Associations of Iron Indicators with Cognitive Impairment Risk in Postmenopausal Women: Evidence from a Matched Case-Control Study and Back Propagation Neural Network Model.

Iron is essential for normal cognitive function, and women have nearly twice the age-matched prevalence of cognitive impairment compared with men. The matched case-control study included 1344 postmenopausal women aged over 60 years, and cognitive status was assessed using Mini-Cog. Plasma ferritin and hepcidin were detected by enzyme-linked immunosorbent assay (ELISA), and total iron binding capacity (TIBC) was determined by colorimetry. Conditional logistic regression and restricted cubic splines (RCS) were used to analyze the relationship between each iron indicator and cognitive impairment risk, with stratified analyses by age and body mass index (BMI) conducted. Additionally, sensitivity analysis was performed to verify the robustness of the results. Back-propagation neural network (BPNN) was utilized as exploratory supplementary tool to quantify the relative importance of multiple factors. The results indicated that elevated hepcidin levels were positively correlated with higher cognitive impairment risk (OR = 1.515, 95% CI: 1.067-2.150). RCS analysis indicated a non-linear association between TIBC and the risk of cognitive impairment (P for nonlinear = 0.002), with exploratory inflection point observed at 100.69 μmol/L. In contrast, ferritin levels showed no statistically significant association with cognitive impairment risk. The BPNN model identified hepcidin (95.5% importance) and TIBC (79.1%) as having relatively high importance, alongside certain blood indicators (albumin, MONO, HDL, TP), listening to the radio and reproductive lifespan. The internal validation of the model AUC reached 0.744. These findings suggested that hepcidin and TIBC were associated with cognitive impairment risk in postmenopausal women, offering evidence for cognitive decline in this population.