Hair neuroendocrine signatures of acute stress reaction symptoms: Cross-national validation and latent profile heterogeneity.

The Acute Stress Reaction (ASR) involves transient emotional, somatic, cognitive, and behavioral symptoms after stressful exposure. Acute stress engages neuroendocrine responses and endocannabinoid systems (ECS), yet their joint links to multidimensional ASR symptoms remain unclear. This study characterizes coupling between hair-based neuroendocrine and ECS markers and ASR symptom dimensions, and tests consistency across countries and symptom severity profiles. Hair steroids, endocannabinoid concentrations and ASR symptoms were assessed in 688 participants from France, China, and Germany during the COVID-19 pandemic. Network analysis examined within-system structure and biomarker-symptom bridging. Differences in networks were compared across countries and ASR severity subgroups, identified through latent profile analysis (LPA). Machine learning (ML) tested out-of-sample transferability of these biomarkers across national and symptom subgroups. The results found that cortisone and N-arachidonoylethanolamine (AEA) were central nodes in the steroid-endocannabinoid network, whereas dissociation showed the highest centrality in the ASR symptom network. AEA, testosterone, avoidance and social function impairment exhibited the strongest bridge connectivity in the biomarker-symptom network. Steroid-endocannabinoid networks varied across countries, and ASR symptom networks diverged between China and Europe. LPA identified low- and high-reaction ASR subgroups, and AEA-avoidance exclusively linked to the low-reaction subgroup. ML indicated robust predictive utility for cortisone and consistent prediction of hyperarousal across countries, but cross-national and cross-subgroup generalization was poor. These findings suggest that stress-related neuroendocrine and ECS markers in hair are linked to ASR symptom dimensions in a coupled network framework, emphasizing the context-dependent stress biology and motivates stratified, population calibrated modelling rather than one-size-fits-all prediction.