Medical-Surgical Epilepsy Unit, Department of Neurology, Clinical Epidemiology Unit, Neurophysiology Service, Department of Neurosurgery, Department of Neuroradiology, Department of Neuropsychology, Hospital de Cruces, Baracaldo, Spain
- Mots-clés : seizures, mesial temporal sclerosis, epilepsy surgery, prognosis, predictors, temporal lobe
- DOI : 10.1684/epd.2011.0413
- Page(s) : 36-46
- Année de parution : 2011
To investigate the outcome of temporal lobe epilepsy surgery and identify the variables which predict a good prognosis with respect to seizures in postoperative follow-up after two and four years. This retrospective study included 115 selected patients who underwent surgery for temporal lobe epilepsy between 1996 and 2007. In the second year after surgery 86.1% of patients had a good prognosis for seizure control (73.9% Engel class I and 12.2% Engel class II) and 89.2% (76.3% Engel class I and 12.9% Engel class II) in the fourth year. Sixty-four of 93 (68.8%) patients were free of disabling seizures (Engel class I) during the entire period and 78 (83.8%) had good prognosis (Engel class I and II). For the second year, logistic regression analysis revealed the following variables to be independently predictive of good seizure control: absence of two or more seizure episodes in the first year after surgery, normal postoperative video-EEG, and age at surgery of less than 35 years. In the fourth year, mesial temporal sclerosis, female sex and normal postoperative video-EEG were the predictive factors. For the group with a good prognosis in both the second and the fourth year, the predictive variables were: absence of two or more seizure episodes in the first year after surgery (OR: 13.762, CI 95%: 2.566-73.808, p<0.002) and normal postoperative video-EEG (OR: 16.301, CI 95%: 3.704-71.740, p<0.001). This study illustrates the sustained benefit of temporal lobe epilepsy surgery. The multivariate logistic regression analysis failed to identify a good predictive model composed of preoperative variables alone, although it was possible to build such a model with either pre- and postoperative variables or only postoperative variables.