Abstract:

Objective To discuss the construction method of phenotypic age （PA） models based on different combinations of multiple clinical biomarkers， and to clarify the influencing factors of PA acceleration. Methods Using the National Health and Nutrition Examination Survey （NHANES） database， the chronological age（CA） and the measurement results of nine clinical biomarkers including albumin（ALB）， alkaline phosphatase（ALP）， creatinine（Cr）， serum glucose（Glu）， C-reactive protein（CRP）， white blood cell count（WBC）， lymphocyte percent（LYM%）， mean cell volume（MCV）， and red cell distribution width（RDW） were collected from NHANES Ⅲ （1988-1994） and NHANES Ⅳ （1999-2018）. In the NHANES Ⅲ dataset， different subset combinations of the nine clinical biomarkers were constructed， and PA models were built and trained for each combination. The trained models were applied in the NHANES Ⅳ dataset to calculate PA and PA acceleration. Pearson correlation analysis and intraclass correlation coefficient（ICC） were used to assess the correlation and consistency between PA of each combination and PA of the full-indicator combination model. A weighted linear regression model was used to analyze the influencing factors of PA acceleration. An interactive website for dynamic PA assessment was established using HBuilder software based on hyper text markup language（HTML）， cascading style sheets（CSS）， and JavaScript technologies. Results A total of 502 combinations of the nine biomarkers were generated， and 324 high-quality combinations were identified. Among them， 142 high-quality combinations had correlation coefficients and ICC greater than 0.990 0. The weighted linear regression model results showed that female and higher education level were protective factors for PA acceleration （β=-0.85， 95%CI： -0.97--0.73； β=-1.14， 95%CI：-1.37--0.90）， while higher body mass index（BMI） and smoking were the risk factors for PA acceleration （β=2.45， 95%CI： 2.26-2.64； β=1.09， 95%CI： 0.94-1.24）. Other Hispanic individuals had significantly higher PA acceleration levels compared with Mexican Americans （β=0.84， 95%CI： 0.55-1.12）. Conclusion PA models based on different combinations of multiple clinical biomarkers have been successfully constructed， and an interactive website for dynamic PA assessment has been established， providing an effective tool for individual aging assessment. Higher BMI and smoking are the risk factors that accelerate biological aging of the body.

Key words: Biological age, Phenotypic age, Biomarker, Aging, Risk factor