Prognostic index for recurrence of seizures after remission of epilepsy. Medical Research Council Antiepileptic Drug Withdrawal Study Group.British Medical Journal 1993; 306 doi: https://doi.org/10.1136/bmj.306.6889.1374 (Published 22 May 1993) Cite this as: British Medical Journal 1993;306:1374
OBJECTIVES--To develop and test a prognostic index for the recurrence of seizures after a minimum remission of seizures of two years in people with a history of epilepsy. DESIGN--Information from a large prospective randomised study of withdrawal of antiepileptic drugs was used to identify clinical and treatment factors of prognostic importance in determining the recurrence of seizures. A split sample approach was used to test the internal validity of predictions made on the basis of identified prognostic factors. SETTING--Centres in six European countries. MAIN OUTCOME MEASURES--Comparison of predicted and observed rates of recurrence of seizure. SUBJECTS--1013 patients randomised to the Medical Research Council study for antiepileptic drug withdrawal. RESULTS--The Cox proportional hazards model identified several factors that increased the risk of seizures recurring. These included being 16 years or older; taking more than one antiepileptic drug; experiencing seizures after starting antiepileptic drug treatment; a history of primary or secondarily generalised tonic-clonic seizures; a history of myoclonic seizures; and having an abnormal electroencephalogram. The risks of seizures recurring decreased with increasing time without seizures. The model allowed estimation of the risk of seizures recurring in the next one and two years under the policies of continued antiepileptic drug treatment and slow withdrawal of drugs. Split sample validation suggested that the model was well calibrated. CONCLUSION--The model is currently the best available aid for counselling the many patients in the community with epilepsy currently in remission who seek advice about the risks of seizures recurring if they stop antiepileptic drug treatment. The model requires validation in a broad population of patients, and such studies are in progress.