Multivariate neural signatures of worry become more differentiated across adolescence and young adulthood: An ultra-high field (7T) fMRI study

Worry is an important contributor to anxiety and depressive disorders, particularly in adolescence. Underlying changes in brain networks are thought to drive development of pathological worry, yet the degree to which neural correlates of worry vary by age is unknown. Here, we use multivariate pattern analysis (MVPA) to investigate individual variance in brain activity during worry in a cross-sectional sample of adolescents and adults varying in worry symptoms. Participants (n=69; ages 12-31) completed a personalized worry-induction fMRI task. Brain activity was averaged across k=268 brain regions from the Shen atlas. Participants also completed self-reports assessing trait worry (PSWQ), anxiety (SCARED), and depression (MFQ) severity. A random-forest model was trained on estimated parameters from worry vs. neutral blocks using leave-one-subject-out cross-validation. Age and symptomatology measures were correlated with model performance metrics. The overall cross-validated model had accuracy=66.71%, balanced accuracy=62.61%, sensitivity=80.43%, and specificity=44.78%. Trial-level accuracy increased with in-scanner worry ratings regardless of condition (b =0.62, p<.001). Accuracy (r=0.30, p=.01) and balanced accuracy (r=0.31, p=.01) significantly correlated with age but did not relate to symptom severity. Brain patterns associated with worry become more differentiated across adolescence, suggesting that neural processes of worry consolidate across this period. Thus, early adolescence may be an opportune time for interventions to alter worry, but interventions adapted from adult populations may require tailoring for adolescent patients.