This study was a subset of the China Hypertension Registry Study (Registration website: http://www.chictr.org.cn/, Number: ChiCTR1800017274, Date: July 20, 2018), a real-world, prospective, and observational study conducted in Wuyuan, Jiangxi Province, China, from March 2018 to August 2018. The methodology for data acquisition and exclusion criteria has been previously described (16, 17). This study adhered to the Declaration of Helsinki and was approved by the Ethics Committee of the Second Affiliated Hospital of Nanchang University (Ethics No. 2018019) and the Institute of Biomedical Sciences, Anhui Medical University (Ethics No. CH1059). Written informed consent was obtained from all enrolled participants.

From the initial cohort of 14,234 hypertensive individuals recruited at baseline, we sequentially excluded individuals lacking baPWV values (n = 5,965) and those with ankle-brachial index (ABI) <0.90 (n = 253), ABI > 1.4 (n = 5), stroke (n = 552), atrial fibrillation (n = 259) and traumatic subdural hemorrhage (n = 3). Finally, this prospective study included 7,198 participants for analysis. The selection process for the analytic sample is presented in Figure 1.

Based on the existing literature and clinical applications, covariates obtained from physical examinations, standard questionnaires, and laboratory measurements were selected as potential confounders. Physical examinations and standard questionnaires were administered by trained staff following a standard operating procedure. A fully adjusted model was developed using the following covariates. Categorical variables included gender, current smoking, current drinking, sleep duration, previous medical diagnoses (hypertension, dyslipidemia, diabetes, coronary heart disease, heart failure, atrial fibrillation, and stroke), and current medication based on drug packaging (antihypertensive, lipoprotein-lowering, glucose-lowering, and antiplatelet drugs). Continuous variables included age, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), homocysteine (Hcy), total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C). The BMI was calculated as weight (kg)/height (m²). Blood pressure was measured thrice using an electronic sphygmomanometer on the right arm positioned at the heart level after a 5-min rest, with a 30-s interval between measurements. The average of the three measurements was used.

The automatic device of Omron-Colin BP-203RPE (Omron Health Care, Japan) was used to measure simultaneously baPWV and ankle-brachial index (ABI). After resting in the supine position for more than 5 min, four cuffs were wrapped around bilateral brachia and ankle arteries of participants, and connected to a plethysmographic sensor and oscillometric pressure sensor. Pulse volume waveform were recorded using semiconductor pressure sensors to assess the time interval between the initial increase in brachial and tibial waves.

The ABI value of each leg was calculated by the ankle SBP divided by the ipsilateral brachial SBP, and the lower value in two legs was used in the analysis. The baPWV was calculated by (La-Lb)/Tba. La is the path distance between the suprasternal notch and the ankle, Lb is the path distance between the suprasternal notch and the brachium, and Tba is the time interval between the brachial and ankle waveform. The larger values of baPWV on the left and right sides were used for the main analysis (18), and the average values on both sides were used for a sensitivity analysis.

Details on the definition and event adjudication were sourced from the literature (19). Stroke event identification followed a rigorous, multistep protocol. First, potential cases were preliminarily identified using retrospective patient interviews, medical record reviews (including hospitalizations and emergency department visits), and linkage with the national health insurance database. Subsequently, all suspected cases were systematically verified by retrieving and analyzing neuroimaging records (for instance, CT/MRI scans). Finally, confirmed cases were reviewed and adjudicated through independent discussion by an Endpoint Adjudication Committee comprising neurologists, neurosurgeons, and public health experts.

The primary outcome was the first occurrence of fatal or nonfatal symptomatic stroke (ischemic or hemorrhagic), excluding subarachnoid hemorrhage and silent strokes. Participants with preexisting stroke at baseline were excluded. Recurrent strokes were documented but not considered primary outcomes. The follow-up period was extended from the baseline survey completion date to August 15, 2022.

Patients with a history of stroke at baseline were excluded. The first attack of symptomatic stroke was considered the first stroke in the patient unless there was evidence against it. Any stroke after the first attack was considered a recurrent stroke, which was not the primary outcome of this study.

Continuous variables are presented as mean ± standard deviation (SD), and categorical variables are presented as frequencies and percentages (%). Differences in baseline characteristics according to baPWV categories were compared using a one-way analysis of variance for continuous variables and a chi-square test for categorical variables. Cox proportional hazard regression was used to evaluate the HRs and 95% CIs for the association between baPWV and first stroke, with adjustment for potential covariates in the three models. The proportional hazards assumption was evaluated using Schoenfeld residual tests, which revealed no violations (P > 0.05), indicating that the assumption of proportionality was satisfied. Potential covariates adjusted in the models were included due to their clinical importance, statistical significance in the univariable analysis, and the potential confounder effect estimates being individually changed by at least 10%. The cumulative hazards of the first stroke by baPWV categories were estimated using the Kaplan–Meier curve, and differences between groups were compared using the log-rank test. The generalized additive model and a restricted cubic spline were used to characterize the dose-response association of baPWV with the first stroke. Additionally, potential modifications to the association between baPWV and first stroke were evaluated using subgroup analyses.

A two-tailed P < 0.05 was considered statistically significant. The R statistical package (version 4.2.3; http://www.r-proje.ct.org) and Empower (R) software (version 4.1; http://www.empow.erstats.com) were used for statistical analyses.

This study included 7,198 individuals with hypertension [mean age, 63.7 (9.7) years], and 47.1% of them were male. The mean ± SD value of baseline baPWV was 18.2 ± 5.4 m/s. The baseline characteristics of all participants stratified by the baPWV quartile are listed in Table 1. Participants with a higher baPWV were more likely to be female and have a higher age, SBP, DBP, Hcy, TC, HDL-C, and LDL-C levels, and a lower BMI. Additionally, they exhibited a longer sleep duration of <5 and >8 h, use of antihypertensive and glucose-lowering drugs, a history of diabetes and congenital heart defect (CHD), a lower rate of current smoking and current drinking, and a history of dyslipidemia (P < 0.05).

During an average follow-up period of four years, 281 cases were identified with the first stroke. The independent effects of baPWV on the first stroke are illustrated in Table 2. After adjusting for potential confounders in model 3, each 1 m/s increment of baPWV was associated with a 2% increase in the risk of the first stroke (HR: 1.02; 95% CI: 1.01, 1.04). When baPWV was measured in quartiles, the HRs (95% CIs) of the first stroke for participants in quartiles 1, 2, and 4 were 0.39 (0.26, 0.59), 0.69 (0.50, 0.96), and 1.00 (0.74, 1.33) respectively, compared with those in quartile 3 (P for trend = 0.810) (Table 2).

The smoothing curve fitting revealed a saturation effect of baPWV on the first stroke (Figure 2). The saturation effect analysis revealed that the inflection point of baPWV was 17.5 m/s. For baPWV < 17.5 m/s, the adjusted HR (95% CI) was 1.31 (1.17, 1.47), and the adjusted HR (95% CI) was 0.99 (0.96, 1.02) for baPWV ≥ 17.5 m/s (P for log-likelihood ratio test < 0.001) (Table 3). In addition, the Kaplan–Meier curves revealed a consistent increase in cumulative hazard for the first stroke from quartile 1 to quartile 3 of baPWV (log-rank P < 0.001), whereas there was a non-significant difference in cumulative hazard between quartiles 3 and 4 (log-rank P = 0.873) (Figure 3). To verify the robustness of our findings, we also conducted a sensitivity analysis using the average of both sides, and the conclusions remained consistent with the main analysis (Supplementary Table S1 and Table S2).

To determine the effect of covariates on the association between baPWV (per 1 m/s increment) and first stroke, stratified analyses were performed in two groups of participants separated by the inflection point of baPWV (17.5 m/s). The stratified analysis results revealed that the association of baPWV with the first stroke in various subgroups was consistent with the findings of the saturation effect analysis (Figure 4). In both the baPWV < 17.5 m/s and the baPWV ≥ 17.5 m/s groups, there were no significant interactions in the following subgroups: gender (male vs. Female), age (60 vs. ≥ 60 years), BMI (<24 vs. ≥ 24 kg/m²), current smokers (no vs. yes), hcy (15 vs. ≥ 15 mmol/L), current smoking (no vs. yes), current drinking (no vs. yes), diabetes (no vs. yes), CHD (no vs. yes), heart failure (no vs. yes), and dyslipidemia (no vs. yes) (P > 0.05 for interaction).