To characterize outcomes following pediatric epilepsy surgery across a broad range of cognitive domains using empirical methods (i.e., reliable change indices: RCIs), compare these outcomes with those based on traditional methods (i.e., standard deviation: SD), and identify factors associated with postoperative cognitive declines and/or improvements.

The cognitive risks of pediatric epilepsy surgery, the most effective treatment option for children with intractable epilepsy, remain unclear. Most existing studies have used non-empiric methods and demonstrated inconsistent results.

This retrospective cohort study included 186 children who underwent surgical resection for treatment of pharmacoresistant epilepsy and who completed pre- and postoperative neuropsychological assessments. Postoperative testing occurred ~6.5 months after surgery and included measures of intelligence, attention/working memory, processing speed, language, executive functioning, visuospatial skills, memory, and academic achievement. Change scores for each patient were classified as decline, no change, or improvement using epilepsy-specific RCIs. Chi-square goodness of fit tests compared the distribution of outcomes as classified with RCIs to those obtained using a traditional one SD cutoff. Multinomial regression analyses were conducted to identify factors associated with cognitive decline and/or improvement.

Most children experienced relatively focal cognitive changes postoperatively (declines and/or improvements in 1-2 domains). Rates of postoperative decline and improvement across individual cognitive domains ranged from 4-35% and 2-31%, respectively. Compared to RCIs, SD methodology often overestimated postoperative improvements and varied with respect to declines. Factors associated with RCI decline or improvement included preoperative performance, age at surgery, surgery site, and postoperative seizures.

Results suggest substantial variability in individual cognitive outcomes ~6.5 months following pediatric epilepsy surgery. The differences in change distributions obtained using RCIs versus SDs highlight the need for studies using empiric methodology to study postoperative cognitive change. Variables associated with postoperative cognitive change may be used to develop multivariable prediction models to aid clinical decision-making and patient counseling.