How Well Can We Predict Medical Intractability in Children with Epilepsy?
Abstract number :
1.157
Submission category :
Year :
2001
Submission ID :
3043
Source :
www.aesnet.org
Presentation date :
12/1/2001 12:00:00 AM
Published date :
Dec 1, 2001, 06:00 AM
Authors :
S.S. Ravid, MD, Child Neurology, Schneider Children[ssquote]s Hospital, New Hyde Park, NY; J. Maytal, MD, Child Neurology, Schneider Children[ssquote]s Hospital, New Hyde Park, NY; G.P. Novak, MD, Child Neurology, Schneider Children[ssquote]s Hospital, Ne
RATIONALE: To determine if predictors of intractability derived from multivariate analysis of one population of patients can be accurately applied to a different population.
METHODS: Two groups of patients from the same medical center were examined by retrospective chart review; group I, recruited from the practice of one author(GN), included 300 patients with at least 2 unprovoked seizures treated for more than 1 year. Group II included 144 patients meeting the same criteria, recruited from the neurology clinic and from the practice of another author (JM). Intractability was defined as more than 1 seizure per year. The presence of several putative risk factors for seizure intractability were examined.
RESULTS: Patients in Group II were more likely to be intractable (49% v 35%; odds ratio (OR)=1.9, p=.001) and to have EEG background abnormalities (40% v 29%, OR=1.6, p=.02); they were less likely to have prior status epilepticus (7% v 20%, OR=.30, p=.0004), EEG spikes (65% v 80%, OR=.45, p=.0004), and many seizures at 3 months into treatment (10% v 20%, OR=.47, p=.04). Univariate analysis (Table) revealed that several factors associated with intractablity in Group I were not associated in Group II (myoclonus, infantile spasms, EEG spikes,neuroimaging, and many seizures at 3 months). Multivariate analysis on Group I revealed 2 variables (etiology and seizures at 3 months) that correctly classified 81% of patients (sensitivity 69%, specificity 88%); applied to Group II, only 56% were correctly classified (sensitivity 26%, specificity 85%).
CONCLUSIONS: Some clinical variables which are associated with intractability in one population of children with epilepsy may not be associated with this outcome in other populations. This limits the ability to use multivariate models as valid predictors of outcome across different populations.[table]