Abstracts

Neurologists must frequently make decisions as to whether or not referral for EEG monitoring is likely to produce fruitful results such as whether or not the attacks recorded are likely to be epileptic or nonepiletic. Data to guide making such decisions are frequently lacking. 191 adults with reported spells underwent EEG monitoring for an average of 5.6 days. Of these, 163 demonstrated attacks which were either epileptic only (n=116) or nonepilelptic (NES; presumed psychogenic) only (n=47). The remaining cases were indeterminant or had both epilepsy and NES. Predictors of epileptic vs. nonepileptic attacks were as follows: 1) report to the social worker of physical, emotional, or sexual abuse from the patient[apos]s perspective during the developmental years (0=no report of abuse; 1=abuse reported in one of the three areas; 2=abuse reported in more than one area); 2) age at onset of repetitive spells (0=before age 21; 1=after age 20); 3) reported frequency of attacks (0=less than 1/d; 1=1/d or more; 4) psychiatric history (0=negative; 1=positive); and 5) gender (0=male; 1=female). A stepwise, buildup, linear regression model was employed. The first four predictors were entered by the computer into the prediction equation in the order described. History of abuse was the most potent predictor (R=.368), but this prediction was significantly improved with the stepwisse additions of early vs. late onset of attacks (R=508), frequency of spells (R=.596), and history of psychiatric problems (R=.619). Gender did not add significantly to this predictive equation. Of interest was the fact that the four predictors were not greatly different in predictive power. Also, surprisingly, the predictors were almost uncorrelated with one another (median intercorrelation = .06). Therefore, a predicitve system was devised by simply adding up the scores from the four predictors (range 0-5 for summary score). This summary score was associated with percentages of epileptic and nonepileptic cases, respectively, as follows: score of 0: 100%, 0%; score of 1: 94%, 6%; score of 2: 69%; 31%; score of 3: 35%, 65%; score of 4: 23%, 77%; score of 5: 0%, 100%. With scores of 2 and less identified as epilepsy and scores of 3 and more identified as characteristic of NES, 82% of patients were correctly classified overall (90% epilepsy; 64% NES). Not surprisingly, the prediction of whether epileptic or NES will be recorded at monitoring remains imperfect. Nevertheless, the predictive formula identified here is easy to use, and the necessary information can be obtained noninvasively at any clinic visit in very few minutes. As such, it should be of value in presaging the likely outcomes of EEG monitorings. The importance of the abuse variable is evident in this study involving NES patients, and it clearly requires additional attention in the future.