An integrative approach to detecting potential blood-based biomarkers of cognitive frailty.

OBJECTIVE: Cognitive frailty, defined by the coexistence of cognitive decline and physical frailty, has been clinically defined, but its biological clues are still vague. This underscores the need for promising blood-based molecular biomarkers. DESIGN: Cross-sectional observational study. SETTINGS AND PARTICIPANTS: Frailty was diagnosed using the Japanese version of the Cardiovascular Health Study (J-CHS), and mild cognitive impairment was assessed with the Japanese version of the Montreal Cognitive Assessment (MoCA-J) and Mini-Mental State Examination-Japanese (MMSE-J). Participants with MMSE-J ≥24, MoCA-J score ≤25, and J-CHS score ≥1 were classified as having cognitive frailty. This study included 87 older adults aged ≥65 years, comprising 44 robust and 43 with cognitive frailty. MEASUREMENTS: Blood samples and associated clinical data were obtained from the National Center for Geriatrics and Gerontology Biobank in Japan. A multi-omics analysis integrating clinical data, RNA-seq, aging-related factors, and metabolomics were conducted to identify potential biomarkers through logistic regression, adjusting for age, sex, and body mass index (BMI). An optimal set of biomarkers was determined by constructing prediction models using the random forest algorithm. RESULTS: Three candidate biomarkers were identified from aging-related factors-growth differentiation factor (GDF15), brain-derived neurotrophic factor (BDNF), and Adiponectin-and three from metabolomics-myristic acid, nicotinamide, and γ-butyrobetaine. Using combinations of these candidates with clinical variables, we constructed risk prediction models. The best model incorporated one aging-related factors (GDF15) and two metabolites (myristic acid, and nicotinamide), achieving a high area under the receiver operating characteristic curve (AUC) of 0.96 in an independent validation cohort. This was significantly higher than models based solely on clinical information (age, sex, and BMI) (Welch's t-test, p <0.001). Among these biomarkers, myristic acid showed the highest influence, with a median Gini importance of 0.38 (95% confidence interval: 0.29-0.47). CONCLUSIONS: We identified three promising biomarkers-GDF15, myristic acid, and nicotinamide-for cognitive frailty. Notably, low plasma myristic acid levels emerged as the most significant contributor to the prediction model. Further refinement and large-scale validation will be essential to support its future clinical application.