Novel combinations of phenotypic biomarkers predict development of epilepsy in the lithium-pilocarpine model of temporal lobe epilepsy in rats

The discovery and validation of biomarkers in neurological and neurodegenerative diseases is an important challenge for early diagnosis of disease and for the development of therapeutics. Epilepsy is often a consequence of brain insults such as traumatic brain injury or stroke, but as yet no biomarker exists to predict the development of epilepsy in patients at risk. Given the complexity of epilepsy, it is unlikely that a single biomarker is sufficient for this purpose, but a combinatorial approach may be needed to overcome the challenge of individual variability and disease heterogeneity. The goal of the present prospective study in the lithium-pilocarpine model of epilepsy in rats was to determine the discriminative utility of combinations of phenotypic biomarkers by examining their ability to predict epilepsy. For this purpose, we used a recent model refinement that allows comparing rats that will or will not develop spontaneous recurrent seizures (SRS) after pilocarpine-induced status epilepticus (SE). Potential biomarkers included in our study were seizure threshold and seizure severity in response to timed i.v. infusion of pentylenetetrazole (PTZ) and behavioral alterations determined by a battery of tests during the three weeks following SE. Three months after SE, video/EEG monitoring was used to determine which rats had developed SRS. To determine whether a biomarker or combination of biomarkers performed better than chance at predicting epilepsy after SE, derived data underwent receiver operating characteristic (ROC) curve analyses. When comparing rats with and without SRS and sham controls, the best intergroup discrimination was obtained by combining all measurements, resulting in a ROC area under curve (AUC) of 0.9592 (P<0.01), indicating an almost perfect discrimination or accuracy to predict development of SRS. These data indicate that a combinatorial biomarker approach may overcome the challenge of individual variability in the prediction of epilepsy.