7T fMRI multivariate pattern analysis shows worry-related brain patterns become more distinguishable with age and are unrelated to symptom severity

During a personalized worry-induction task, multivariate patterns of activity across 268 brain regions distinguished worry from neutral states with 66.71% cross-validated accuracy, 62.61% balanced accuracy, 80.43% sensitivity, and 44.78% specificity. Prior work on the neural correlates of worry largely described group averages; this study characterizes individual variance with multivariate pattern analysis and reports quantifiable classification performance. A random-forest model with leave-one-subject-out cross-validation in 69 participants aged 12 to 31, using activity averaged across 268 Shen-atlas regions; sensitivity and specificity are asymmetric, with lower specificity.

Trial-level classification accuracy increased with in-scanner worry ratings regardless of condition (b=0.62, p<.001). It links model performance directly to participants' momentary subjective worry, indicating the classifier captures signals tied to worry experience rather than only task condition. A trial-level regression result that is statistically significant; the association is observational and does not establish causal direction.

Classification accuracy (r=0.30, p=.01) and balanced accuracy (r=0.31, p=.01) increased significantly with age but were unrelated to symptom severity. It provides an initial indication that neural patterns of worry become more differentiated from adolescence into young adulthood, and that this differentiation does not track current symptom levels. Cross-sectional sample of n=69 spanning ages 12 to 31, with small-to-moderate correlations; symptom measures were self-reports (PSWQ, SCARED, MFQ).

The authors propose that early adolescence may be an opportune window for interventions to alter worry, and that interventions adapted from adult populations may require tailoring for adolescent patients. It connects developmental differentiation of neural patterns to intervention timing, offering a hypothesis for age-adapted adolescent anxiety interventions. An inference from cross-sectional correlational findings; no intervention trial data are presented.

The results apply to adolescents and adults aged 12 to 31 with varying worry symptoms, obtained in a personalized worry-induction 7T fMRI setting; data are publicly available on the NIMH Data Archive and code on GitHub, supporting reproduction and reanalysis. For researchers, this offers a testable starting point for using multivariate neural markers in developmental psychopathology; for clinicians and intervention designers, it suggests that worry interventions for adolescents may need to differ from adult protocols.

As a fast parse containing only the abstract, this summary lacks figures, regional weights, and model details, so it cannot identify which brain regions or networks drive the age-related increase in distinguishability, nor assess how low specificity affects the conclusions. The cross-sectional correlational design cannot separate age effects from cohort effects, and the null association between symptom severity and neural patterns needs testing in larger and longitudinal samples. Implementation details of the worry-induction task and self-report measures are not expanded in the abstract, so readers assessing ecological validity should consult the full article.