The Accuracy of Outcome Prediction Models for Childhood‐onset Epilepsy

Abstract

Summary:Purpose: Two large prospective cohort studies of childhood epilepsy (Nova Scotia and the Netherlands) each developed a statistical model to predict long-term outcome. We sought to evaluate the accuracy of a prognostic model based on the two studies combined.Methods: Analyses with classification tree models and stepwise logistic regression produced predictive models for the combined dataset and the two separate cohorts. The resulting models were then externally validated on the opposite cohort. Remission was defined as no longer receiving daily medication for any length of time at the end of follow-up.Results: The combined cohorts yielded 1,055 evaluable patients. At the end of follow-up (≥5 years in >96%), 622 (59%) were in remission. By using the combined data, the classification tree model and the logistic regression model predicted the outcome correctly in ∼70%. The classification tree model split the data on epilepsy type and age at first seizure. Predictors in the logistic regression model were: seizure number before treatment, age at first seizure, absence seizures, epilepsy types of symptomatic generalized and symptomatic partial, preexisting neurologic signs, intelligence, and the combination of febrile seizures and cryptogenic partial epilepsy. When the prediction models from each cohort were cross-validated on the opposite cohort, the outcome was predicted slightly less accurately than did the model from the combined data.Conclusions: Based on currently available clinical and EEG variables, predicting the outcome of childhood epilepsy may be difficult and appears to be incorrect in about one of every three patients.

The Influence of Design and Definition on the Proportion of General Epilepsy Cohorts with Remission and Intractability

“…The proportion with remission off AED was not affected by the fact that patients with generalised absence seizures and known poor prognosis, including progressive neurological disorders, were excluded from the cohort with 54% in remission off medication [26] compared to 55% when the previously excluded patients were later included in the analysis [18] . Possibly, the two groups reintroduced into the cohort (good prognosis, i.e.…”

“…The study with 48% of the cohort in immediate remission at 2 years excluded Table 3. Proportion of patients in remission off AED arranged according to potential for bias 52% retrospective ͉ hospital ͉ off AED at last follow-up ͉ Oskoui et al [9], 2005 47% mixed ͉ population ͉ 5 years off AED at last follow-up ͉ Sillanpaa et al [25], 1998 1 54% mixed ͉ population ͉ off AED at last follow-up ͉ Camfield et al [26], 1993 55% mixed ͉ population ͉ off AED at last follow-up ͉ Geelhoed et al [18], 2005 59% mixed/prospective ͉ population/hospital ͉ off AED at last follow-up ͉ Geelhoed et al [18], 2005 65% prospective ͉ hospital ͉ off AED at last follow-up ͉ Geelhoed et al [18], 2005 56% prospective ͉ population ͉ 5 years off AED at last follow-up ͉ Sillanpaa et al [25], 1998 1 1 Physicians were reluctant to discontinue AED. patients with progressive neurological disorders associated with an increased risk of seizures, e.g.…”

“…Two cohorts were from low-and middle-income countries (Bangladesh and India), 2 from Hong Kong and 1 from Saudi Arabia. There was a cohort formed from mixing 2 earlier cohorts, which were also included in this review [18] . No cohort was based in Australasia or Africa.…”

The Need for Broad Spectrum and Safe Anti-epileptic Drugs in Childhood Epilepsy

“…8 However, the problem is that the delineation of the syndrome is not always possible at the onset of the seizures. Who can predict that a four-yearold normal child presenting with two tonic-clonic seizures will develop the devastating Lennox Gastaut syndrome?…”

Validation of Two Prognostic Models Predicting Outcome at Two Years after Diagnosis in a New Cohort of Children with Epilepsy: The Dutch Study of Epilepsy in Childhood

“…These prognostic models may not perform as well in other settings as in the setting in which they have been developed and should therefore be validated in a new cohort (10). Surprisingly, only a few of these models have been validated with external patient samples (8,12), although sometimes split‐sample approaches have been used (11). This lack of validation may be a result of the time it takes and the expense to recruit and follow up new patient populations.…”