Cadmium-associated cognitive impairment in adults: Proteomics-guided blood-based risk stratification with external validation and rat multimodal corroboration.

Abstract

Cadmium exposure is associated with cognitive decline, yet affected communities lack compact tools linking pollutant burden to cognitive-risk triage. We developed an exposure-informed blood model for screening-defined cognitive impairment (CI) and evaluated transportability and biological plausibility. Among 307 screened adults, 198 were eligible; propensity-score matching yielded 25 CI and 25 cognitively normal participants for derivation. Plasma proteomics and stability-oriented selection identified a four-marker model comprising serum cadmium (Cd), APOE ε4, high-density lipoprotein cholesterol, and plasma purine nucleoside phosphorylase (PNP). The locked model was tested in 93 participant-independent adults from two hospitals in the same regional network, with PNP measured by enzyme-linked immunosorbent assay (ELISA). Cd showed the most consistent inverse association with Mini-Mental State Examination scores, particularly in APOE ε4 carriers and older adults. The model achieved an internal area under the curve (AUC) of 0.92 and external AUC of 0.778 (average precision, 0.729). Intercept updating improved the Brier score from 0.591 to 0.214; a calibration slope of 0.366 indicated that absolute probabilities required local adjustment. Decision-curve analysis supported locally calibrated triage across 0.20-0.50 thresholds. Cadmium-exposed male rats showed spatial-memory retention deficits, hippocampal-prefrontal desynchronization, and mitochondrial-synaptic proteomic disruption involving PNP/WDR81. Parsimonious modeling, bootstrap stability assessment, locked two-site validation, and sensitivity analyses reduced dependence on a single split, algorithm, site, or predictor specification. Remaining uncertainty concerns coefficient precision, lifetime-dose representation, temporal attribution, geographic calibration, and sex-generalizable translation. These findings support an assay-compatible, locally recalibrated framework for prioritizing cognitive evaluation in comparable cadmium-affected settings.