• Center on Health Equity & Access
  • Clinical
  • Health Care Cost
  • Health Care Delivery
  • Insurance
  • Policy
  • Technology
  • Value-Based Care

Machine-Learning Approach Identifies 3 Behavioral Phenotypes of TLE

Article

Patients in this study, who had overall significantly higher scores than controls, fell into 3 categories of psychological risk from temporal lobe epilepsy (TLE) based on analysis with unsupervised machine learning.

A group of researchers have characterized 3 distinct groups of patients based on patterns of behavior in temporal lobe epilepsy (TLE), underscoring the heterogeneity of the disorder.

Called the Epilepsy Connectome Project, the study included 114 patients and 83 controls. Findings were published recently in Brain Communications.

All participants completed the Achenbach System of Empirically Based Assessment inventory, offering insights into behaviors such as depression, anxiety, somatic problems, avoidant personality problems, attention deficiency/hyperactivity, and antisocial personality problems.

The patients who had overall significantly higher scores than controls fell into 3 categories of psychological risk based on analysis with unsupervised machine learning:

  • Cluster 1: unaffected with no scale elevations compared with controls
  • Cluster 2: mildly symptomatic with significant elevations across some scales compared with controls
  • Cluster 3: severely symptomatic with significant elevation across all scales compared with controls and the other groups

“The person-centered analytic approach undertaken here underscores the heterogeneity of behavioral risk among patients with TLE and resulted in a more clinically meaningful separation of groups than the traditional TLE vs control comparison,” commented the researchers, noting that, to date, just 2 other groups have taken this same approach.

The researchers simultaneously validated these cluster findings by identifying abnormalities on measures included in quality of life (QOL) metrics and the NIH Toolbox Emotion Battery of 10 subtests of negative affect, psychological well-being, stress and self-efficacy, and social relationships.

Data from the NIH Toolbox Emotion Battery showed that compared with controls, patients who fell into Cluster 2 showed significantly impaired life satisfaction (P < .001) and self-efficacy (P = .001), reporting worse scores for loneliness (P = .008), perceived stress (P < .001), fear somatic (P < .001), and anger hostility (P = .017). Compared with Cluster 1, patients in Cluster 2 scored significantly worse on all but fear effect and anger-physical aggression scales.

Patients in Cluster 3 had significantly worse scores compared with controls on all scales, with a range of P = .01 for the anger-physical aggression scale to P < .001 on remaining scales. Similarly, patients in Cluster 3 had significantly worse scores than patients in Cluster 1, with all scales at P ≤0.001.

QOL, measured with the Patient-Weighted Quality of Life in Epilepsy Questionnaire scale, showed that patients who fell into Cluster 1 had significantly better QOL than patients in Cluster 2 in all but 1 scale and that patients who fell into Cluster 3 had significantly worse QOL than patients in Cluster 2 across all scales.

“These independent measures, examined as a function of the identified behavioral phenotypes, mirrored the identical stepwise patterns of abnormality reflected in the latent groups,” explained the researchers. “Clearly, the behavioral phenotypes have implications for other dimensions of behavior and function. In this investigation, we did not have independent psychiatric diagnoses that would have been a valuable measure of concurrent validity. But prior investigations have shown that identified behavioral clusters do have external validity demonstrated through their relationship with formal psychiatric assessments.”

The researchers observed no apparent effects of behavioral cluster type across various cognition measures, including on a reading recognition test, picture vocabulary test, or a list sorting memory test. However, the patients in Cluster 1 were significantly faster than patients in Cluster 3 for the Pattern Comparison Processing Speed Test.

Reference

Struck A, Garcia-Ramos C, Nair V, et al. The presence, nature and network characteristics of behavioural phenotypes in temporal lobe epilepsy. Brain Commun. Published online March 30, 2023. doi:10.1093/braincomms/fcad095

Related Videos
Screenshot of Susan Wescott, RPh, MBA
4 KOLs are featured in this series
4 KOLs are featured in this series
Justin Oldham, MD, MS, an expert on IPF
Justin Oldham, MD, MS, an expert on IPF
Screenshot of Stephanie Hsia, PharmD
1 KOL is featured in this series.
Female doctor in coat with stethoscope on blue background - Pixel-Shot - stock.adobe.com
4 KOLs are featured in this series
5 KOLs are featured in this series.
Related Content
© 2024 MJH Life Sciences
AJMC®
All rights reserved.