Methodological issues in estimating the cost of epilepsy