Abstracts

Rationale: Greater neurocognitive symptomatology may reflect more widespread brain dysfunction or higher stress levels, and therefore potentially a higher chance of seizure recurrence. We aimed to determine whether a patient-perceived neurocognitive symptoms questionnaire could predict seizure control following initiation of epilepsy treatment, and if so if this was independent to that of other clinical, imaging and genomic predictors. Methods: 138 newly treated epilepsy patients completed the A-B Neuropsychological Assessment Scale (ABNAS) prior to initiation of anti-epileptic drug (AED) therapy and were prospectively followed for 12 months. Patients were classified as medication non-responsive based on whether they had one or more recurrent seizures despite treatment, not explained by medication non-compliance or other significant provoking factors. The pre-treatment ABNAS scores and additional clinical covariates were compared between AED responsive and non-responsive patients. Subsequently our recently reported multigenic classifier was incorporated into a final logistic model to determine whether the neurocognitive questionnaire provided predictive value for seizure control independent of the genomic classifier. The ABNAS predictive value was subsequently validated in a separate prospective cohort of 74 newly treated epilepsy patients. Results: Non-responsive patients (n=45) had a significantly higher pre-treatment ABNAS score (i.e. worse neurocognitive symptomatology) than patients in whom seizures were controlled (n=93) (median, 10 vs. 5, p=0.005, Mann-Whitney U-test). The presence of a lesion on MRI was also associated with an increased chance of seizure recurrence (p=0.003). The χ² for the overall likelihood estimate of the multivariate logistic regression model with the inclusion of the multigenic classifier was 56.7 (p<0.0001), demonstrating that the pre-treatment ABNAS (p=0.01) along with the MRI focal lesions (p=0.02) were both independently predictive of patient response to AED. The diagnostic performance of the logistic model for AED pharmacoresponsiveness had sensitivity 91%; specificity 64%; positive predictive value 84% and negative predictive value 78%. The correlation between the questionnaire and pharmacoresponse remained significant in the subsequent validation cohort (median, 23 vs. 7, p=0.004, Mann-Whitney U-test). Conclusions: Pre-treatment assessment of neurocognitive symptomatology provides prognostic information regarding the chance of successful seizure control in patients with newly treated epilepsy. These results further demonstrate the multifactorial nature of the determinants of anti-epileptic drug response, indicating that genomic, structural (MRI) and neuropsychological factors all independently contribute to the complex drug response phenotype.