Predictive factors for success in surgical treatment for partial epilepsy: a multivariate analysis.

Abstract

We conducted a longitudinal self-controlled study of 131 patients aged 4-60 treated with resective surgery for medically uncontrolled partial epilepsy from 1949 to 1988. Using multivariate logistic regression, we showed that pre- and perioperative variables can be used to predict "success" or "failure" of surgical resective treatment in approximately 79% of cases. If the predicted probability is > 0.75 or < 0.25, the model predicts a correct result in 87% of cases. Eight predictive factors emerged with a backward multivariate logistic regression model with the likelihood-ratio (LR) test to exclude variables from the equation: (a) the influence of the surgical team and surgical procedure, (b) the presence of paresis preoperatively, (c) duration of disease, (d) age at treatment, (e) positive neuroradiologic findings in preoperative investigations, (f) preoperative complex partial seizures (CPS), (g) nonepileptic EEG abnormalities, and (h) generalized spike activity in EEG preoperatively. Sex, age at first seizure, area of resection, presence of simple or generalized seizures preoperatively, preoperative seizure frequency, tissue pathology, use of computed tomography/nuclear magnetic resonance (CT/NMR) in preoperative investigations, degree of preoperative neurologic deficit, perioperative electrocorticographic results, and bilateral EEG spikes did not have predictive value in the model.