Outcome of Absence Epilepsy

Abstract

Article abstrac+We performed a meta-analysis of studies on absence epilepsy (AE) to ascertain whether the outcome of this well-defined type of epilepsy can be stated unequivocally. Using the Zndex Medicus and Medline CD+, we identified 1,619 publications. After applying the criteria of the International League against Epilepsy (ILAE) for AE, 26 publications on 23 study cohorts with a total of 2,303 patients were included. Remission rates ranged from 0.21 to 0.89; they differed substantially due to heterogeneity between the studies in inclusion criteria, methods, follow-up length, and outcome definitions. One half of the patients developed generalized tonic-clonic seizures (TCS) in the course of the disease. The proportion seizure free was 0.78 for patients with absence seizures (AS) only, and 0.35 for those who developed TCS. The outcome of AE may be worse than previously stated due to the considerable proportion of patients developing TCS in the course of their disease. Early prediction of outcome in patients who present with AS cannot be provided with certainty. Outcome of epilepsy may be expressed as remission of seizures, with or without medication; as evolution from one seizure type to another; or as intellectual, social, or vocational outcome. A number of patients become seizure free during treatment with antiepi-leptic drugs (AEDs) and remain so after their discon-tinuation. Berg and Shinnar' recently performed a meta-analysis on discontinuation studies and found an overall risk of relapse after 2 years of 0.29. Pre-dictors of a poor outcome were adolescent or adult age a t onset of seizures, remote symptomatic epilepsy , and a n abnormal interictal EEG. Although these results may help to estimate prognosis, a reliable early prediction of outcome in an individual patient is still not feasible. Moreover, accurate prediction of outcome is most urgent at the time of diagnosis, and not after 2 or more years of AED treatment. Outcome studies differ in study populations, designs , seizure types, and diagnostic criteria. This heterogeneity between studies limits the identification of prognostic factors. The predictors of outcome found by Berg and Shinnar' are characteristics of the various epilepsy types according to the 1989 classification of the International League against Epilepsy (ILAE),2 suggesting that epilepsy type is the main predictor of outcome. Estimates of outcome might therefore be determined more reliably for a single epilepsy type. To test this hypothesis, we conducted a meta-analysis of studies on the outcome of absence epilepsy (AE), because this epilepsy syndrome is well defined in age at onset and seizure and EEG characteristics. If outcome and its predictors prove to be consistent in AE, this will have important implications for the individual patient with this syndrome. Methods. Studies were identified using Index Medicus and Medline CD+ (from 1966 to 1994, inclusive). The search strategy consisted of the terms absence epilepsy, or petit mal, or pyknolepsy, and prognosis, or remission, or outcome (as MESH headings and text words). Reviews and preliminary reports were used to search for further references , mainly concerning studies published before 1966. References in English, German, French, Spanish, Italian, and Dutch were included. Criteria for inclusion of studies were that the reports should give clear descriptions of methods and results and that definition of absence epilepsy used in the study had been given. The outcome had to be expressed as remission or relapse rates as main study endpoint. Patient selection criteria had to be clearly stated. Studies on absence status o r myoclonic, atonic or atypical absences, or mixed types alone were excluded. The criteria for "typical" absence epilepsy (AE) were taken from the ILAE definition,2 adapted by Loiseau3 and W0lf3,~ (table 1). Outcome was expressed as the mean estimates of remission rate (i.e., the proportion of patients with AE who were free of seizures at the end of the study), with or without AEDs. Proportions were calculated if not presented. Confidence intervals (95% confidence limits) were calculated for remission rates. Survival analysis was used, when possible, to express estimates of remission rates relative to duration of follow-up. Sample size, age, gender, overall outcome, duration of follow-up, EEG findings , medication, and evolution to other seizure types and their outcome were ascertained and expressed as the mean estimates of the single studies when possible. When groups of patients were identified specifically according to differences in disease characteristics, the groups From the Departments of Neurology (Drs. Bouma, van Dijk, and Brouwer) and Clinical Epidemiology (Dr.