw clonus, and forelimb clonus. METHODS: A convulsive current that elicits these seizure behaviors in 97% of rats (CC97 ) was determined using a Probit analysis. Numerous prototype ASDs were evaluated in this model using stimulus intensities of 1.5× and 2× the CC97 , which is comparable to the approach used in the mouse 6 Hz seizure model (e.g., 32 and 44 mA stimulus intensities). The ASDs evaluated include carbamazepine, clobazam, clonazepam, eslicarbazepine, ethosuximide, ezogabine, gabapentin, lacosamide, lamotrigine, levetiracetam, phenobarbital, phenytoin, rufinamide, tiagabine, topiramate, and sodium valproate. Median effective dose (ED50 ) and median toxic (motor impairment) dose (TD50 ) values were obtained for each compound. RESULTS: Compounds that were effective at the 1.5 × CC97 stimulus intensity at protective index (PI) values >1 included clobazam, ethosuximide, ezogabine, levetiracetam, phenobarbital, and sodium valproate. Compounds that were effective at the 2 × CC97 stimulus intensity at PI values >1 included ezogabine, phenobarbital, and sodium valproate. SIGNIFICANCE: In a manner similar to the use of the mouse 6 Hz model, development of a rat 6 Hz test will aid in the differentiation of ASDs, as well as in study design and dose selection for chronic rat models of pharmacoresistant epilepsy. The limited number of established ASDs with demonstrable efficacy at the higher stimulus intensity suggests that, like the mouse 6 Hz 44 mA model, the rat 6 Hz seizure model may be a useful screening tool for pharmacoresistant seizures.

The impact of nonadherence to antiseizure drugs on seizure outcomes in an animal model of epilepsy.

OBJECTIVE: Nonadherence to prescribed dosing regimens is a significant problem in the treatment of pediatric and adult chronic epilepsy, and can result in severe consequences to patient outcomes. In this first-of-kind preclinical study, the impact of nonadherence on seizure control was studied by simulating human patterns of nonadherence in an animal epilepsy model. METHODS: In study 1, three different patterns of nonadherence were modeled in newly diagnosed epileptic rats treated with carbamazepine: perfect adherence (100% of pellets contained carbamazepine), variable nonadherence (50% of pellets contained carbamazepine with different dosing patterns between animals), and complete nonadherence (0% of pellets contained carbamazepine). In study 2, a cohort of newly diagnosed epileptic rats were subjected to a "drug holiday" nonadherence paradigm, that is, a 2-week on (100%), 2-week off (0%), and 2-week on (100%) carbamazepine paradigm. RESULTS: In the first experiment, the 100% (0.3 ± 0.2 SD convulsive seizures per day) adherent cohort demonstrated better seizure control than either the 0% (1.1 ± 0.8 SD) or 50% (0.8 ± 0.6 SD) adherent cohorts, which had similar levels of seizure control. In the second study, poor seizure control was exhibited during the second 2 weeks; that is, the drug holiday epoch; however, this did not negatively affect restoration of seizure control upon reinstatement of CBZ. SIGNIFICANCE: The results from this pilot investigation suggest that nonadherence to carbamazepine is associated with significant negative but reversible effects on seizure control in an animal model of epilepsy. Furthermore, these results demonstrate that animal studies of nonadherence can yield potentially important and translatable insights into the consequences of nonadherence on seizure control.

Repeated low-dose kainate administration in C57BL/6J mice produces temporal lobe epilepsy pathology but infrequent spontaneous seizures.

More efficient or translationally relevant approaches are needed to model acquired temporal lobe epilepsy (TLE) in genetically tractable mice. The high costs associated with breeding and maintaining transgenic, knock-in, or knock-out lines place a high value on the efficiency of induction and animal survivability. Herein, we describe our approaches to model acquired epilepsy in C57BL/6J mice using repeated, low-dose kainate (KA) administration paradigms. Four paradigms (i.p.) were tested for their ability to induce status epilepticus (SE), temporal lobe pathology, and the development of epilepsy. All four paradigms reliably induce behavioral and/or electrographic SE without mortality over a 7d period. Two of the four paradigms investigated produce features indicative of TLE pathology, including hippocampal cell death, widespread astrogliosis, and astrocyte expression of mGluR5, a feature commonly reported in TLE models. Three of the investigated paradigms were able to produce aberrant electrographic features, such as interictal spiking in cortex. However, only one paradigm, previously published by others, produces spontaneous recurrent seizures over an eight week period. Presentation of spontaneous seizures is rare (N=2/14), with epilepsy preferentially developing in animals having a high number of seizures during SE. Overall, repeated, low-dose KA administration improves the efficiency and pathological relevance of a systemic KA insult, but does not produce a robust epilepsy phenotype under the experimental paradigms described herein.

A novel open-source drug-delivery system that allows for first-of-kind simulation of nonadherence to pharmacological interventions in animal disease models.

BACKGROUND: Nonadherence to a physician-prescribed therapeutic intervention is a costly, dangerous, and sometimes fatal concern in healthcare. To date, the study of nonadherence has been constrained to clinical studies. The novel approach described herein allows for the preclinical study of nonadherence in etiologically relevant disease animal model systems. NEW METHOD: The method herein describes a novel computer-automated pellet delivery system which allows for the study of nonadherence in animals. This system described herein allows for tight experimenter control of treatment using a drug-in-food protocol. Food-restricted animals receive either medicated or unmedicated pellets, designed to mimic either "taking" or "missing" a drug. RESULTS: The system described permits the distribution of medicated or unmedicated food pellets on an experimenter-defined feeding schedule. The flexibility of this system permits the delivery of drug according to the known pharmacokinetics of investigational drugs. COMPARISON WITH OTHER METHODS: Current clinical adherence research relies on medication-event monitoring system (MEMS) tracking caps, which allows clinicians to directly monitor patient adherence. However, correlating the effects of nonadherence to efficacy still relies on the accuracy of patient journals. CONCLUSION: This system allows for the design of studies to address the impact of nonadherence in an etiologically relevant animal model. Given methodological and ethical concerns of designing clinical studies of nonadherence, animal studies are critical to better understand medication adherence. While the system described was designed to measure the impact of nonadherence on seizure control, it is clear that the utility of this system extends beyond epilepsy to include other disease states.

The expression of kainate receptor subunits in hippocampal astrocytes after experimentally induced status epilepticus.

Astrocytes have emerged as active participants of synaptic transmission and are increasingly implicated in neurologic disorders including epilepsy. Adult glial fibrillary acidic protein (GFAP)-positive hippocampal astrocytes are not known for ionotropic glutamate receptor expression under basal conditions. Using a chemoconvulsive status epilepticus (SE) model of temporal lobe epilepsy, we show by immunohistochemistry and colocalization analysis that reactive hippocampal astrocytes express kainate receptor (KAR) subunits after SE. In the CA1 region, GluK1, GluK2/3, GluK4, and GluK5 subunit expression was observed in GFAP-positive astrocytes during the seizure-free or "latent" period 1 week after SE. At 8 weeks after SE, a time after SE when spontaneous behavioral seizures occur, the GluK1 and GluK5 subunits remained expressed at significant levels. Kainate receptor subunit expression was found in astrocytes in the hippocampus and surrounding cortex but not in GFAP-positive astrocytes of striatum, olfactory bulb, or brainstem. To examine hippocampal KAR expression more broadly, astroglial-enriched tissue fractions were prepared from dissected hippocampi and were found to have greater GluK4 expression after SE than controls. These results demonstrate that astrocytes begin to express KARs after seizure activity and suggest that their expression may contribute to the pathophysiology of epilepsy.

Human neocortical electrical activity recorded on nonpenetrating microwire arrays: applicability for neuroprostheses.

OBJECT: The goal of this study was to determine whether a nonpenetrating, high-density microwire array could provide sufficient information to serve as the interface for decoding motor cortical signals. METHODS: Arrays of nonpenetrating microwires were implanted over the human motor cortex in 2 patients. The patients performed directed stereotypical reaching movements in 2 directions. The resulting data were used to determine whether the reach direction could be distinguished through a frequency power analysis. RESULTS: Correlation analysis revealed decreasing signal correlation with distance. The gamma-band power during motor planning allowed binary classification of gross directionality in the reaching movements. The degree of power change was correlated to the underlying gyral pattern. CONCLUSIONS: The nonpenetrating microwire platform showed good potential for allowing differentiated signals to be recorded with high spatial fidelity without cortical penetration.

Hardware considerations of a spatial filter for decorrelating high-density multielectrode neural recordings.

High density implantable microelectrode arrays record large amounts of highly correlated data, which causes large strains on limited bandwidth telemetry systems. Previous work has shown that the use of a spatial filter can significantly reduce the number of channels that must be transmitted to adequately represent the data. However, the limitations on power and size for an implantable neuroprosthetic device impose significant limitations on the computational complexity of the spatial filter. We assess the performance of the floating point operations of spatial filtering and show that it can be approximated to integers with negligible losses to signal fidelity, thus reducing the computational complexity.