Transparency and Openness
The data were selected based on sample size estimation from the Healthy Aging Longitudinal Study in Taiwan (HALST), conducted by the National Health Research Institutes (ClinicalTrials.gov: NCT02677831), which focuses on healthy aging [13]. This study was approved by the institute's ethics committee (protocol number: EC0970608 and EC1020805). HALST recruited community-dwellers aged range 55 and above from seven living areas near the selected hospitals. In the first-wave (baseline; 2009 - 2013) of recruitment, 94% subjects completed interviews and hospital-based examinations [13]. Participants were deemed healthy if they did not have highly contagious diseases, severe illnesses (e.g., active cancer), physician-diagnosed dementia, bedridden status, severe mental disorders or cognitive impairment (Mini-Mental State Examination, MMSE score <16), mental retardation, or severe hearing loss, and were not hospitalized or institutionalized at recruitment.
Participants, assessment of frailty, matching criteria
A total of 288 subjects, with an average age of 65 (Group 1), a total of 196 subjects with an average age of 71 (Group 2), and a total of 80 subjects with an average age of 79 (Group 3), were selected from HALST. This study first selected subjects from the HALST database depending on the frailty status. Frailty status was categorized as either frail or robust based on the presence (≥1 score) or absence of frailty phenotype components. Frail cases were those classified as frail in both Group 1 and Group 2 and average aged 65-71. In contrast, Group 3 was selected according to slow walking speed, and weak grip strength. Age, sex, metabolic syndromes, education level, and the residency area were recorded among the three groups. There were 115 frailty in Group 1, 79 frailty in Group 2, and 32 frailty in Group 3 (Fig. 1).
Global cognitive function was assessed using the MMSE score. Frailty status was evaluated with the Fried frailty phenotype, including unintentional weight loss (> 4.5kg in the past year), weak grip strength (lowest 20% of sex- and height-adjusted values), self-reported exhaustion (CES-D scale), slow walking speed (slowest 20% of sex- and BMI-adjusted values), and low physical activity (lowest 20% of sex-specific physical activity, < 105.2 kcal/week for men and < 46.1 kcal/week for women) [8]. Each component contributed one point. Scores of 1-3 or more indicated frailty, and 0 indicated robustness.
Blood pressure measurement
Participants emptied their bladders before measurements. They provided information on recent smoking of the last cigarette timing, exercise in the past 2 weeks, hypertension medication use. After rest for five minutes, three consecutive blood pressure (BP) measurements were taken, with one-minute intervals. With the first reading discarded, the average of the second and third readings represented the BP levels, including systolic (SBP) and diastolic BP (DBP). Pulse pressure (PP), calculated as the difference between SBP and DBP, and mean arterial pressure (MAP), calculated as 1/3(SBP) + 2/3(DBP), were also determined [14].
Biochemistry measurement
Blood samples were taken after 8 hours of fasting. All blood samples were centrifuged for plasma or serum collection, and stored in a -80℃ refrigerator for biochemical factors analyses [13].
The serum levels of low-density and high-density lipoprotein cholesterol (LDL-C and HDL-C) and triglycerides were determined on an ADVIA XPT chemical analyzer by Enzymatic GPO method (Triglycerides) and Elimination/ catalase principle (HDL-C and LDL-C). The Siemens ADVIA XPT system was used to measure glucose, high sensitivity C-reactive protein (hsCRP) and creatinine. Serum hsCRP were assayed by immunoturbidimetry method and creatinine levels were assayed by Jaffe reaction with rate-blanked method. Hemoglobin levels were determined by the Sysmex XN9000 system utilizing the Laser Flow Cytometry method. HbA1c levels were detected using the Bio-Rad D-100, a high-throughput high-performance liquid chromatography system.
Plasma levels of IGF-1 and TNFR1 were measured using enzyme-linked immunosorbent assays with the Human ELISA kit from R&D Systems, Inc. The inter- and intra-assay coefficients of variation (CVs) of IGF-1 were 2.75%–12.84% and 6.38%, respectively. The inter- and intra-assay CVs of TNFR1 were 2.12%–14.25% and 8.42%, respectively.
Statistical analysis
Characteristics of data at Group 1, Group 2, and Group 3 were reported by continuous and categorical variables. Group comparisons for continuous variables were performed using the Kruskal-Wallis test, followed by pairwise comparisons using the Mann-Whitney U test, where appropriate. For categorical variables, either the Chi-square test or Fisher’s exact test was applied. Following these group comparisons, we further examined the differences in robustness and frailty within each group, utilizing the same statistical methods described above (Mann-Whitney U test for continuous variables and Chi-square or Fisher’s exact test for categorical variables). Multiple regression models were employed to examine the associations between independent variables, specifically metabolic and biochemical indicators, and the dependent variables, including the MMSE and TSF. Variables that demonstrated a p-value of less than 0.05 in the simple regression analyses were considered potential confounders and were subsequently included in the multiple regression models. After the multiple regression, we further used the five components of TSF as predictors for MMSE and TNFR1. Given that the MMSE, TSF, and TNFR1 were not normally distributed, a rank transformation was applied in the regression analysis to account for this non-normality [15]. All analyses were conducted using SAS software, Version 9.4 (SAS Institute, Inc., Cary, NC), and the figure was created with BioRender.com.