Scholars have long been interested in the association between arsenic (As) exposure and neurological disorders; however, existing systematic epidemiological investigations are insufficient and lack the inclusion of diagnostic or predictive biological markers. This study sought to evaluate the association between As exposure and cognitive impairment and identify potential biomarkers by developing predictive models. Here, we found that logarithm (Ln)-transformed urinary As concentrations were negatively linearly related to the mini-mental state examination (MMSE) score exposure-response curves. Subsequently, we identified a unique plasma neurometabolite profile in subjects exposed to As compared with the reference group. Further analyses showed that tryptophan, tyrosine, dopamine, epinephrine, and homovanillic acid were all significantly associated with both urinary As concentrations and MMSE scores. Notably, the association between As exposure and MMSE scores was partly mediated by tryptophan, tyrosine, dopamine, and epinephrine. Importantly, an unprecedented prediction model utilizing neurotransmitters was established to assess the risk of cognitive impairment due to As exposure. A 91.1% consistency rate was found between the predicted and the actual probabilities. Additionally, machine learning models also produced highly accurate predictions. Overall, this study revealed a dose-dependent cognitive decline in As-exposed adults accompanied by a disturbance in the signature of neurotransmitter metabolites, offering new predictive insights.
Keywords: MMSE score; arsenic; cognitive impairment; neurotransmitter; risk assessment.