Abstracts

Epilepsy management typically relies on patient report of seizure frequency and temporal pattern. To evaluate their accuracy, we compared these reports with the patterns observed in a one-year, prospective, clinic-based seizure diary study. Subjects [gt]= 18 years with localization-related epilepsy and [gt]=1 seizure during the prior year were recruited from the Epilepsy Mgmt Center/Neurology clinic at MMC. A comprehensive interview at recruitment included a full epilepsy history, with detailed questions regarding typical seizure frequency per week and month prior to the intake interview, and typical seizure pattern. Subjects were asked to maintain a yearlong daily seizure diary, including dates and times of all seizures. Patients who completed [gt]=30 days of the diary were eligible for analysis. We compared the initial self-report of patients to the daily reports turned in by the patients. Seizure clustering by self-report was defined as those subjects who initially reported that they typically experienced 3 sz in 24 hrs. Actual seizure clustering from diary data was defined by both this definition (clinical definition), and by a statistical analysis of the pattern of seizures for each subject in which clustering was defined as greater than expected variability from a Poisson process model that assumed no clustering (Boots [amp] Getis, Point Pattern Analysis, 1988) (statistical definition). Of 55 subjects who completed the study, 9 subjects had no seizures. In comparison with diary-based frequency estimates, correlations of self-reports of average weekly ([rho]=0.63) and monthly ([rho]=0.62) seizure frequency were high. Using the diary-based clinical definition of clustering as the gold standard, self-reported clustering had a sensitivity of 62%, specificity of 79%, PPV of 73% and NPV of 70%. Using the diary-based statistical definition of clustering as the gold standard, self-reported cluster had moderate sensitivity (67%), high specificity (88%), a good positive predictive value (PPV, 62%) and an excellent negative predictive value (NPV, 90%). Patient self-report provides relatively accurate estimates of seizure frequency and clustering and provides a sound basis for initial management. Some of the observed differences between self-reported and recorded seizures may reflect changes in therapy during the observational phase, and correlations may actually be higher. Though weekly and monthly estimates of seizure number by patients correlate well with diary-based estimates of seizure rates, diaries provide more accurate information that could improve patient management. In those who report lack of clustering, diaries rarely reveal it (as reflected by the high NPVs). Those who report clustering are likely to have it (as shown by the high PPVs) though diary confirmation is recommended. (Supported by K23NSO2192 (PI Dr. Haut))