Abstracts

RATIONALE: Technical advances have not erased the value of seizure description by clinical history. It is still one of the most important tools in the evaluation of patients with epilepsy. There has not been a formal comparison of the accuracy, sensitivity and specificity of seizure identification by history alone, versus the gold standard, seizure identification by long term video-EEG monitoring. We performed this analysis in patients with suspected temporal lobe epilepsy. At the end of this activity, participants will have an evidence based notion of the accuracy of epileptologitsts[ssquote] assessment of the clinical history to correctly identify seizures in patients with probable temporal lobe epilepsy.
METHODS: The clinical and EEG monitoring data of 88 patients with suspected temporal lobe epilepsy were reviewed. Detailed descriptions of the different types of clinical events reported by these patients were registered in a database. All events were also classified according to the International League Against Epilepsy seizure classification. The EEG telemetry monitoring data of all recorded, seizure-like events in these patients were extracted and classified as a seizure or not a seizure. Seizure origin and seizure type were noted. Each clinical event reported by history was matched with the corresponding clinical event recorded in the epilepsy unit (gold standard). Each clinical event was adjudicated by three blinded (no access to EEG data), independent epileptologists as being a seizure or not. Sensitivity, specificity, overall accuracy and interrater agreement for the clinical assessment were obtained.
RESULTS: Of 361 clinically different events described by 88 patients (average, 4 per patient), 175 (48.4%) were recorded in the epilepsy unit. Forty (23%) were simple partial seizures, 97 (55%) complex partial seizures, 28 (16%) primarily or secondarily generalized seizures and 10 (6%) non-epileptic events. Only 10 events (6%) were misidentified as a seizure or not by blinded epileptologists, resulting in an overall clinical accuracy of 94%. Epileptologists[ssquote] sensitivity for seizure identification was 96% (95% confidence intervaI [CI], 94-99%), but their specificity was only 50% (95% CI, 15-85%). Raw agreement among the 3 epileptologists was 90%. Agreement beyond chance was good, with a kappa of 0.7 for paired evaluations.
CONCLUSIONS: The overall clinical accuracy of epileptologists in identifying an event as a seizure or not by clinical semiology alone was high in patients with suspected temporal lobe epilepsy. Epileptologists rarely missed a seizure (high sensitivity,96%), but they often overcalled non-seizures as seizures (low specificity, 50%). Agreement among different epileptologists was good. Features that influence clinical accuracy, and implications for research and clinical practice are discussed.