Review-of-systems questionnaire as a predictive tool for psychogenic nonepileptic seizures

Abstract

Patients with refractory epilepsy undergo video electroencephalography for seizure characterization; among whom approximately 10–30% will be discharged with the diagnosis of psychogenic non-epileptic seizures (PNES). Clinical PNES predictors have been described, but in general are not sensitive or specific. We evaluated whether multiple complaints in routine review of system (ROS) questionnaire could serve as a sensitive and specific marker of PNES. We performed a retrospective analysis of standardized ROS questionnaire completed by patients with definite PNES and epileptic seizures (ES) diagnosed in our adult epilepsy monitoring unit. A multivariate analysis of covariance (MANCOVA) was used to determine whether PNES and ES groups differed with respect to the percentage of complaints on ROS. Ten-fold cross-validation was used to evaluate the predictive error of a logistic regression classifier for PNES status based on percentage of positive complaints on ROS. A total of 44 patients were included for analysis. Patients with PNES had a significantly higher number of complaints on the ROS questionnaire compared to patients with epilepsy. A threshold of 17% positive complaints achieved a 78% specificity and 85% sensitivity for discriminating between PNES and ES. We conclude that routine ROS questionnaire may be a sensitive and specific predictive tool for discriminating between PNES and ES.

Diagnostic implications of review-of-systems questionnaires to differentiate epileptic seizures from psychogenic seizures

“…In addition to our larger sample, these results suggest that patients with mixed PNES and ES appear may be similar to patients with PNES only. All patients with PNES therefore may have increased somatic sensitivity, somatic dysfunction, or increased disability or distress caused by somatic symptoms [13, 17, 19]. This increased somatic sensitivity has been used to explain why patients with PNES tend to be diagnosed with more medical conditions [18, 20–22].…”

“…All predictive models were trained using the retrospective dataset alone and performance was assessed on the prospective dataset. Predictive performance was compared to the naïve assumption that all patients had ES, to the performance of the 17% cut-off proposed by Robles and colleagues, and to a predictive approach similar to what was done by Asadi-Pooya and colleagues [13, 14]. Comparison to a naïve assumption is more conservative than a comparison to a random classifier with accuracy of 50% because it takes into account the prevalence difference between PNES and ES.…”

“…Almost all clinics ask patients to fill out standardized review-of-systems questionnaires (RoSQ) before they speak to a physician; evidence that these are effective in diagnosis of seizures is promising but limited [13, 14]. Patients with PNES experience increased somatic sensitivity, as evidenced by medically unexplained symptoms and reporting more disability for less severe symptoms [1, 12, 13, 15, 16]. Additionally, most frequently, PNES are a component of conversion disorder in which patients convert psychological stressors into somatic symptoms or findings, one of which can be seizures [3, 12, 17].…”

“…There is retrospective evidence that RoSQs may help identify patients at risk for PNES [13, 14]. In a small retrospective dataset with many different formats of RoSQs, Robles and colleagues demonstrated recently that patients who noted more than 17% of symptoms on RoSQs were more likely to have PNES than ES with an area under the received operating curve (AUC) of 84% [13].…”

“…In a small retrospective dataset with many different formats of RoSQs, Robles and colleagues demonstrated recently that patients who noted more than 17% of symptoms on RoSQs were more likely to have PNES than ES with an area under the received operating curve (AUC) of 84% [13]. Their sample size was limited, however, by inconsistent availability of RoSQs in the electronic health record.…”