To determine the etiology and clinical predictors of intractable epilepsy, a case-control study comprising 50 patients and 50 control subjects was performed. Patients included children who had more than one seizure per month over at least 6 months. Control subjects included children with epilepsy who had been seizure-free for more than 6 months. Patients were evaluated with special reference to birth history and development. Clinical examination and neurodevelopmental assessment were performed in all the patients. Drug monitoring was performed to exclude pseudointractability. Epilepsy in the study group was caused by perinatal problems (48%) and sequelae of central nervous system infection (24%) and was idiopathic in 20%. In the control group, epilepsy was idiopathic in 72%, a result of calcified granuloma in 22%, and perinatal problems comprised 6% of the subjects. On univariate analysis, strong association was evident between intractable epilepsy and several factors, including age at onset of seizure, remote symptomatic epilepsy, initial seizure type, history of neonatal seizure, high initial seizure frequency, microcephaly, and neurologic impairment. On multivariate analysis, neurologic impairment (odds ratio [OR] 12.25; 95% confidence interval [CI] 3.58-41.89), age at onset of seizure less than 1 year (OR 11.70; 95% CI 2.95-46.43), myoclonic seizure/infantile spasm (OR 10.36; 95% CI 2.39-44.93), and remote symptomatic epilepsy (OR 2.9; 95% CI 1.13-7.43), were independent predictors of intractability.