Linking post-stress brain connectivity to acute cortisol reactivity using network-based inference and prediction

Acute cortisol responses to psychosocial stress differ greatly among individuals, but the relationship between cortisol levels and post-stress brain connectivity is not well understood. Previous studies have been limited by small sample sizes, region-of-interest approaches, and a focus on group-level analyses. This study aimed to determine if acute cortisol increase is associated with and can be predicted from whole-brain resting-state functional connectivity (rsFC) using network-based inference and prediction techniques.

The researchers analyzed 339 healthy participants from two ScanSTRESS datasets. First, they used the Network-Based Statistic (NBS) to identify connected rsFC networks associated with acute cortisol increase, while controlling for age, site, and sex/hormonal status. Next, they used NBS-Predict and Connectome-Based Predictive Modeling (CPM) to predict participants' acute cortisol increase from their connectivity patterns.

The results showed that greater cortisol responses were linked to lower post-stress rsFC within a significant distributed network of 258 connections among 78 regions. This network was centered on thalamic nuclei and pallidal regions and extended to default-mode, limbic, orbitofrontal, and cerebellar regions. The study found a significant negative association between cortisol responses and rsFC in females, although formal sex-difference contrasts were not significant.

Furthermore, NBS-Predict and CPM yielded modest but significant out-of-sample prediction of acute cortisol increase. The predictive networks converged on subcortical and posterior cingulate regions, suggesting a network-level perspective on neural-endocrine coupling following acute stress. In conclusion, post-stress rsFC carries both inferential and predictive information about individual hypothalamic-pituitary-adrenal (HPA) axis reactivity.