A Multimodal Approach to Predicting Cognitive Impairment and Elucidating Neural Mechanisms in an HPA Axis Hyperactivation Model of Cushing’s Disease

Abstract Background: Chronic hyperactivation of the HPA axis is closely linked to cognitive impairment. Using CD as a natural model, this study assessed cognitive function, integrated multi-omics features, and explored related brain structural mechanisms. Methods: A total of 103 CD patients and 84 healthy controls underwent MoCA assessments, structural MRI, and multi-omics profiling. LASSO and logistic regression were used to select key predictors and build a cognitive impairment model. Partial correlation and mediation analyses were conducted to explore how HPA axis activation affects cognition via brain structure. Results：CD patients showed impairments in executive function, attention, language, and memory. The final model identified COR8am, CINGPOSTL, and HIPPL as key predictors (AUC = 0.872). Mediation analysis showed COR8am influenced cognition indirectly through changes in CINGPOSTL and HIPPL volumes. Conclusion: This study established an accurate multi-omics prediction model and revealed a structural brain pathway linking HPA axis hyperactivation to cognitive decline, offering a basis for early detection and targeted interventions.