Abstract
Objective
We sought to determine which combination of clinical and electroencephalography (EEG) characteristics differentiate between an antiseizure medication (ASM)–resistant vs ASM-responsive outcome for patients with idiopathic generalized epilepsy (IGE).
Methods
This was a case-control study of ASM-resistant cases and ASM-responsive controls with IGE treated at five epilepsy centers in the United States and Australia between 2002 and 2018. We recorded clinical characteristics and findings from the first available EEG study for each patient. We then compared characteristics of cases vs controls using multivariable logistic regression to develop a predictive model of ASM-resistant IGE.
Results
We identified 118 ASM-resistant cases and 114 ASM-responsive controls with IGE. First, we confirmed our recent finding that catamenial epilepsy is associated with ASM-resistant IGE (odds ratio [OR] 3.53, 95% confidence interval [CI] 1.32–10.41, for all study subjects) after covariate adjustment. Other independent factors seen with ASM resistance include certain seizure-type combinations (absence, myoclonic, and generalized tonic-clonic seizures [OR 7.06, 95% CI 2.55–20.96]; absence and generalized tonic-clonic seizures [OR 4.45, 95% CI 1.84–11.34]), as well as EEG markers of increased generalized spike-wave discharges (GSWs) in sleep (OR 3.43, 95% CI 1.12–11.36 for frequent and OR 7.21, 95% CI 1.50–54.07 for abundant discharges in sleep) and the presence of generalized polyspike trains (GPTs; OR 5.49, 95% CI 1.27–38.69). The discriminative ability of our final multivariable model, as measured by area under the receiver-operating characteristic curve, was 0.80.
Significance
Multiple clinical and EEG characteristics independently predict ASM resistance in IGE. To improve understanding of a patient’s prognosis, clinicians could consider asking about specific seizure-type combinations and track whether they experience catamenial epilepsy. Obtaining prolonged EEG studies to record the burden of GSWs in sleep and assessing for the presence of GPTs may provide additional predictive value.
OCT