Abstracts

Rationale: Research on cognition and behavior in epilepsy is moving beyond simple correlations between test scores and focal brain lesions, toward a more sophisticated network-based understanding. This shift requires precise definitions of cognitive and behavioral dysfunction and careful study of longitudinal symptom trajectories. Veterans with epilepsy, especially those with a history of traumatic brain injury (TBI), represent a high-risk population where the interplay of cognitive, behavioral, and neurological factors remains poorly understood. This study aimed to identify distinct neurobehavioral phenotypes in post-9/11 veterans with epilepsy and examine their socio-demographic correlates, comorbidities, and TBI history.

Methods: We conducted a detailed chart review supported by natural language processing (NLP) to identify cognitive dysfunction indicators from medical records. Extracted concepts were categorized into seven Research Domain Criteria (RDoC) domains. Using Uniform Manifold Approximation and Projection (UMAP) for clustering and dimensionality reduction, we analyzed the frequency of 39 behavioral and cognitive concepts across medical notes. Four study groups (n=200 each) were compared: Epilepsy Only, TBI Only, TBI before Epilepsy, and TBI after Epilepsy (post-traumatic epilepsy; PTE).

Results: Groups were sociodemographically similar, though the TBI before Epilepsy group had the highest marriage rates (52%). Significant health disparities emerged. Seizures were most prevalent in TBI before Epilepsy (64%, p< 0.001). Headaches (74%) and insomnia (47.5%) were most frequent in combined epilepsy-TBI groups. Mental health differences were striking: anxiety (65.5%), suicidal ideation (27%), bipolar disorder (39%), and PTSD (79.5%) peaked in the TBI before Epilepsy group (all p< 0.05). Clustering revealed four phenotypes defined by symptom severity and domain (cognitive vs. behavioral). Veterans with TBI before Epilepsy disproportionately exhibited high cognitive risk, while those with PTE clustered toward high behavioral risk.

Conclusions: This study advances our understanding of neurobehavioral heterogeneity in veterans with epilepsy and TBI by identifying distinct phenotypic clusters that transcend traditional diagnostic categories. Severity and symptom profiles varied within conventional groups in clinically meaningful ways. Veterans with post-traumatic epilepsy exhibited significant behavioral challenges, while those with preceding TBI faced pronounced cognitive difficulties. These findings have important implications for risk stratification and treatment planning. By leveraging advanced computational methods to parse clinical complexity, this work provides a foundation for developing more precise, personalized approaches to managing the neurobehavioral sequelae of epilepsy and TBI. Importantly, it emphasizes the need to explore integrated, phenotype-informed care models that could help address longstanding challenges in improving quality of life for veterans affected by these conditions.

Funding: Congressionally Directed Medical Research Programs.