Abstracts

Rationale: Both FDG-PET and structural MRI (sMRI) can reveal abnormalities related to epilepsy in patients with medication resistant epilepsy. However, some of the characteristic findings in epilepsy are progressive therefore neuroimaging is less sensitive after soon after the patient begins to experience seizures. We seek to improve the cost-benefit ratio of clinical neuroimaging by providing a quantitative model to determine the pre-test probability that a potentially diagnostic abnormality will be observed on a patient s interictal FDG-PET or sMRI.Methods: We used Bayesian logistic regression to estimate the effect of age, gender, duration of seizure disorder, frequency of seizures and the etiology of their seizures on the interictal FDG-PET and sMRI. This model was based upon the expert analysis of neuroimages from 485 patients admitted to the UCLA video-EEG monitoring unit from 2006 to 2013. 135 patients with non-epileptic seizures were included as negative controls. Missing data was multiple imputed based on other input data using Markov chain Monte Carlo estimation.Results: Our model showed that for both FDG-PET and sMRI, the probability of a potentially diagnostic abnormality increased logarithmically with the duration of seizure disorder (p<0.05), controlling for all other factors. The duration of epilepsy, as compared to non-epileptic seizure disorder, further increased the likelihood of structural abnormalities (p<0.05) but not metabolic abnormalities (p>0.5). Patients with temporal lobe epilepsy were much more likely to exhibit abnormalities compared to all other patients (p<0.01). Counterintuitively, a diagnosis of extratemporal epilepsy did not increase the likelihood of metabolic abnormalities (p>0.5) but it did tend to increase the likelihood of structural abnormalities (p<0.10). No other effects were significant in either model (p>0.20).Conclusions: This model both confirms and challenges the concept that neuroimaging will only reveal diagnostic changes 10 to 15 years after the onset of seizure disorder. This guidance is most appropriate for predicting the presence of structural abnormalities in patients with extratemporal epilepsies. In contrast, patients with temporal lobe epilepsies may exhibit either structural or metabolic changes much earlier. This, however, presents a dilemma that neuroimaging could provide evidence that the seizures come from a temporal focus but, in order to justify neuroimaging, there should be suspicion of a temporal focus. The lack of effect of a diagnosis of extratemporal epilepsy on the presence of metabolic abnormalities may reflect a relative insensitivity to extratemporal metabolic lesions. Further, the finding that the duration of both epileptic and non-epileptic seizure disorder increased likelihood for both structural and metabolic abnormalities suggests that antiepileptic drugs or other factors common to both patient populations may be associated with the presence of abnormalities. This model may help neurologists better understand when diagnostic neuroimaging may be useful to diagnose and/or treat patients with seizures.