Abstracts

Rationale: Latent profile analysis (LPA) is a person-centered statistical approach for identifying underlying subgroups within a population and may help uncover latent phenotypes that characterize heterogenous clinical samples. LPA has been applied in adults with temporal lobe epilepsy. This is the first study to apply LPA to a multi-site sample of pre-surgical pediatric epilepsy patients.

Methods: 791 pediatric patients between ages 6 years, 0 months and 21 years, 8 months underwent neuropsychological assessment (see Table 1). Summary scores across nine cognitive domains (adaptive, attention/working memory, executive function, language, motor, processing speed, verbal memory, visual memory, and visuospatial) were calculated using multiple measures administered within each domain. LPA were conducted using maximum likelihood estimation with robust standard errors. Multiple fit indices (e.g., negative log likelihood, Bayesian Information Criterion, Consistent Akaike Information Criterion, etc.) were calculated and examined for each model with the final model selected based on the convergence of evidence. This full profile enumeration procedure was replicated in the full sample (n = 791), as well as two subsets: patients with focal epilepsy (n = 643), and patients with temporal lobe epilepsy (TLE; n = 259).

Results: In the full sample and the TLE only sample, fit indices supported a 4-profile solution. Profiles were driven by general level of cognitive functioning across domains, with cognitive phenotypes characterized by broadly average, low average, below average, and exceptionally low functioning. Greater profile differentiation based on specific domains was supported by the fit indices from the patients with focal epilepsy. A 5-profile solution was still broadly driven by general cognitive functioning (Figure 1); however, one low average profile had worse language and verbal memory functioning (profile 4), while the other low average profile had worse motor and executive functioning (profile 3). Average posterior probabilities were high, suggesting well-separated profiles. Of note, across all latent profiles, fine motor functioning consistently emerged as an area of weakness.

Conclusions: Consistent with finding with adult epilepsy patients, LPA identified subgroups of pediatric epilepsy patients ranging from generalized neurocognitive impairments to intact neurocognitive functioning. This was evident in the full sample as well as those with focal or TLE only. Unique profiles with domain-specific impairment were identified in the sample when it was restricted to focal epilepsies suggesting that future work is needed to understand neuroanatomical, neurophysiological, and pharmacologic factors that may drive phenotypical group membership. Identification of these cognitive phenotypes has the potential to support more targeted and personalized assessment and cognitive intervention.

Funding: This study was funded by the Pediatric Epilepsy Research Foundation.