The Kailuan Study (Registration Number: ChiCTR-TNC-11001489) was a community-based cohort study. Participants were recruited from Kailuan communities (Tangshan City, northern China) and follow-up information was obtained biennially from 2010 to 2016 (17). A total of 5,182 participants were involved in the final analyses after excluding those who did not complete a minimum of three assay repetitions of baPWV (n = 30,927); those with history of stroke, myocardial infarction, heart failure, or tumor (n = 100); and those with atrial fibrillation and an ankle-brachial index ≤0.9 (n = 160). This study was conducted according to the guidelines of the Helsinki Declaration and was approved by the Ethics Committee of Kailuan General Hospital. Signed informed consent was obtained from all participants.

Clinical examination and questionnaire baseline data were collected by trained research staff according to standard operating procedures. The questionnaire included items on participant age, sex, smoking and drinking history, physical activity, current prescription medication, and medical history. Clinical examination included measurements of weight, height, waist circumference, resting heart rate, systolic blood pressure (SBP), and diastolic blood pressure (DBP). BMI was calculated as weight (kg)/height (m²) (18). Blood pressure and resting heart rate were measured via an Omron® electronic automated blood pressure monitor. Hypertension was defined as a SBP ≥140 mmHg or a DBP ≥90 mmHg, a history of hypertension, or treatment with antihypertensive medications (19). Diabetes mellitus was a self-reported diagnosis, a fasting blood glucose (FBG) level of ≥7.0 mmol/L (126 mg/dl), or use of antidiabetic agents (20).

Blood samples of all participants were collected by venipuncture. FBG, low-density lipoprotein cholesterol (LDL-C), serum uric acid, and high-sensitivity C-reactive protein (Hs-CRP) were determined by the Friedewald formula (11). The estimated glomerular filtration rate (eGFR) was calculated according to the Chronic Kidney Disease Epidemiology Collaboration formula (21).

BaPWV was measured in some participants from 2010 (the third follow-up), and the baPWV of the study population was repeatedly assessed during each follow-up. It was measured by an Omron Colin BP-203RPE III device (Omron Health Care, China). Before the measurement, the participants were instructed not to smoke and had to rest for at least 5 min. The temperature of the examination room was maintained at 22–25°C. The participant was placed in a supine position, and cuffs were wrapped on both arms and ankles. The heart sound collection device was placed in the pre-heart area. The electrocardiogram (ECG) collection device was clamped on the left and right wrists and connected to the plethysmography sensor and the oscillation pressure sensor. We repeated the measurement twice for each participant, and recorded the second data as the final result; we used the larger value of baPWV on the left and right sides for analysis. We categorized the levels of baPWV into three grades: low (<14.0 m/s), intermediate (14.0–17.9 m/s), and high (≥18.0 m/s) (22).

The trajectories of baPWV were identified using PROC TRAJ in SAS® v.9.4 (SAS Institute Inc, Cary, NC, United States) (23). The best typing model was selected based on the lowest Bayesian information criterion. After the typing was determined, the highest power of each trajectory curve was determined according to the probability (>70%) that the observed object meets the trajectory curve. The Kolmogorov-Smirnov test was used to test for a normal distribution of continuous variables. Data with a normal distribution were expressed as mean ± standard deviation and analyzed by one-way analysis of variance. Categorical data were presented as frequencies and percentages (%), and differences between types were evaluated by χ² test.

Multivariate regression analysis was used to evaluate the association of baPWV trajectory with baseline data. We adjusted covariates including age, sex, mean arterial pressure (MAP, computed as 1/3 SBP + 2/3 DBP), resting heart rate, BMI, waist circumference, FBG, LDL-C, Hs-CRP, uric acid, and eGFR. Multiple logistic regression analysis was used to explore factors associated with baPWV longitudinal trajectory. The forward stepwise method was used for multiple logistic regression (enter probability: 0.05, remove probability: 0.10). In addition, we performed a multiple analysis with adjustment of the following variables: Δ MAP, Δ resting heart rate, Δ BMI, Δ waist circumference, Δ FBG, ΔLDL-C, ΔHs-CRP, Δ uric acid, and Δ eGFR to assess their effect on the baPWV longitudinal trajectory during the follow-up. A likelihood ratio test was used to assess the significance of the model fit. The effectiveness of the model was estimated with the Nagelkerke pseudo-R² statistic (a measure of explained variation in the model). Sensitivity analysis was performed based on baseline antihypertensive drugs, new-onset hypertension, diabetes, and participants with cardiovascular disease during the follow-up period. A two-sided p < 0.05 was considered statistically significant. Statistical analyses were performed using SAS® v.9.4 (SAS Institute Inc).

The study included totally 5,182 participants, of whom 2,445 (47.18%) were male. The average age was 46.87 ± 9.57 years at baseline. The median follow-up year is 5.66 years (Interquartile range, 4.12–6.52 years). Supplementary Table S1 summarizes the baseline and follow-up characteristics of the participants. Compared with baseline levels, the following variables significantly increased (p < 0.001): MAP, resting heart rate, LDL-C, BMI, waist circumference, FBG, and uric acid. The prevalence of hypertension and diabetes also increased. More participants during follow-up received antihypertensive medications, anti-diabetic medications, or lipid-lowering medications. In contrast, the proportion of physical exercisers, smokers, and alcohol drinkers decreased. All the above trends were statistically significant (p < 0.001). A comparison between the baseline characteristics of the included participants and the general population who only underwent a single baPWV measurement is shown in Supplementary Table S2. Compared with the general population, the participants tended to have lower levels of MAP, resting heart rate, waist circumference, BMI, LDL-C, FBG, Hs-CRP, and uric acid, with a relatively elevated eGFR level. In addition, there was a lower prevalence of high/low-density lipoproteinemia and diabetes in the study population, and relatively fewer smokers, but more alcohol drinkers.

A total of 17,331 baPWV measurements were performed among 5,182 participants, with a mean of 3.34 ± 0.48 times for each. The annual growth rate of baPWV was 0.14 ± 0.40 m/s/y. Four baPWV trajectories were identified, including a low group, medium-low group, medium-high group, and high group. In the low group (1,961 cases, 37.84%), the mean baPWV increased from 11.70 to 12.24 m/s, with an increase of 0.54 m/s and an annual growth rate of 0.10 m/s/y. In the medium-low group (1,846 cases, 35.62%), the mean baPWV increased from 14.07 to 14.94 m/s, with an increase of 0.87 m/s and an annual growth rate of 0.16 m/s/y. In the medium-high group (1,024 cases, 19.76%), the mean baPWV increased from 16.53 to 17.75 m/s, with an increase of 1.22 m/s and an annual growth rate of 0.21 m/s/y. In the high group (351 cases, 6.77%), the mean value of baPWV increased from 19.90 to 21.09 m/s, with an increase of 1.19 m/s and an annual growth rate of 0.15 m/s/y. The probability of each type (low group, medium-low group, medium-high group, and high group) meeting the trajectory curve was 90.72%, 84.86%, 87.33%, and 91.93%, respectively (Figure 1 and Table 1). The trajectories of males and females were consistent with that of the total participants. The initial baPWV of each trajectory type in females was lower than that in males, while there was no difference observed in the baPWV annual growth rate (0.14 m/s/y) (Supplementary Figure S1 and Tables S3, S4).

A comparison of the baseline characteristics among different baPWV trajectory groups is shown in Table 2. With an increasing baPWV longitudinal trajectory level, the following covariates showed an upward trend: age; male proportion; SBP; DBP; MAP; resting heart rate; BMI; waist circumference; LDL-C; FBG; Hs-CRP; uric acid; proportion of smokers, alcohol drinkers, and physical exercisers; and participants with history of hypertension and diabetes, and antihypertensive, anti-diabetic, and lipid-lowering medications. In contrast, the level of eGFR decreased. All the above trends were statistically significant (p < 0.001). Changes in the main independent variable parameters among different trajectory groups during follow-up are presented in Table 3. Δ FBG and Δ uric acid significantly increased (p < 0.01), and the percentage of participants with new-onset hypertension, diabetes, and cardiovascular disease also increased (p < 0.001). However, levels of Δ BMI, Δ waist circumference, Δ LDL-C, and Δ eGFR significantly decreased during follow-up (p < 0.001).

We also explored the factors relevant to the baPWV longitudinal trajectory. Compared to the low trajectory group, older participants, males, alcohol drinkers, and those with higher levels of MAP, resting heart rate, FBG, LDL-C, Hs-CRP, and uric acid tended to have a higher risk of inclusion in the high trajectory groups. We further investigated the effect of changes in the above-mentioned variables on the baPWV longitudinal trajectory during the follow-up. After fully adjusting, baseline MAP, resting heart rate, FBG, and uric acid were positively associated with the baPWV trajectory. In addition, the participants with increased ΔMAP, Δ resting heart rate, ΔFBG, and Δ uric acid had an increased risk of inclusion into the higher trajectory group. The fitting degree of the model increased by 5.1%–59.1% after adjustment (Table 4).

The 54% variation of the baPWV trajectory was interpreted by baseline data: 33.5% by MAP, 16.4% by age, 1.3% by sex, and 1.3% by resting heart rate. The rest of the variation was interpreted by a combination of FBG, uric acid, Hs-CRP, LDL-C, and drinking. After adding the adjusted value of the independent variable, the interpretation degree of the variation of the baPWV trajectory increased by 5.1%, of which 49.9% was still interpreted by MAP (33.5%) and age (16.4%), while the rest was interpreted by ΔMAP (5.4%), resting heart rate (1.2%), and other variables (Supplementary Table S5). In addition, the effect relationship between age and baPWV was found by restricted cubic spline plot (RCS). A positive correlation exists between age and baPWV, and the baPWV value gradually increases with age (Supplementary Figure S3).

A sensitivity analysis was conducted on baseline antihypertensive medication, new-onset hypertension, new-onset diabetes, and new-onset CVD population. We found that advanced age; male sex; high levels of MAP, resting heart rate, FBG, LDL-C, and uric acid; low BMI; and increased MAP, resting heart rate, FBG, and uric acid remained risk factors for high baPWV level (Supplementary Table S6).