Poster #12 - Identifying Phenotypic Clusters in Relation to Psychiatric Disorders During Early Adolescence Using Latent Profile Analysis: an ABCD Study

• In Event: Poster Session 20

Abstract

Background
Rates of severe psychiatric symptoms in children have steadily increased over time (Cocozza and Skowyra, 2000). Several risk factors may lend vulnerability to mental illness in adolescents, including psychological, cognitive, and health-related factors (O’Connell, et al., 2009). Importantly, 75% of mental illnesses emerge before the age of 25. Without early prevention and treatment, psychiatric disorders affect both the adolescent years and decrease adulthood productivity (McGorry and Mei, 2018). To establish effective early interventions, clinicians need to identify children who are the greatest risk for future psychiatric challenges. Here, we aimed to examine whether psychological, cognitive, and health-related variables can be used to generate phenotypically distinct clusters of young adolescents that differentiate between those with and without psychiatric clinical relevance.

Methods
The Adolescent Brain and Cognitive Development (ABCD) Study is a multi-site, longitudinal study of adolescent brain and cognitive development in the United States. After removing subjects with missing data, we applied an exploratory factor analysis (EFA) using psychological, cognitive, and health features (42 variables) to extract latent factors underlying this phenotypic space (N=9,587, 48% girls, mean age=9.9 years).We then used latent profile analysis (LPA) to identify data-driven clusters based on the factors extracted from the EFA. We compared the identified clusters on measures not included in the cluster procedure, such as demographics and lifetime psychiatric disorder diagnostics.

Results
Twelve latent factors were extracted from the EFA broadly spanning psychological, cognitive, and reward-seeking domains (TLI=0.924, CFI=0.924, RMSEA=0.034). The LPA generated eight clusters of adolescents with different phenotypic profiles. Three of these clusters (Cluster 2, 4, and 8) displayed increased psychological symptoms. Cluster 2 (N=602, 47% girls) had a higher proportion of adolescents with internalizing, externalizing, eating disorder diagnoses (p<0.001), and high family income (​≥$100,000) relative to other clusters (p<0.001). Cluster 4 (N=636, 35% girls) displayed lower performance on executive function tasks, lack of planning/perseverance and increased externalizing behavior. In addition to having an increased incidence for the same psychiatric diagnoses as Cluster 2, Cluster 4 also had a high proportion of Psychotic, Bipolar, and Post-Traumatic Stress Disorder (p<0.001). This cluster had a higher proportion of adolescents with low family income (<$50,000; p<0.001). Finally, Cluster 8 (N=1113, 44% girls) displayed high externalizing behavior and psychiatric diagnoses identical to Cluster 4 (p<0.001), except for eating disorders, which was not present. Similar to Cluster 4, this cluster had a higher proportion of adolescents with low family income (<$50,000; p<0.001).

Conclusion
Our results suggest that a data-driven analysis of highly dimensional behavioral, cognitive, and health factors can identify clusters of young adolescents with distinct, clinically relevant phenotypic profiles. Future research will comprehensively characterize the eight phenotypic clusters as well as follow the trajectories of these high psychiatric clusters over time. Understanding the trajectories of these clusters will facilitate the identification of early risk factors for later psychiatric disease and promote early interventions.

Authors

• Zixuan Zheng

• Leyla R Brucar, University of Minnesota - Twin Cities

• Andrea M Maxwell, University of Minnesota - Twin Cities
  Presenting Author

• Gregory Niklason, University of Minnesota - Twin Cities

• Anna Zilverstand, University of Minnesota - Twin Cities