Introduction

Atherosclerotic cardiovascular disease (ASCVD) has shown increasing rates of morbidity and mortality worldwide in recent decades [1]. Early detection and risk-based intervention strategies are crucial for secondary prevention of ASCVD. Arterial stiffness, which increases with aging, is an important determinant of cardiovascular (CV) health, and it is an independent predictor of risk for future cardiovascular events [2,3,4].

Pulse wave velocity (PWV) is commonly used as an effective noninvasive indicator for assessing arterial stiffness. The predictive value of aortic PWV for CV events and mortality has been verified in a variety of clinical scenarios [5, 6]. Brachial-ankle pulse wave velocity (baPWV) is widely used in Asia due to its advantages of reproducibility, simplicity, and good correlation with aortic PWV [7]. High baPWV is associated with an increased risk of CV events or all-cause mortality in different populations, such as patients with CV disease risk [8], hypertension [9], diabetes mellitus [10], and the general population [11]. Although some relevant studies [4, 12,13,14,15,16,17,18,19,20,21,22,23,24,25,26] on the association between baPWV and the prognosis of established ASCVD have been published, the results of the risk estimates have not been consistent, and no overall quantitative estimate of the association exists. Therefore, we conducted the present systematic review and meta-analysis to provide an overview of these studies and to quantitatively evaluate the association between baPWV and different outcomes, including CV events, CV mortality, and all-cause mortality in ASCVD patients. We also investigated risk estimates separately in patients with coronary artery disease (CAD) and stroke.

Methods

The review protocol was registered prospectively in PROSPERO (Registration number PROSPERO: CRD42020181378), and this meta-analysis was performed in compliance with the checklist of the Meta-Analysis of Observational Studies in Epidemiology [27].

Data sources and search strategy

The PubMed, EMBASE, and Cochrane Library databases were searched with English language restriction from inception to 24 April, 2020 via a combination of free-text terms and subject terms. The search strategy is available in the Supplementary Materials. Reference lists of the retrieved studies were manually screened for additional relevant literature.

Study selection and outcomes

The following inclusion criteria were applied. (1) Cohort studies performed in patients with at least 1 year of follow-up; (2) the study subjects were patients with ASCVD who were aged ≥18 years; (3) arterial stiffness was assessed by measurement of baPWV; and (4) the study included relevant outcome data, including a composite of CV events, CV mortality, or all-cause mortality. For multiple studies published from the same population cohort, only the study with the longest follow-up period or the largest cohort size was included. The details are shown in the Supplementary Materials. Case reports, reviews, letters, congress reports’ abstracts, and studies that comprised overlapping data were excluded. The outcomes of interest were as follows: (1) CV events; (2) CV mortality; and (3) all-cause mortality.

Data extraction and quality assessment

Study selection and data extraction were performed independently by two investigators. Divergences were resolved by consensus. Cohort characteristics, including first authors, publication year, study design, study location, participant characteristics (subtype of ASCVD, sample size, age, percentage of men, CV risk factors), baPWV measurements, baPWV cutoff, follow-up duration, endpoints, adjustment factors, effect size, and 95% confidence intervals (CIs), were extracted according to a standardized data extraction form. Data were extracted by the software Engauge Digitizer from Kaplan–Meier curves if they were not shown directly in the articles. All data were imported into an Excel file for subsequent data synthesis.

We assessed the quality of eligible studies on the basis of the Newcastle-Ottawa Scale (NOS) for cohort studies [28]. This scale is composed of three aspects (selection of participants, comparability of study groups, and outcomes) with a maximum of nine stars. We identified studies as being of good, fair, or poor quality when they were graded as seven to nine stars, four to six stars, and zero to three stars, respectively.

Data synthesis and statistical analysis

Summarized effect sizes with 95% CIs were calculated using hazard ratios (HRs) and 95% CIs for clinical events associated with baPWV. HRs and 95% CIs were transformed into the natural logarithm and standard error, respectively. Because there were no available uniform cutoff values for baPWV, patients were classified into the high baPWV group and low baPWV group on the basis of cutoffs provided by each study (e.g., median, highest quartile, or optimal cutoff derived by receiver-operator characteristic [ROC] curve analysis). In the studies using baPWV as a continuous variable, HRs were uniformly translated into HRs per standard deviation (SD) increase in baPWV if the SD was available.

We assessed heterogeneity using the I² statistic, which was classified as low (I² value of 25% to <50%), moderate (I² value of 50% to <75%), or high (I² value ≥75%) heterogeneity [29]. If statistical heterogeneity among the studies was observed, the DerSimonian and Laird random-effect model was used for the effect estimates. To investigate the underlying cause of heterogeneity, subgroup analyses were performed on the basis of follow-up duration, determination methods for cutoff values of baPWV, subtypes of ASCVD, adjustment for age and hypertension (or systolic blood pressure [SBP]), definition of outcomes and NOS score. In addition, sensitivity analysis was conducted to ascertain the robustness of pooled estimates. Due to the high impact of age on baPWV, we evaluated moderator effects of average age via subgroup analysis in the main analysis.

Because <10 studies were included for different clinical outcomes of the meta-analysis, funnel plots were not created. We performed Egger’s test (linear regression method) [30] and Begg’s test (the adjusted rank correlation method) [31] to evaluate the possibility of publication bias.

All analyses were performed using Stata software version 15.1 (StataCorp, College Station, TX, USA) and RevMan 5.3 (The Cochrane Collaboration, Copenhagen, Denmark). A p < 0.05 was considered statistically significant.

Results

Literature search

The literature selection process is shown in Fig. 1. Briefly, using the literature search strategy, a total of 1126 publications were retrieved. After screening the title and abstract, 70 duplicates and 1024 irrelevant articles were excluded. Of the remaining 32 studies that underwent full-text evaluation, 16 were excluded because they did not meet the inclusion criteria (n = 6), had incomplete data (n = 6), or had duplicate datasets (n = 4). A total of 16 studies [4, 12,13,14,15,16,17,18,19,20,21,22,23,24,25,26] were eligible for systematic review. In the six studies using baPWV as a continuous variable [4, 16,17,18,19, 25], considering that the variation in CV event rates corresponding to a 1 cm/s increase in baPWV was expected to be minuscule, HRs were uniformly converted into HRs per 1 SD increase in baPWV, except that for one article [4] in which HR per 1 cm/s increase of baPWV could not be converted because the SD was not available. Finally, 15 studies were included in further meta-analysis.

Fig. 1
figure 1

PRISMA flow diagram of the study

Full size image

Study characteristics

Details of the study characteristics are summarized in Table 1. All of the studies were hospital-based cohort studies, and were published since 2005, from the Asian population (Japan and Korea), and had a mean/median follow-up duration of 348 days [26] to 5.9 years [22]. Patients with CAD or stroke were included in the meta-analysis. The included studies involved a total of 6968 participants, with study sample sizes ranging from 91 [19] to 1765 [15] individuals. The average age of participants in most studies was in the 60 s (ranging from 58 to 70.8 years), and more than half of the participants were men. Regarding the subtypes of ASCVD, the mean or median levels of baPWV in stroke patients (1997–2105 cm/s) were generally higher than those in CAD patients (1478–1730 cm/s).

Table 1 Characteristics of the studies included in the present meta-analysis
Full size table

There were minor differences in the definition of clinical outcomes among the 16 studies. HRs were used as effect sizes according to endpoint events. Cutoff values of baPWV as a categorical variable varied from 1619 [24] to 2856 cm/s [15]. The cutoff points of baPWV in CAD patients ranged from 1619 cm/s to 1730 cm/s, while those in stroke patients ranged between 2040 and 2856 cm/s. A total of 14 reported HRs (8 studies [12, 13, 20,21,22,23,24, 26] used baPWV as a categorical variable and 6 studies [4, 16,17,18,19, 25] used baPWV as a continuous variable) were reported for CV events (5203 individuals). The HR for CV events in one study [13] was estimated by the software Engauge Digitizer from Kaplan–Meier curves and the full text. The multiple-adjusted HRs for high versus low baPWV for CV death events (3896 individuals) and all-cause death events (2112 individuals) were pooled from four [14, 15, 21, 26] and three [13, 15, 22] articles, respectively. In addition, baPWV in one article [4] was reported as a continuous variable for the outcome of CV death. Therefore, this study was not included in the combined HR value for CV mortality risk associated with high baPWV.

Twelve articles were judged to be of good quality, and four were judged to be of fair quality based on the NOS score. Most of these studies lacked representativeness of the exposed cohort. The fair-quality studies [16, 17, 19, 23] did not control for age or other CV risk factors. The mean quality score was 7.25 (Table 2).

Table 2 Quality assessments: Newcastle-Ottawa scale for cohort studies
Full size table

CV events

HRs in eight studies were pooled for the risk estimate of CV events according to the categorical variable of baPWV. The random-effects model showed that high baPWV was significantly associated with an increased risk of CV events (HR: 2.55, 95% CI: 1.61–4.03, Fig. 2) in ASCVD patients. Sensitivity analysis suggested that the pooled HR remained robust when each study was removed by turn (Supplementary Fig. 1). In the analysis of publication bias, neither Begg’s test (p = 0.266) nor Egger’s test (p = 0.246) showed significant publication bias.

Fig. 2
figure 2

Forest plot of the association between high baPWV and CV events. Each box and its extending lines represent the HR and the 95% CI for individual studies, respectively. The overall HR and its 95% CI are presented as a diamond and its width, respectively

Full size image

Because of significant heterogeneity (I2 = 72.4%, p = 0.001), subgroup analyses were conducted according to follow-up duration, determination methods for cutoff values of baPWV, subtypes of ASCVD, adjustment for age and hypertension (or systolic blood pressure [SBP]), definition of outcomes and NOS score (Table 3). According to the follow-up duration, the pooled HR showed a stronger association in the >3 years of follow-up subgroup (HR: 3.36, 95% CI: 1.85–6.10) than in the <2 years of follow-up subgroup (HR: 1.41, 95% CI: 0.80–2.49). We also found that the pooled HRs varied according to the determination methods for cutoff values of baPWV, with a significantly higher HR in the subgroup of cutoff points by the ROC curve method (HR 3.78, 95% CI: 1.82–7.85) than that by the percentile method (HR 1.62, 95% CI: 1.19–2.19, p for interaction = 0.04). Significant heterogeneity (I2 = 76.6%) was observed in the subgroup of cutoff points by ROC curves. However, after excluding the study by Park et al. in 2017, whose mean follow-up duration was 11.6 months, heterogeneity represented by I2 among the other studies turned out to be 0%, and the pooled HR was 5.38 (95% CI: 3.54–8.17) (Supplementary Fig. 2). Therefore, we speculated that the source of heterogeneity among the studies may be the determination methods used for cutoff values of baPWV.

Table 3 Subgroup analyses of association between high baPWV and the risk of CV events in ASCVD
Full size table

The HR for the association between high baPWV and the risk of CV events was numerically higher in CAD patients (HR: 2.93, 95% CI: 1.65–5.18) than in stroke patients (HR: 1.58, 95% CI: 1.04–2.40), with no significant difference (p for interaction = 0.10). After adjustment for age and hypertension (or SBP), the association between high baPWV and the risk of CV events remained significant. Similar results were observed in the subgroup analyses stratified by definition of outcomes and NOS score.

We evaluated the moderation effects of average age. Average age was a significant moderator of the association of high baPWV with CV events. The impact of high baPWV on future CV events was more prominent in patients with an average age of <65 years (HR 4.88, 95% CI: 2.96–8.05) than in elderly patients (HR 2.03, 95% CI: 1.28–3.22, p for interaction = 0.01, Table 3).

When baPWV was analyzed as a continuous variable, five studies [16,17,18,19, 25] were included in the pooled HR calculation. The results showed a significant association between each 1-SD increase in baPWV and an increased risk of CV events (HR: 1.41, 95% CI: 1.24–1.60, Fig. 3). There was no significant heterogeneity among the studies.

Fig. 3
figure 3

Forest plot of the associations of baPWV as a continuous variable with CV events. Each box and its extending lines represent the HR and the 95% CI for individual studies, respectively. The pooled HR and its 95% CI are presented as a diamond and its width, respectively

Full size image

CV mortality and all-cause mortality

The pooled HR for the association of high baPWV with the risk of CV mortality was 2.66 (95% CI: 1.88–3.76, Fig. 4A), with no heterogeneity (I2 = 0%). In addition, there was no significant publication bias, as shown by Begg’s test (p = 0.308) and Egger’s test (p = 0.317).

Fig. 4
figure 4

Forest plot of the associations of high baPWV with CV mortality and all-cause mortality. A displays the overall HR for the association between high baPWV and CV mortality. B represents the overall HR for the association between high baPWV and all-cause mortality. Each box and its extending lines represent the HR and the 95% CI for individual studies, respectively. Each pooled HR and its 95% CI are presented as a diamond and its width, respectively

Full size image

For the analysis of all-cause mortality, significant heterogeneity (I2 = 48.1%) was not observed among the studies. The random-effects model showed that high baPWV was significantly associated with an increased risk of all-cause mortality in ASCVD patients (HR: 1.77, 95% CI: 1.09–2.87, Fig. 4B). There was no evidence of publication bias for the association between high baPWV and the risk of all-cause mortality (Begg’s test: p = 0.296, Egger’s test: p = 0.339).

Discussion

Little clinical evidence is available for risk stratification based on baPWV for the secondary prevention of ASCVD. The current meta-analysis, which involved 15 studies, showed that those with high baPWV had a 2.5 times higher risk for CV events, 2.6 times higher risk for CV mortality, and 1.7 times higher risk for all-cause mortality than patients with low baPWV. The strengths of this meta-analysis are the inclusion of a relatively large number of patients with ASCVD, adjustment for confounders, and subgroup analyses of CV events. These data provide evidence on the association of baPWV with poor prognosis in patients with ASCVD.

Increasing evidence has suggested the association of arterial stiffness measured by baPWV with the risk of CV diseases and mortality. A previous meta-analysis [32] reported that baPWV predicted the risk of total CV events and all-cause mortality in various populations (general population or populations with risk factors or disease). J-BAVEL [Japan Brachial-Ankle Pulse Wave Velocity Individual Participant Data (IPD) Meta-Analysis of Prospective Studies] [33] demonstrated that each SD increase in baPWV was associated with a 1.19-fold increase in the risk of future CV disease in subjects without cardiovascular disease, while our meta-analysis showed that a 1 SD increase in baPWV corresponded to a 41% increase in CV event risk in patients with CAD. The predictive ability of baPWV for the risk of CV events seems to be higher in patients with CAD than in the general population, which increases the probability of the application of baPWV in the management of secondary prevention of CAD.

Sensitivity analysis and subgroup analyses were conducted considering that the reliability of the results was affected, to some extent, by significant heterogeneity among the studies for the risk estimate of CV events. The results indicated that the difference in the definition of high baPWV may be one of the causes of heterogeneity in our meta-analysis. In addition, there was a more marked association of high baPWV with CV events in the subgroups of long-term (>3 years) follow-up duration and an average age of <65 years, which suggested that arterial wall stiffening may have a persistent and chronic effect on cardiovascular outcomes, and other cardiovascular risk factors attenuate the impacts of baPWV on the prognosis in elderly patients. Therefore, baPWV should be monitored and interventions should be implemented as early as possible for the secondary prevention of ASCVD.

In the present study, the mean/median levels of baPWV widely varied between studies with CAD patients and studies with stroke patients. This may be because stroke patients were older, and had higher systolic blood pressure, as well as more unreported risk factors, than the CAD patients. The differences in the cutoff values of high baPWV may be due to the different levels of baPWV in both groups. More risk factors with higher levels make baPWV insensitive to the identification of poor prognosis in stroke patients. BaPWV appeared to be a stronger predictor in patients with CAD than in patients with stroke, although no significant difference was observed in the interaction analysis, which might be affected by few included relevant studies or underrepresentation of selected populations. More studies are necessary to further draw a definitive conclusion.

The cutoff values of baPWV may depend on multiple factors, such as patient characteristics (age and prevalence of hypertension), disease states, definition of outcomes, and statistical methods. In our meta-analysis for CV events, the cutoff values of baPWV in patients with CAD were close to the levels proposed by the Physiological Diagnosis Criteria for Vascular Failure Committee [34], with gradient cutoffs (<1400 cm/s for normal, ≥1400 cm/s and <1800 cm/s for borderline, ≥1800 cm/s for abnormal). Due to a lack of adequate clinical evidence, the cutoff values of baPWV from the overall population may not be applicable to patients with stroke. Because of the higher prevalence of hypertension and poorer blood pressure control in patients with stroke [35,36,37,38], we speculated that the cutoff values of baPWV in patients with stroke might be higher than the current recommended threshold.

There are several possible mechanisms by which arterial stiffness predicts cardiovascular prognosis and death. Arterial stiffness plays a central role in the vascular aging process. Stiffened arteries could damage the cushioning function and increase the cardiac afterload, resulting in myocardial hypertrophy and left ventricular dysfunction [39]. Increased large artery stiffness amplifies pulse pressure, causing decreased coronary perfusion, cerebral microvascular impairment (lacunar infarctions and large white matter hyperintensities), and other target organ damage [40]. The interplay between macro- and microcirculation ultimately contributes to CV events and even death.

Carotid–femoral pulse wave velocity (cfPWV), another well-recognized indicator of arterial stiffness in Western countries, is strongly associated with cardiovascular risk [41] and is highly correlated with baPWV [42]. In contrast to cfPWV, baPWV is measured over a longer arterial path length and is a combination index reflecting central and peripheral arterial stiffness [34]. A previous study showed that baPWV correlated better with parameters of LV and arterial structure and function than cfPWV [43]. Moreover, because of its confirmed predictive value in cardiovascular disease and mortality, as well as its simplicity and reproducibility [39], baPWV may be considered a valuable risk assessment tool in the management of ASCVD. The distribution of baPWV did not show significant ethnic differences [44, 45], which suggests that baPWV measurement might be applicable to a wider-ranging study population, rather than only Asian populations.

Strategies to reduce arterial stiffness warrant additional studies. Arteries become stiffer with increasing pathological states, such as hypertension, diabetes, and atherosclerosis. Therefore, for therapeutic pharmacological options, some antihypertensive drugs and oral antidiabetic agents may reduce arterial stiffness. Renin-angiotensin system inhibitors appear to be superior to all other antihypertensive drugs in reducing arterial stiffness [46]. A double-blind, randomized, placebo-controlled clinical trial showed that treatment with dapagliflozin (an SGLT2 inhibitor) for 12 weeks significantly reduced PWV in patients with type 2 diabetes [47]. Moreover, a recent large-scale multidatabase cohort study demonstrated that SGLT2 inhibitors reduced the risk of cardiovascular events compared with the use of dipeptidyl peptidase-4 (DPP-4) inhibitors [48]. In the future, more direct evidence is needed to explore the utility of medication intervention for baPWV for cardiovascular protection and secondary prevention in patients with ASCVD.

The present meta-analysis has some limitations. First, significant heterogeneity was observed for the risk estimate of CV events. We performed sensitivity analysis and subgroup analyses to investigate the sources of heterogeneity. The heterogeneity might have been attributable to the difference in the definition of high baPWV, as well as other factors (such as baseline risk levels and medications after discharge) that we were unable to investigate due to the absence of data. Second, the participants of the included studies were from Asian ASCVD populations (Japan and Korea). Therefore, the generalizability of our findings to other non-Asian populations is limited. Third, the number of included studies was small, which may have reduced the robustness of the results. Accordingly, large, multicenter, and long-term follow-up studies in the future will be required to draw definite conclusions regarding predictive value. Despite these methodological concerns, our pooled results consistently showed that high baPWV was associated with an increased risk of CV events and mortality in ASCVD patients.

Conclusion

In conclusion, this study showed a positive association between baPWV and future CV events, CV mortality, and all-cause mortality in ASCVD patients. These findings indicate that screening for baPWV might be valuable in the prognosis and risk assessment of patients with ASCVD.