We agree with Semah and colleagues that MRI data will be very important for any clinical prediction model designed to identify patients at high risk for refractory epilepsy. However, there are 2 caveats regarding neuroimaging data originating from tertiary care epilepsy centers. First, precise classification systems (such as one that distinguishes between HS and HA, as proposed by Semah and colleagues) will need to demonstrate reproducibility between institutions. The more esoteric the classification system, the more challenging this becomes. Second, prognostic or predictive models generated in tertiary care centers must be generalizable to more diverse patient populations. Patients referred to tertiary care centers may have more severe disease than patients not seen or studied in such centers. Using patients with the widest possible spectrum of disease severity is critical in the development of prediction models. The prevalence of HA in the general population, including patients without seizures and patients with well-controlled epilepsy, is unknown.1,2 We agree with Semah and colleagues that large prospective studies in children are needed to evaluate the prognostic value of MRI abnormalities in children with epilepsy. A major challenge will be designing population-based studies to address this important issue.
Dlugos DJ. In reply. Arch Neurol. 2002;59(6):1042-1043. doi: