The role of circulating metabolome in cognitive impairment is inconclusive, and whether the associations are in the severity-dependent manner remains unclear. We aimed to identify plasma metabolites associated with cognitive impairment and evaluate the added predictive capacity of metabolite biomarkers on incident cognitive impairment beyond traditional risk factors. In the Rugao Longevity and Ageing Study (RuLAS), plasma metabolome was profiled by nuclear magnetic resonance spectroscopy. Participants were classified into the cognitively normal, moderately impaired, and severely impaired groups according to their performance in two objective cognitive tests. A two-step strategy of cross-sectional discovery followed by prospective validation was applied. In the discovery stage, we included 1643 participants (age: 78.9 ± 4.5 years) and conducted multinomial logistic regression. In the validation stage, we matched 68 incident cases of cognitive impairment (moderately-to-severely impaired) during the 2-year follow-up with 204 cognitively normal controls by age and sex at a 1:3 ratio, and conducted conditional logistic regression. We identified 28 out of 78 metabolites cross-sectionally related to severely impaired cognition, among which IDL particle number, ApoB in IDL, leucine, and valine were prospectively associated with 28%, 28%, 29%, and 33% lower risk of developing cognitive impairment, respectively. Incorporating 13 metabolite biomarkers selected through Lasso regression into the traditional risk factors-based prediction model substantially improved the ability to predict incident cognitive impairment (AUROC: 0.839 vs. 0.703, P < 0.001; AUPRC: 0.705 vs. 0.405, P < 0.001). This study identified specific plasma metabolites related to cognitive impairment. Incorporation of specific metabolites substantially improved the prediction performance for cognitive impairment.
Read full abstract