Two clinical parameters measurable at seizure onset appear to predict a return to baseline after prolonged status epilepticus (SE), based on a study of patients who presented to a single, tertiary academic medical center over a 12-year period.

Absence of nonconvulsive SE with coma and a decreasing Charlson Comorbidity Index were the only independent predictors for return to baseline in patients with SE duration greater than 48 hours, the researchers found. However, the research fell short of developing a model for identifying patients at risk for prolonged SE.

“These findings are of great clinical importance, as up to now, clinicians have had no reliable prediction tools to direct decisions regarding the level of care with progressive SE duration. Early and reliable identification of patients with potential favorable outcome despite having SE for several days is of utmost clinical importance, as this insight may urge clinicians to intensify treatment rather than consider care withdrawal as systemic and neurologic sequelae increase, and chances of SE termination decrease over time,” first author Raoul C. Sutter, MD, of University Hospital Basel (Switzerland), and his colleagues wrote about their findings in Epilepsia.

The researchers identified 467 adult patients with prolonged SE at University Hospital Basel during 2005-2016 – excluding those with SE as a consequence of hypoxic‐ischemic brain injury – who had a median age of 66.7 years and median SE duration of 1 day. While 11.8% of patients died in the hospital and 12.4% at 30 days after SE onset, 40.9% made a complete neurologic and functional recovery to their premorbid status.

There were significant differences in in-hospital outcomes between patients with different SE durations. For example, rates of returning to baseline differed significantly at 55.6% of those with a SE duration of 0-12 hours, 36.8% with 12-24 hours’ duration, 34.6% with 24-48 hours’ duration, and 25.5% with more than 48 hours.

A multivariable regression model identified absence of nonconvulsive SE with coma and a decreasing Charlson Comorbidity Index as the only independent predictors for return to baseline in patients with SE duration greater than 48 hours, and both remained significant predictors after adjustment for use of anesthetics and vasopressors. These predictors of a return to baseline after prolonged SE remained significant after excluding patients who died. This two-variable prediction model had an area under the receiver operating curve (AUROC) of 0.82, “indicating good discrimination,” and an AUROC of 0.76 following cross-validation.

The investigators also sought to develop a model to identify patients at risk for prolonged SE, but the model showed relatively poor discriminative ability with AUROCs of just 0.67-0.72 for predicting no termination of SE within 12, 24, or 48 hours. “Our attempt to generate a highly reliable prediction model for early recognition of patients at increased risk for developing prolonged SE failed, as demonstrated by the rather small AUROC and the fact that sensitivity analyses after exclusion of patients who died revealed inconsistent association of the identified predictors,” they wrote.

Prior reports identified younger age, absence of acute brain lesions at presentation, and the absence of multiple concomitant medical problems as factors associated with favorable outcome after prolonged SE, but “none of the studies performed multivariable regression models and generated or tested predictions models in this context,” they noted.

The authors cautioned that “although internal cross-validation of the final prediction model indicated adequate performance [based on an AUROC of 0.76], further external validation of our prediction model is warranted before our prediction model can be implemented and used for decision making in daily clinical practice.”

Some authors reported receiving research, travel, and/or personal grants or speaker fees from companies marketing antiepileptic drugs, such as UCB, Eisai, and GlaxoSmithKline.

jevans@mdedge.com

SOURCE: Sutter RC et al. Epilepsia. 2018 Nov 22. doi: 10.1111/epi.14603

Model may predict prolonged status epilepticus outcomes