Abstracts

Rationale: Most pharmacogenomic studies have attempted to identify single nucleotide polymorphism (SNP) markers that are predictive for treatment outcomes. However, it is unlikely in complex diseases such as epilepsy, affecting heterogeneous populations, that a single SNP will adequately explain treatment outcomes. This work reports a multi-SNP model that classifies treatments outcomes for such a disease, and compares this with single SNP models. Methods: A prospectively collected dataset of outcomes in 115 patients newly treated for epilepsy, with genotyping for 4,041 SNPs in 279 candidate genes, was utilized for the model development. Of these 115 patients, 83 were phenotyped as responders (seizure free for 12 months) and 32 as non-responders. Phenotyping was performed by a committee of two trained neurologists. A cross-validation based methodology on the developmental cohort identified SNPs most influential in predicting seizure control after one year of drug treatment and then incorporated these into a multi-SNP classification model. The classifier was cross-validated to determine it’s effectiveness in predicting treatment outcome in the developmental cohort and then in two independent validation cohorts. In each the classification by the multi-SNP model was compared with that of models utilizing the individual SNPs alone. Results: Five SNPs were selected for the multi-SNP model. Cross-validation showed that the multi-SNP model had a predictive accuracy of 83.5% in the developmental cohort, detecting 74 of 83 responders correctly and 22 out of 32 non-responders correctly. Using cross-validation within the developmental cohort the classifier had a sensitivity of 89% and specificity of 69%. These findings were supported by replication analyses of 63 subsequently recruited newly treated epilepsy patients and 108 chronic epilepsy patients. In the newly treated replication cohort the classifier predicted 42 of 46 responders correctly (sensitivity 91%) and 9 of 17 non-responders (specificity 53%). In the chronic Tasmanian epilepsy cohort the classifier predicted 63 of 78 responders (sensitivity 81%) and 15 of 30 non-responders (specificity 50%). In all cases the multi-SNP model classified the treatment outcomes better than those using any individual SNPs alone. Importance of significance with a low number of patients supports the clinical applicability of such a tool on a patient-by-patient basis. Conclusions: The results demonstrate that a multi-SNP classifier can successfully predict treatment outcome more reliably than single SNP models. It would be useful to apply this model to other populations of newly treated epilepsy for further validation. Future pharmacogenomic research in complex diseases such as epilepsy should attempt to develop classifiers that utilize multiple genetic determinants.