Iron metabolism and clinical characteristics for early identification of frailty in elderly hemodialysis patients: a machine learning and mediation analysis approach

To investigate the association between iron metabolism and frailty in elderly maintenance hemodialysis (MHD) patients and to identify potential biomarkers and predictive models. Transcriptomic datasets were analyzed to screen iron metabolism–related genes, followed by WGCNA and PPI to identify core genes. Clinical validation was performed using qRT-PCR, ROC analysis, and correlation with handgrip strength. Machine learning models and a nomogram were developed to predict frailty, and mediation analysis was applied to assess the role of muscle strength. Three core genes (CCL5, CD8A, and ERBB2) were identified and validated, all significantly downregulated in frail patients and correlated with handgrip strength. The combined gene panel achieved good predictive performance (AUC = 0.843). Machine learning models integrating iron indices demonstrated robust discrimination (AUCs > 0.80), and mediation analysis suggested handgrip strength as a partial mediator between TSAT and frailty. Iron metabolism and muscle function are closely associated with frailty and may be involved in its underlying biological processes, and integrated models may enable early recognition and intervention in MHD patients.