Early Identification of Candidates for Epilepsy Surgery: A Multicenter, Machine Learning, Prospective Validation Study.

BACKGROUND AND OBJECTIVES: Epilepsy surgery is often delayed. We previously developed machine learning (ML) models to identify candidates for resective epilepsy surgery earlier in the disease course. In this study, we report the prospective validation. METHODS: In this multicenter, prospective, longitudinal cohort study, random forest models were validated at a pediatric epilepsy center consisting of 2 hospitals and 14 outpatient neurology clinic sites and an adult epilepsy center with 2 hospitals and 27 outpatient neurology clinic sites. The models used neurology visit notes, EEG and MRI reports, visit patterns, hospitalizations, and medication, laboratory, and procedure orders to identify candidates for surgery. The models were trained on historical data up to May 10, 2019. Patients with an ICD-10 diagnosis of epilepsy who visited from May 11, 2019, to May 10, 2020, were screened by the algorithm and assigned surgical candidacy scores. The primary outcome was area under the curve (AUC), which was calculated by comparing scores from patients who underwent epilepsy surgery before November 10, 2020, against scores from nonsurgical patients. Nonsurgical patients' charts were reviewed to determine whether patients with high scores were more likely to be missed surgical candidates. Delay to surgery was defined as the time between the first visit that a surgical candidate was identified by the algorithm and the date of the surgery. RESULTS: A total of 5,285 pediatric and 5,782 adult patients were included to train the ML algorithms. During the study period, 41 children and 23 adults underwent resective epilepsy surgery. In the pediatric cohort, AUC was 0.91 (95% CI 0.87-0.94), positive predictive value (PPV) was 0.08 (0.05-0.10), and negative predictive value (NPV) was 1.00 (0.99-1.00). In the adult cohort, AUC was 0.91 (0.86-0.97), PPV was 0.07 (0.04-0.11), and NPV was 1.00 (0.99-1.00). The models first identified patients at a median of 2.1 years (interquartile range [IQR]: 1.2-4.9 years, maximum: 11.1 years) before their surgery and 1.3 years (IQR: 0.3-4.0 years, maximum: 10.1 years) before their presurgical evaluations. DISCUSSION: ML algorithms can identify surgical candidates earlier in the disease course. Even at specialized epilepsy centers, there is room to shorten the time to surgery. CLASSIFICATION OF EVIDENCE: This study provides Class II evidence that a machine learning algorithm can accurately distinguish patients with epilepsy who require resective surgery from those who do not.

Early identification of epilepsy surgery candidates: A multicenter, machine learning study.

OBJECTIVES: Epilepsy surgery is underutilized. Automating the identification of potential surgical candidates may facilitate earlier intervention. Our objective was to develop site-specific machine learning (ML) algorithms to identify candidates before they undergo surgery. MATERIALS & METHODS: In this multicenter, retrospective, longitudinal cohort study, ML algorithms were trained on n-grams extracted from free-text neurology notes, EEG and MRI reports, visit codes, medications, procedures, laboratories, and demographic information. Site-specific algorithms were developed at two epilepsy centers: one pediatric and one adult. Cases were defined as patients who underwent resective epilepsy surgery, and controls were patients with epilepsy with no history of surgery. The output of the ML algorithms was the estimated likelihood of candidacy for resective epilepsy surgery. Model performance was assessed using 10-fold cross-validation. RESULTS: There were 5880 children (n = 137 had surgery [2.3%]) and 7604 adults with epilepsy (n = 56 had surgery [0.7%]) included in the study. Pediatric surgical patients could be identified 2.0 years (range: 0-8.6 years) before beginning their presurgical evaluation with AUC =0.76 (95% CI: 0.70-0.82) and PR-AUC =0.13 (95% CI: 0.07-0.18). Adult surgical patients could be identified 1.0 year (range: 0-5.4 years) before beginning their presurgical evaluation with AUC =0.85 (95% CI: 0.78-0.93) and PR-AUC =0.31 (95% CI: 0.14-0.48). By the time patients began their presurgical evaluation, the ML algorithms identified pediatric and adult surgical patients with AUC =0.93 and 0.95, respectively. The mean squared error of the predicted probability of surgical candidacy (Brier scores) was 0.018 in pediatrics and 0.006 in adults. CONCLUSIONS: Site-specific machine learning algorithms can identify candidates for epilepsy surgery early in the disease course in diverse practice settings.

Neurology steering board effects change for a major electronic health record vendor.

Neurologists are among the least satisfied physicians with their current electronic health record (EHR), with many known pain points and great opportunities for improved tools and workflows. Improved EHR functionality can have major implications for patient care, physician efficiency, and prevention of burnout. We describe the advocacy of the American Academy of Neurology for improved EHR usability and the resultant formation and subsequent accomplishments of a Neurology Subspecialty Steering Board at 1 major EHR vendor (Epic).

Stimulus-Induced Rhythmic, Periodic, or Ictal Discharges in Coma-Incidence and Interrater Reliability of Continuous EEG After a Standard Stimulation Protocol: A Prospective Study.

PURPOSE: Stimulus-induced rhythmic, periodic, or ictal discharges (SIRPIDs) are often seen during continuous electroencephalographic (cEEG) monitoring in coma. Given their uncertain clinical significance, our prospective study evaluated incidence of SIRPIDs in comatose patients in the neuroscience intensive care unit (NSICU) who underwent a standard stimulation protocol and defined interreader reliability for cEEG. METHODS: Of 146 patients prospectively screened who underwent cEEG during a 6-month period, 53 patients were included and 93 patients were excluded. Our protocol used a sequence of auditory, mild tactile, and painful stimuli tested in a quiet room. Continuous electroencephalogram were then reviewed offline by blinded experts, with interrater agreement assessed by kappa statistic. By Pearson χ and Wilcoxon rank-sum tests, we then compared binary and numerical clinical features between those with and without SIRPIDs. RESULTS: Of 53 patients who underwent our protocol, one patient with a corrupt cEEG file was excluded. Traumatic brain injury was the most common diagnosis. Moderate interrater agreement was observed for 66 total stimulations: 20 patients (38.5%) had possible or definite SIRPIDs by minimum one reviewer. For 19 stimulations reviewed by a third reviewer, consensus was reached in 10 cases making the incidence of SIRPIDs 19.3% in our cohort. There was a moderate interrate agreement with kappa of 0.5 (95% confidence interval: 0.1, 0.7). Median intensive care unit stay was 15 days in patients with SIRPIDs versus 6.5 days in those without (P = 0.021). CONCLUSIONS: Our prospective study of SIRPIDs in the neuroscience intensive care unit found a 19% incidence by cEEG using a standard stimulation protocol, most often rhythmic delta activity, and showed a moderate interrater agreement.

Spreading depression in continuous electroencephalography of brain trauma.

OBJECTIVE: Cortical spreading depolarizations are a pathophysiological mechanism and candidate target for advanced monitoring in acute brain injury. Here we investigated manifestations of spreading depolarization in continuous electroencephalography (EEG) as a broadly applicable, noninvasive method for neuromonitoring. METHODS: Eighteen patients requiring surgical treatment of traumatic brain injury were monitored by invasive electrocorticography (ECoG; subdural electrodes) and noninvasive scalp EEG during intensive care. Spreading depolarizations were first identified in subdural recordings, and EEG was then examined visually and quantitatively to identify correlates. RESULTS: A total of 455 spreading depolarizations occurred during 65.9 days of simultaneous ECoG/EEG monitoring. For 179 of 455 events (39%), depolarizations caused temporally isolated, transient depressions of spontaneous EEG amplitudes to 57% (median) of baseline power. Depressions lasted 21 minutes (median) and occurred as suppressions of high-amplitude delta activity present as a baseline pattern in the injured hemisphere. For 62 of 179 (35%) events, isolated depressions showed a clear spread of depression between EEG channels with delays of 17 minutes (median), sometimes spanning the entire hemisphere. A further 188 of 455 (41%) depolarizations were associated with continuous EEG depression that lasted hours to days due to ongoing depolarizations. Depolarizations were also evidenced in EEG as shifts in direct current potentials. INTERPRETATION: Leão's spreading depression can be observed in clinically standard, continuous scalp EEG, and underlying depolarizations can spread widely across the injured cerebral hemisphere. These results open the possibility of monitoring noninvasively a neuronal pathophysiological mechanism in a wide range of disorders including ischemic stroke, subarachnoid hemorrhage, and brain trauma, and suggest a novel application for continuous EEG.

The incidence of seizures in patients undergoing therapeutic hypothermia after resuscitation from cardiac arrest.

STUDY OBJECTIVE: Non-convulsive seizures/status epilepticus occur in approximately 20% of comatose, non-cardiac arrest intensive care unit (ICU) patients, and are associated with increased mortality. The prevalence and clinical significance of seizures in comatose survivors of cardiac arrest undergoing therapeutic hypothermia is not well described. METHODS: At this urban level I trauma center, every patient undergoing therapeutic hypothermia is monitored with continuous video encephalography (cvEEG). We abstracted medical records for all cardiac arrest patients treated with therapeutic hypothermia during 2010. Clinical data were extracted in duplicate. cvEEGs were independently reviewed for seizures by two board-certified epileptologists. RESULTS: There were 33 patients treated with therapeutic hypothermia after cardiac arrest in 2010 who met inclusion criteria for this study. Median age was 58 (range 28-86 years), 63% were white, 55% were male, and 9% had a history of seizures or epilepsy. During cooling, seizures occurred in 5/33 patients (15%, 95%CI 6%-33%). 11/33 patients (33%, 95% CI 19%-52%) had seizures at some time during hospitalization. 13/33 (39%) survived to discharge and of these, 7/13 (54%) survived to 30 days. 9/11 patients with seizures died during hospitalization, compared with 11/22 patients without seizures (82% vs. 50%; difference 32%, CI 951%-63%). No patient with seizures was alive at 30 days. CONCLUSIONS: Seizures are common in comatose patients treated with therapeutic hypothermia after cardiac arrest. All patients with seizures were deceased within 30 days of discharge. Routine use of EEG monitoring could assist in early detection of seizures in this patient population, providing an opportunity for intervention to potentially improve outcomes.

The Tasmanian Epilepsy Register--a community-based cohort. Background and methodology for patient recruitment from the Australian national prescription database.

BACKGROUND/AIMS: Centralized prescription databases may provide an efficient mechanism for recruitment of community-treated disease. METHODS: The Australian federal government agency, the Health Insurance Commission (HIC), invited patients to participate in the Tasmanian Epilepsy Register (TER). Eligible patients included those who received at least one anticonvulsant above a 'reportable' price threshold between July 1, 2001 and June 30, 2002. Patients were asked to disclose their medical indication for anticonvulsant treatment with additional demographic and prescription information obtained from the HIC. RESULTS: 7,541 were eligible for recruitment. After two mail invitations over 6 months, 3,375 (46.6%) had responded, but TER enrollment amongst those indicating treatment for epilepsy was 1,180 (78.3%). TER participants were more likely to obtain their prescriptions exclusively from their general practitioner (70.9%) or from combined sources (19.1%) rather than from pediatrician (4.2%), neurologist (1.4%) or general physician (1.0%) sources. Patients were more likely to respond with increasing age (linear trend p < 0.001), when from a higher socioeconomic area (linear trend p < 0.001), or if their prescription was obtained from a neurologist (p < 0.001). CONCLUSION: The national Australian prescription database represents community-treated epilepsy and provides an effective and efficient method for patient recruitment for clinical epidemiological research.

Racial/ethnic disparities in the treatment of epilepsy: what do we know? What do we need to know?

We examine current understanding of the minority disadvantage in the clinical management of epilepsy. We performed an online literature search using several keywords (race, ethnicity, epilepsy, treatment, and quality of life) and identified additional literature through cross-referencing/manual search. The search produced 58 items published between 1977 and 2005. Of 49 original research studies, 38 were quantitative, 7 were qualitative, and 4 used mixed methods. Three or more articles were published in Epilepsia, Epilepsy &Behavior, Epilepsy Research, Neurology, and Seizure. Research concerning racial/ethnic differences in epilepsy treatment is scarce and limited by methodology, but suggests underutilization of state-of-the-art therapies by minorities. Racial/ethnic minorities also appear to have limited knowledge about epilepsy and its treatment, experience barriers to care, lack social support, and seek alternative therapies for epilepsy. We propose a framework to identify the array of disparities, points of intervention, and interventions.

Human blood genomics: distinct profiles for gender, age and neurofibromatosis type 1.

Application of gene expression profiling to human diseases will be limited by availability of tissue samples. It was postulated that germline genetic defects affect blood cells to produce unique expression patterns. This hypothesis was addressed by using a test neurological disease-neurofibromatosis type 1 (NF1), an autosomal dominant genetic disease caused by mutations of the NF1 gene at chromosome 17q11.2. Oligonucleotide arrays were used to survey the blood gene expression pattern of 12 NF1 patients compared to 96 controls. A group of genes related to tissue remodeling, bone development and tumor suppression were down-regulated in NF1 blood samples. In addition, there were blood genomic patterns for gender and age: Y chromosome genes showing higher expression in males, indicating a gene-dosage effect; and genes related to lymphocyte functions showing higher expression in children. The results suggest that genetic mutations can be manifested at the transcriptional level in peripheral blood cells and blood gene expression profiling may be useful for studying phenotypic differences of human genetic diseases and possibly providing diagnostic and prognostic markers.

Assessment of adverse events and quality of life in epilepsy: design of a new community-based trial.

Improving health-related quality of life (HRQOL) has become recognized as an essential component of treating patients with epilepsy. In recent years, several rating scales have been developed that focus on both common adverse effects and various aspects of HRQOL that are more relevant to this patient population. Increasingly, such assessments are being incorporated into clinical trials, as it becomes clear that improvements in overall quality of life are an important feature of drug therapy. Here we present the design of a large, community-based trial evaluating the effects of switching from immediate-release carbamazepine to twice-daily, beaded, extended-release carbamazepine (Carbatrol). As this trial involves switching formulations of the same compound, we expect to find only small differences in efficacy but significant differences in tolerability and quality-of-life measures. To identify appropriate instruments that could measure these factors, here we review several epilepsy-specific scales used to monitor adverse events and HRQOL and discuss their potential utility in the context of the proposed trial.

Psychopathology and quality of life: psychogenic non-epileptic seizures versus epilepsy.

BACKGROUND: HRQOL is lower in patients with psychogenic non-epileptic seizures (PNES) than in epilepsy patients. Although psychopathology may reduce HRQOL, it is not known whether patients with PNES and epilepsy are similarly affected. We aimed to compare the relationship between psychopathology and HRQOL in PNES and treatment resistant epilepsy. MATERIAL/METHODS: 106 patients with definite diagnosis of PNES or epilepsy were recruited from Epilepsy Monitoring Unit. Patients completed QOLIE-89, Profile of Mood States (POMS), and Adverse Events Profile (AEP). Total Mood Disturbance (TMD) was derived from POMS. We used chi-square and t tests and hierarchical multiple regression to compare HRQOL and its mental status correlates in patients with PNES and epilepsy. RESULTS: Psychiatric history was more prevalent and depression/dejection and TMD were higher in PNES than epilepsy (P<=0.003). PNES patients had a lower adjusted mean HRQOL than epilepsy patients (P<0.01). Mood problems and AEP were strong predictors of HRQOL (P<0.001) and explained the lower HRQOL in PNES vis-l-vis epilepsy. Decreases in HRQOL due to mood problems were similar in both groups. The model explained 62% of the variation in HRQOL. CONCLUSIONS: Although more severe psychopathology in PNES explains the lower HRQOL in PNES relative to epilepsy, the negative association between psychopathology and HRQOL remains stable across the groups. PNES patients with severe mood problems show similar, low levels of HRQOL as patients with severe mood problems who have epilepsy. Future studies should examine causal linkages between psychopathology and PNES and other explanations in seizure-related QOL.

Quality of life in psychogenic nonepileptic seizures.

PURPOSE: Psychogenic nonepileptic seizures (PNESs) are events that alter or seem to alter the neurologic function and, in their appearance, resemble epileptic seizures (ESs). In patients with ESs the psychological and medical aspects of epilepsy greatly influence the health-related quality of life (HRQOL). The relation between these factors and PNESs is not well established. In this study, we compared HRQOL in patients with PNESs with that of patients with ESs. METHODS: We evaluated 105 patients admitted to the Epilepsy Monitoring Unit of University Hospital between January 20, 2001, and January 20, 2002. Only patients with the definite diagnosis of ESs or PNESs were analyzed (n = 85). Patients completed an epilepsy-specific quality-of-life instrument (QOLIE-89), the Profile of Mood States (POMS), and Adverse Events Profile (AEP). We used t tests and regression analyses to contrast HRQOL in PNESs and ESs and to elucidate the main factors associated with HRQOL in patients with PNESs. RESULTS: In our sample, 45 patients had PNESs, and 40 had ESs. The overall HRQOL and scores on 13 of 19 QOLIE-89 subscales were significantly lower (i.e., worse) in PNES than in ES patients. AEP and scores on five of six POMS subscales also were worse in PNES patients than in ES patients. PNES versus ES diagnosis, POMS depression/dejection, and AEP were significant predictors of HRQOL, jointly explaining 65% variation in HRQOL. The lower HRQOL in PNESs versus ESs was in part explained by depression and AEP. CONCLUSIONS: Patients with PNESs have a lower HRQOL and worse mood problems than do patients with ESs. This disadvantage is primarily due to depression and medication side effects, although these factors influence QOL in much the same way in PNES and ES patients. These baseline HRQOL data on patients with PNESs can be used to evaluate the effects of treatment in this patient population.

Levetiracetam: a different approach to the pharmacotherapy of epilepsy.

OBJECTIVE: To review the pharmacology, pharmacokinetics, efficacy, and adverse effects of levetiracetam in the treatment of epilepsy. DATA SOURCES: A MEDLINE search restricted to English-language publications was conducted (January 1993-October 2000). Unpublished data provided by the manufacturer and information found in proceedings of professional meetings were also included. DATA EXTRACTION/STUDY SELECTION: Information regarding basic pharmacology was collected from studies in animals. Pharmacokinetic data were collected from human trials. Only randomized, placebo-controlled clinical trials were included to describe the efficacy and safety of levetiracetam. DATA SYNTHESIS: Levetiracetam is a new antiepileptic drug (AED) that appears to work by a unique mechanism. Animal studies have shown that levetiracetam may prevent epileptogenesis. Levetiracetam is rapidly and completely absorbed, minimally bound to plasma proteins, eliminated through the kidneys, and has a half-life of 6-8 hours. Doses must be adjusted for varying degrees of renal function. In clinical trials, levetiracetam significantly decreased seizure frequency compared with placebo when added to existing AED regimens. One clinical trial indicated that levetiracetam may be effective as monotherapy. Few major adverse effects were reported in the clinical trials; however, several patients reported psychological and psychotic reactions. CONCLUSIONS: Levetiracetam is a safe and effective new AED. Its apparent unique mechanism of action makes levetiracetam an important addition to therapy with older medications. Caution should be exercised when administering levetiracetam to individuals who may be prone to psychotic or psychiatric reactions.