Outcome predictors for surgical treatment of temporal lobe epilepsy with hippocampal sclerosis

Abstract

Purpose: To study long-term postoperative course and identify predictors for postoperative seizure control in patients with medically intractable temporal lobe epilepsy (TLE) associated with hippocampal sclerosis (HS), diagnosed by magnetic resonance imaging (MRI), and ascertained histopathologically. To compare patients becoming seizure-free (i.e., cured from epilepsy) and patients experiencing prolonged seizure-free periods interposed with recurring seizures.

Methods: One hundred thirty-five patients (74 women) underwent complete evaluation for epilepsy surgery. The predictive value of duration of epilepsy, age at onset, age at surgery, gender, febrile convulsion history, ictal dystonic posturing, unilateral interictal electroencephalography (EEG) discharges (IED), preoperative secondarily generalized tonic-clonic seizures (SGTCS), and preoperative seizure frequency for short- and long-term postoperative seizure control were evaluated with two classification systems: Classification 1 (seizure-freedom with or without auras during 12-months before observation points) and the stringent classification 2 [International League Against Epilepsy (ILAE) Ia; absolute absence of seizures and auras after operation].

Results: Unilateral IED at year 1 and 2 (p = 0.037 and p = 0.034), male gender and low seizure frequency at year 2 (p = 0.013 and p = 0.046) were significant predictors for seizure freedom using classification 1. All variables (except male gender at year 2; p = 0.035) lost their predictive power, applying classification 2. The proportion of seizure-free patients remained stable between 70% to 79% with classification 1, but decreased from 64.4% at year 1 to 45.8% at year 5 with classification 2.

Discussion: Positive predictors of short-term outcome do not predict long-term outcome in patients with TLE associated with HS. Absolute freedom of seizures and auras cannot be predicted by conventional preoperative variables.