Background: The onset of psychopathology peaks during adolescence; further, mental health disorders with onset during adolescence are associated with more severe and recurring symptoms in adulthood. Adolescence is also characterized by continued brain development, including refinement of functional networks (i.e., collections of brain regions with similar functions) that support cognitive and affective processing. While previous network studies of adolescent mental health have provided foundational knowledge of neurobiological abnormalities associated with specific mood symptoms and disorders, they have primarily examined contemporaneous correlations between the time series of brain regions and ignored temporal dynamics of fMRI signals. While correlations between time-series are simpler to calculate, simulation studies show that spurious connections are identified when researchers do not account for the temporal dependencies between brain regions. In this context, the Group Iterative Multiple Model Estimation (GIMME) is a data-driven approach that uses unified structural equation modeling to explain patterns of temporal covariation among variables. Using GIMME, we modeled directed functional connections at the group level and identified subgroups of individuals based on the similarity of their directed connectivity maps across the executive control and default mode brain networks. We then used Generalized Additive Mixed Models to test whether resulting neural subgroups differ in internalizing and externalizing symptom trajectories throughout puberty.
Method: In 173 adolescents (99 females), ages 9-14 years (M=11.30, SD=1.08), we applied GIMME to the executive control (ECN), and ventral and dorsal default mode (v/dDMN) networks. Participants reported their internalizing and externalizing symptoms at up to three time-points (each spaced approximately two years apart). 80% of the participants were in beginning stages of puberty at the first time-point, and over 80% of the participants were near or completed puberty by the third time-point. We used Generalized Additive Mixed Models to test whether connectivity subgroups differentiate trajectories of internalizing and externalizing symptoms throughout puberty.
Results: On average, participants increased in internalizing and externalizing symptoms throughout puberty, however their trajectories varied significantly. Participants who belonged to ECN and DMN subgroups with lower levels of connections between fronto-parietal and fronto-thalamic regions had higher initial levels of internalizing and externalizing symptoms, and showed no significant changes throughout puberty. One ECN subgroup, with greater frontal->parietal and parietal->frontal connections did not show an association between early life stress severity and internalizing symptoms. Conversely, participants who belonged to more inter-connected ECN and DMN subgroups had lower initial levels of internalizing symptoms that increased significantly throughout puberty. Neural subgroup differences in externalizing trajectories did not survive multiple comparisons corrections.
Conclusions: Neural heterogeneity in early puberty is meaningfully related to the course of internalizing symptoms throughout puberty. Directed connections may reveal which youth may benefit from interventions offered at different developmental windows.