Can Semiology Predict Psychogenic Nonepileptic Seizures? A Prospective Study

Objective: Reducing health and economic burdens from diagnostic delay of psychogenic nonepileptic seizures (PNES) requires prompt referral for video electroencephalography (VEEG) monitoring, the diagnostic gold standard. Practitioners make VEEG referrals when semiology suggests PNES, although few semiological signs are supported by well-designed studies, and most VEEG studies neglect to concurrently measure how accurately seizure witnesses can ascertain semiology. In this study, we estimate the value of eyewitness-reported and video-documented semiology for predicting PNES, and we measure accuracy of eyewitness reports. Methods: We prospectively interviewed eyewitnesses of seizures in patients referred for VEEG monitoring, to inquire about 48 putative PNES and ES signs. Multiple, EEG-blinded, epileptologists independently evaluated seizure videos and documented the presence/absence of signs. We used generalized estimating equations to identify reliable video-documented PNES and ES signs, and we compared eyewitness reports with video findings to assess how accurately signs are reported. We used logistic regression to determine whether eyewitness reports could predict VEEG-ascertained seizure type. Results: We analyzed 120 seizures (36 PNES, 84 ES) from 35 consecutive subjects. Of 45 video-documented signs, only 3 PNES signs (preserved awareness,'' eye flutter,'' and bystanders can intensify or alleviate'') and 3 ES signs (abrupt onset,'' eye-opening/widening,'' and postictal confusion/sleep'') were significant and reliable indicators of seizure type. Eyewitness reports of these 6 signs were inaccurate and not statistically different from guessing. Consequentially, eyewitness reports of signs did not predict VEEG-ascertained diagnosis. We validated our findings in a second, prospective cohort of 36 consecutive subjects. Interpretation: We identified 6 semiological signs that reliably distinguish PNES and ES, and found that eyewitness reports of these signs are unreliable. We offer suggestions to improve the accuracy of eyewitness reports. ANN NEUROL 2011;69:997-1004