Study population
This was a community-based cross-sectional study. The study sample was derived from participants in the baseline examination of the randomized controlled Multimodal INterventions to delay Dementia and disability in rural China (MIND-China), which is part of the World-Wide FINGERS Network [26, 27]. In brief, MIND-China targets all registered residents who were aged ≥60 years by the end of 2017 and living in the 52 villages of Yanlou Town, Yanggu County, western Shandong Province. In March-September 2018, a total number of 5,765 residents (74.9% of all eligible persons) participated in the baseline examination. In August 2018-December 2020, a subsample of 2,505 participants in MIND-China underwent the ActiGraph examination. Of these, we excluded 404 participants due to insufficient wearing time (<3 valid days) of ActiGraph (n=301) and dementia, severe mental diseases or major depressive disorders (n=103). Of the remaining 2,101 dementia-free participants, data on plasma NfL were available in a subsample of 1,029 persons. Among these, 38 were excluded due to missing data on serum inflammatory biomarkers, leaving 991 persons for the analysis involving systemic low-grade inflammatory. Figure 1 shows flowchart of the study participants.
Insert Figure 1 here
ActiGraph data collection and processing
Participants were instructed to wear an ActiGraph wGT3X-BT triaxial accelerometer (ActiGraph, LLC, Pensacola, FL) on their hip, affixed to an elastic belt, during all waking hours for seven consecutive days, and to remove it only for bathing [28]. ActiGraph recorded data on counts per minute (CPM) and lasting time for each value of CPM. We analyzed data from the ActiGraph accelerometer for participants who wore the device for at least 10 hours per day for a minimum of 3 days [29]. We defined SB, LPA, and MVPA as <100, 100-1,040, and ≥1,041 CPM, respectively, according to previous reports [30]. Wearing seasons of accelerometer were categorized as spring, summer, autumn, and winter.
Plasma NfL and serum cytokines
EDTA plasma was collected, aliquoted, and stored at -80ºC according to standard procedures until further analysis. In the biomarker subsample, plasma NfL concentration was measured on a single molecule array (SIMOA) platform (Quanterix Corp, MA, USA) with the NF-light® advantage Kit following the manufacturer’s instructions (Wayen Biotechnologies Inc., Shanghai, China) [31]. Serum cytokines were measured with commercially available meso scale discovery (MSD) V-PLEX® Proinflammatory Panel, as previously described [32]. The proinflammatory panel included interferon gamma (IFN-γ), IL-6, IL-8, interleukin-10 (IL-10), TNF-α, interleukin-17A (IL-17A), eotaxin-3, monocyte chemotactic protein-1 (MCP-1), ICAM-1, and vascular cell adhesion molecule-1 (VCAM-1).
Measurement of covariables
Data on demographics, lifestyles, medical history, and use of medications were collected through face-to-face interview, clinical examinations, and laboratory tests [33]. All medications were classified and coded according to the Anatomical Therapeutic Chemical (ATC) Classification System [34]. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2) and obesity was defined as BMI ≥28 kg/m2 [35]. Education was categorized into illiterate (no formal schooling education), primary school, and middle school or above. Smoking and drinking status were categorized as never, former, and current. Hypertension was defined as systolic pressure ≥140 mmHg or diastolic pressure ≥90 mmHg or current use of antihypertensive medications (ATC codes C02, C03, and C07-C09), diabetes as self-reported history of diabetes diagnosed by a physician or fasting blood glucose ≥7.0 mmol/L or current use of blood glucose-lowering medications (ATC code A10), and dyslipidemia as total cholesterol (TC) ≥6.22 mmol/L or low-density lipoprotein cholesterol (LDL-C) ≥4.14 mmol/L or triglyceride (TG) ≥2.27 mmol/L or high-density lipoprotein cholesterol (HDL-C) <1.04 mmol/L or having received medications for high cholesterol or dyslipidemia (ATC code C10). Apolipoprotein E (APOE) genotyping was performed using multiple-polymerase chain reaction amplification, and APOE genotype was dichotomized into carriers vs. non-carriers of the APOE ε4 allele.
Statistical analysis
Characteristics of the study participants were presented with frequencies (%) for categorical variables and mean (standard deviation, SD) or median (interquartile range, IQR) for continuous variables depending on their distributions. We compared characteristics of the study participants by sex using the chi-square test for categorical variables, t-test for continuous variables with normal distribution, and Kruskal-Wallis test for those with skewed distribution.
Plasma NfL concentration was log-transformed before model fitting owing to skewed distribution. The association patterns of SB, LPA, and MVPA with plasma NfL were examined using restricted cubic spline (RCS) analysis. We ran the RCS models with 3 knots placed at the 10th, 50th, and 90th percentiles to test if daily SB, LPA, and MVPA were linearly associated with plasma NfL. If RCS analysis suggested a linear relationship, daily SB, LPA, and MVPA were analyzed as continuous variables using the general linear regression models. When the non-linear association was detected, we plotted the dose-response trajectories in RCS analysis for each outcome. Inflection points were defined as the number (approximately equal to an integer) of daily hours of SB, LPA, and MVPA at which the relationships with plasma NfL changes [36]. Then, we further analyzed the association of SB, LPA, and MVPA with plasma NfL after grouping the participants according to the respective inflection points. We reported the main results from three models: Model 1 was adjusted for age, sex, education, and ActiGraph wear season and daily wear time; Model 2 was additionally adjusted for APOE ε4 allele, smoking, alcohol intake, obesity, hypertension, dyslipidemia, diabetes, stroke, and coronary heart disease (CHD); and to assess the mutual impact of movement behaviors on the results, in Model 3, the analyses of MVPA and LPA were mutually adjusted and the analyses of SB were adjusted for MVPA.
A composite score of the low-grade inflammation markers was calculated according to predefined clusters of conceptually related biomarkers, including IL-6, IL-8, TNF-a, and ICAM-1 [37]. Specifically, measures of serum cytokines were screened for outliers (above mean plus 3 SDs), log-transformed due to skewed distributions of original data (except TNF-α and ICAM), and then converted to a standardized z-score. The individual cytokine z-scores of IL-6, IL-8, TNF-a, and ICAM-1 were averaged to yield a composite score for low-grade inflammation.
We used Stata Statistical Software, Release 15.0 (Stata Corp LLC., College Station, TX, USA) and R-3.6.3 for Windows (R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/) for all the analyses. Two-tailed P<0.05 was considered to be statistically significant