Investigators at the Division of Neurology and Clinical Neurophysiology, Hospital General de Catalunya, Barcelona, Spain, evaluated the withdrawal of antiepileptic treatment over a one year period in 608 epileptics who had been seizure-free for 5 years. Relapses occurred in 144 (23.7%) of the total group. Among 474 patients in whom details of seizure types were available, relapses occurred in 119 (25.1%) of the total. Idiopathic generalized epilepsies showed a 19.6% relapse reate whereas symptomatic generalized epilepsies relapsed in 37.9% of patients. The lowest relapse rate occurred in patients with absence seizures (9.9%) and the highest rates in those with West (50%) and Lennox-Gastaut (43.7%) syndromes. Grand mal and complex partial seizures recurred in 27.8% and 23.3% of patients, respectively. The majority of relapses were single isolated seizures and occurred in the first year after drug withdrawal.
Risk factors for seizure relapse included 1) delay in initiating anticonvulsant therapy, 2) a symptomatic etiology, 3) mixed seizure types, 4) status epilepticus occurrence, and 5) signs of permanent neurologic damage. 
COMMENT: The 5 year seizure-free period before drug withdrawal in this study is longer and approximately twice that employed by other investigators. Shinnar S et al  reported the same 25% relapse rate among 88children who had been seizure free for only 2 years.
The withdrawal of antiepileptic medication should not be determined by a set seizure-free time-period for all types of epilepsy. Each patient should be evaluated individually, having regard to several factors, including psychological, predictive of potential relapse. Withdrawal is probably contra-indicated or likely to be unsuccessful in patients with the following: 1) symptomatic epilepsies with radiological or neurological evidence of structural cerebral defect, mixed seizure types, Lennox-Gastaut syndrome, focal seizures, complex partial with secondary generalized tonic-clonic patterns, 3) history of status epilepticus and 4) psychological stress, especially in adolescents and young adults.
Th prognostic value of the EEG is controversial; these authors and Thurston JH et al  found a negative correlation whereas others regard the EEG as an important factor, predictive of a good outcome when normal (Shinnar S et al 1985). An abnormal EEG was one of the most significant predictors of relapse after drug withdrawal in the study by Emerson R et al.