Peripheral vs. Central Arterial Stiffness and Cardiovascular Events in Older Adults: The Atherosclerosis Risk in Communities (ARIC) study

Abstract

Aims

Carotid-femoral pulse wave velocity (cfPWV) is a representative measure of central arterial stiffness and an independent predictor of cardiovascular disease (CVD). Femoral-ankle PWV (faPWV) represents peripheral arterial stiffness, but its association with CVD has not been specifically investigated.

Methods

We analyzed 3,402 ARIC participants without prior coronary heart disease (CHD), heart failure (HF), or stroke at Visit 5 (2011–13) (mean age 74.8 [4.9] years, 36.1% male, 22.0% Black). faPWV and cfPWV were measured by Omron VP-1000 Plus. The primary outcome was CVD (CHD, HF, and stroke). We used multivariable Cox proportional hazards models.

Results

During a median 9.0-year follow-up, 607 CVD events occurred. Overall, faPWV showed an inverse association with CVD, with hazard ratio (HR) for top vs. bottom quartile 0.80 (95%CI 0.64–1.01) and p-for-trend 0.017 in Model 1 (demographically adjusted) and HR 0.86 (0.68–1.09) and p-for-trend 0.096 in Model 2 (further adjusted for CVD risk factors). In contrast, cfPWV was positively associated with CVD in both Models (HR for top vs. bottom quartile 1.22 [0.95–1.56], p-for-trend=0.043 in Model 2). The ratio of cfPWV to faPWV (“cf-fa ratio”) showed a stronger association with CVD (HR 1.37 [1.07–1.74], p-for-trend=0.005) than cfPWV. Examining CVD subtypes, the significant contrast in Model 2 was cf-fa ratio and HF.

Conclusions

faPWV showed a borderline significant inverse association with CVD, and cf-fa ratio appeared more strongly associated with CVD than cfPWV. Our findings indicate distinct prognostic implications of central vs. peripheral arterial stiffness and support cf-fa ratio as an alternative measure for CVD risk assessment.

Abstract word count: 250 words

Lay summary

Central arterial stiffness (i.e., aortic stiffness) is known to predict future cardiovascular disease (CVD), but this study examined the prognostic value of peripheral arterial stiffness (i.e., stiffness of leg arteries), using data from 3,402 older adults free of prior CVD at baseline. Carotid-femoral pulse-wave velocity (cfPWV) and femoral-ankle PWV (faPWV) were used to measure central and peripheral arterial stiffness, respectively, and faster velocity indicates greater stiffness.

Key findings include:

• As anticipated, cfPWV was positively associated with the risk of CVD whereas faPWV tended be associated inversely. This finding suggests that stiffness of different locations have different implications for cardiovascular health.

• The ratio of cfPWV to faPWV (“cf-fa ratio”) demonstrated stronger associations with CVD compared to cfPWV, the current reference measure of central arterial stiffness. The cf-fa ratio could be an alternative measure of arterial stiffness for classifying the future risk of CVD in older adults.

Introduction

Arterial stiffness is a well-established independent predictor of mortality and cardiovascular disease (CVD).¹ Aortic stiffness is typically assessed using pulse wave velocity (PWV).^2,3 Since pressure waves travel faster in stiffer arteries, a higher PWV value reflects greater arterial stiffness.⁴ Although PWV can be measured at various arterial segments, carotid-femoral PWV (cfPWV) is considered the reference standard measure of arterial stiffness, given its robust association with future CVD events.^1,5 cfPWV mainly represents the stiffness of the descending and abdominal aorta and is therefore considered a marker of central arterial stiffness.

In contrast, femoral-ankle PWV (faPWV) represents the stiffness of the peripheral arteries of the lower extremity. Although the investigation of faPWV is much more limited than cfPWV, several studies have reported counterintuitively inverse associations between faPWV and CVD status. For example, the Atherosclerosis Risk in Communities (ARIC) study found that higher faPWV was associated with a lower prevalence of coronary heart disease (CHD) and heart failure (HF)⁶; a Japanese study reported consistent results.⁷ Similar inverse associations were reported with diabetes,⁸ chronic kidney disease,⁹ cardiac biomarkers such as troponin,¹⁰ and vascular calcification.¹¹ Only a few studies have explored the association of faPWV with future CVD events, and the findings remain inconclusive.^12,13

Therefore, we comprehensively evaluated the associations of faPWV with CVD outcomes in a large community-based cohort of older adults in the United States. We repeated the same analysis with cfPWV as a contrast. We hypothesized that faPWV would show inverse associations with CVD outcomes while cfPWV would demonstrate positive associations.

Methods

Study design

Details of the ARIC study have been reported previously.^14,15 In brief, the ARIC study is a community-based prospective cohort that enrolled 15,792 participants aged 45 to 64 years living in four U.S. communities (Forsyth County, North Carolina; Jackson, Mississippi; suburban Minneapolis, Minnesota; and Washington County, Maryland) in 1987–89 (Visit 1). We used Visit 5 (2011–2013) as a baseline for the present study since PWV was assessed for the first time in ARIC. The Institutional Review Boards at each site approved the study design to ensure ethical standards were met, including obtaining written informed consent from all participants. The study followed the ethical principles outlined in the Declaration of Helsinki.

Study population

Out of 6,538 participants at Visit 5, we excluded participants with missing faPWV and cfPWV data (n=1,450), those with clinical conditions potentially influencing PWV data (body mass index [BMI] ≥40 kg/m², arrhythmias, aortic aneurysm, history of aortic or peripheral artery procedures, moderate or severe aortic stenosis and regurgitation, and ejection fraction <30%) (n=560),⁶ those with outliers of faPWV or cfPWV (>3 SD from average) (n=61). We also excluded non-White and non-Black participants because of the small sample size (n=14), those with missing covariates (n=238), and those with a history of CHD, HF, or stroke (n=838). The final study sample included 3,402 participants (Figure S1).

faPWV and cfPWV

After resting in supine position for 5–10 minutes, PWV measurements were conducted using the Omron VP-1000 Plus system (Omron, Kyoto, Japan).¹⁶ PWV at several segments was calculated by dividing the distance between arterial sites by the transit time of the wave. For cfPWV, applanation tonometry sensors were used to capture arterial waveforms at the left common carotid and femoral arteries, and distances were measured directly. For faPWV, distances were automatically calculated based on participant height.^17,18 The repeatability has been largely similar between cfPWV and faPWV.¹⁹ All PWV measurements were performed at least twice for each participant and averaged. The mean of the left and right faPWV was used in the analysis.

Covariates

All covariates were collected at Visit 5 unless specified otherwise. Age at the time of Visit 5 was determined from the date of birth reported at Visit 1. Sex and race were self-reported at Visit 1. Educational level was collected at Visit 1 and categorized as basic (less than completed high school), intermediate (high school graduate or equivalent), and advanced (at least some college). BMI was defined as weight in kilograms divided by the square of height in meters. Diabetes was defined as fasting plasma glucose ≥126 mg/dL, non-fasting plasma glucose ≥200 mg/dL, hemoglobin A1c ≥6.5%, use of medication for diabetes, or self-reported physician diagnosis of diabetes. Blood pressure was measured three times by a certified technician using an automatic sphygmomanometer (OMRON HEM-907 XL) after 5 minutes of rest in the sitting position, and the average of the last two measurements was used for the analysis. Estimated glomerular filtration rate (eGFR) was calculated by the Chronic Kidney Disease Epidemiology Collaboration creatinine-cystatin equation.²⁰ Total cholesterol and high-density lipoprotein (HDL) cholesterol were assessed using the Olympus Cholesterol reagent and Olympus HDL-Cholesterol test, respectively. The ankle-brachial index (ABI), a measure of peripheral artery disease, was measured with the same device as the PWV measurement. Participants were asked to report medication use in the prior four weeks and bring medication containers for verification.

Outcomes

ARIC conducted biannual telephone interviews with participants and active surveillance of local hospitals to identify potential CVD events. Deaths were also identified through the linkage of state vital statistics records and the National Death Index.¹⁴ The primary outcome of the present study was a composite CVD (CHD, HF, and stroke). We also analyzed the first event of CHD, HF, and stroke individually. In ARIC, all CVD outcomes were adjudicated by a physician panel through medical chart reviews. CHD was defined as definite or probable myocardial infarction, CHD death, or coronary revascularization procedure. HF was defined as definite or probable acute decompensated HF.²¹ Stroke included definite or probable ischemic or hemorrhage stroke. Participants were followed until the occurrence of CVD, loss-to-follow-up, death, or administrative censoring on December 31, 2021, whichever came first.

Statistical analysis

Baseline characteristics were summarized across quartiles of faPWV and cfPWV. Continuous and categorical variables were presented as means (SD) and counts (percentages), respectively. Those variables were compared using ANOVA and the chi-square test as appropriate. To assess the relationship between faPWV and cfPWV, we created a scatter plot with a linear regression line and calculated Pearson’s correlation coefficient.

We used the Kaplan-Meier method to estimate the cumulative incidence of CVD outcomes across quartiles of PWV measures and a log-rank test to compare their estimates overall and in each sex. We ran multivariable Cox proportional hazards models to quantify the association of quartiles of PWV measures with CVD outcomes, with the lowest quartile as a reference. We obtained p-for-trend by modeling quartiles as numerical variables (from 1–4). Two models were constructed: Model 1 adjusted for demographic variables (age, sex, race, education level) and field center, and Model 2 additionally accounted for clinical characteristics measured at Visit 5 (BMI, systolic blood pressure, antihypertensive medication use, eGFR, total cholesterol, HDL cholesterol, current smoking status, and diabetes). We also modeled PWV measures as restricted cubic splines to visualize potentially non-linear associations. For each PWV measure, we placed knots at the 5th, 35th, 65th, and 95th percentiles and set the reference at the median in the lowest quartile. Since faPWV was inversely associated with CVD outcomes in general, as shown subsequently, we additionally analyzed the ratio of cfPWV to faPWV (“cf-fa ratio”).

We have conducted a few sensitivity analyses to confirm the robustness of our primary findings. First, we excluded participants with ABI ≤0.9 (n=106) since lower limb stenosis or occlusion may interfere with the measurement of faPWV. Second, we ran Fine and Gray subhazard models accounting for competing risk of non-CVD deaths. Third, we performed subgroup analyses by age (<74 versus ≥74 years [median]), sex, race (Black versus White), and systolic blood pressure (<128.5 versus ≥128.5mmHg [median]). To obtain reliable estimates, in this subgroup analysis, we modeled PWV measures as a continuous variable using the overall SD of the total study population (1-SD: 168 cm/s for faPWV, 295 cm/s for cfPWV, and 0.33 for cf-fa ratio). We assessed statistical interactions by a likelihood ratio test comparing models with and without a relevant interaction term. To assess the predictive value of each PWV measure, we used the likelihood ratio test, Akaike’s Information Criterion (AIC), and Bayesian Information Criterion (BIC) to compare nested Cox models built on the fully adjusted model (Model 2). We also calculated Harrell’s C-statistic to evaluate model discrimination, as well as the difference from the base model using bootstrap resampling with 1,000 replications. A two-tailed p-value <0.05 was considered statistically significant. All analyses were performed with R version 4.3.2 (The R Foundation for Statistical Computing, Vienna, Austria).

Results

Baseline characteristics

Among 3,402 participants, the mean age was 74.8 (SD 4.9) years, with 36.1% male and 22.0% Black. Participants with higher faPWV tended to have favorable risk factor profiles (Table 1). Specifically, they were less likely to be current smokers, on antihypertensive medication, and to have diabetes and were more likely to have higher levels of HDL cholesterol, eGFR, and ABI. A few exceptions were higher systolic and diastolic blood pressure and heart rate along with higher faPWV. In contrast, participants in higher cfPWV quartiles tended to have poorer risk factor profiles with older age, lower educational attainment, higher prevalence of antihypertensive medication use and diabetes, and lower levels of HDL cholesterol, eGFR, and ABI (Table S1). The correlation between faPWV and cfPWV was weak (Figure S2), with a correlation coefficient of 0.023 (95%CI: −0.010–0.057, p=0.171).

Table 1.

Baseline characteristics according to quartiles of faPWV

	Overall	Q1	Q2	Q3	Q4	p
n	3402	851	851	849	851
faPWV range, cm/s	553 – 1658	553 – 977	978 – 1077	1078 – 1193	1194 – 1658
Age, y	74.8 (4.9)	74.9 (4.9)	74.4 (4.7)	74.7 (4.9)	75.3 (5.0)	0.002
Male (%)	1229 (36.1)	318 (37.4)	309 (36.3)	302 (35.6)	300 (35.3)	0.807
Black race (%)	750 (22.0)	255 (30.0)	191 (22.4)	168 (19.8)	136 (16.0)	<0.001
Education (%)						0.017
Basic	376 (11.1)	107 (12.6)	94 (11.0)	87 (10.2)	88 (10.3)
Intermediate	1445 (42.5)	348 (40.9)	335 (39.4)	358 (42.2)	404 (47.5)
Advanced	1581 (46.5)	396 (46.5)	422 (49.6)	404 (47.6)	359 (42.2)
Field center						<0.001
Forsyth County, NC	672 (19.8)	131 (15.4)	126 (14.8)	180 (21.2)	235 (27.6)
Jackson, MS	698 (20.5)	241 (28.3)	176 (20.7)	156 (18.4)	125 (14.7)
Minneapolis, MN	1101 (32.4)	229 (26.9)	306 (36.0)	301 (35.5)	265 (31.1)
Washington County, MD	931 (27.4)	250 (29.4)	243 (28.6)	212 (25.0)	226 (26.6)
Body mass index, kg/m2	27.8 (4.5)	29.2 (4.7)	28.1 (4.4)	27.4 (4.3)	26.6 (4.1)	<0.001
Current smoker (%)	194 (5.7)	69 (8.1)	53 (6.2)	35 (4.1)	37 (4.3)	0.001
Antihypertensive medication (%)	2070 (60.8)	558 (65.6)	528 (62.0)	513 (60.4)	471 (55.3)	<0.001
Diabetes (%)	945 (27.8)	272 (32.0)	249 (29.3)	222 (26.1)	202 (23.7)	0.001
Systolic blood pressure, mmHg	129.9 (17.2)	125.5 (16.0)	128.2 (16.1)	130.5 (16.5)	135.2 (18.0)	<0.001
Diastolic blood pressure, mmHg	66.6 (10.1)	62.75 (9.5)	65.7 (9.5)	67.5 (9.6)	70.3 (10.3)	<0.001
Heart rate, bpm	62.2 (10.1)	61.1 (9.6)	61.5 (10.1)	62.5 (10.1)	63.8 (10.6)	<0.001
Total cholesterol, mmol/L	4.89 (1.05)	4.80 (1.01)	4.86 (1.06)	4.87 (1.02)	5.01 (1.09)	0.001
HDL cholesterol, mmol/L	1.40 (0.37)	1.35 (0.33)	1.41 (0.36)	1.41 (0.36)	1.42 (0.39)	<0.001
eGFR, ml/min/1.73m2	69.3 (17.0)	66.4 (17.7)	69.7 (17.4)	70.5 (16.3)	70.7 (16.2)	<0.001
Ankle-brachial index	1.14 (0.11)	1.10 (0.15)	1.14 (0.10)	1.15 (0.09)	1.16 (0.09)	<0.001
cfPWV, cm/s	1150 (295)	1150 (315)	1134 (288)	1148 (289)	1168 (288)	0.116
faPWV, cm/s	1091 (168)	890 (73)	1028 (29)	1131 (33)	1313 (103)	<0.001
cf-fa ratio	1.08 (0.33)	1.30 (0.39)	1.10 (0.28)	1.02 (0.26)	0.89 (0.23)	<0.001

Continuous variables with mean (SD) and categorical variables with n (%) are presented.

cfPWV, carotid-femoral pulse wave velocity; eGFR, estimated glomerular filtration rate; faPWV,　femoral-ankle pulse wave velocity; HDL, high-density lipoprotein; and Q, quartile.

Composite cardiovascular disease

During a median follow-up of 9.0 years (interquartile interval 7.2–8.0 years, maximum 10.6 years), there were 607 CVD events (CHD: 176, HF: 298, and stroke: 133), with an incidence rate of 22.4 per 1,000 person-years. Cumulative incidence of CVD was highest in the lowest quartile of faPWV, followed by the second lowest, the highest, and the second highest, with p log-rank of 0.07 (Figure 1A). On the other hand, as anticipated, for cfPWV, the cumulative incidence was highest in the highest quartile, followed by the second highest, the second lowest, and the lowest (Figure 1B), with p log-rank of <0.001 (Figure 1B). The results were broadly similar for cf-fa ratio (Figure 1C). When stratified by sex, for faPWV, women showed similar patterns (log‑rank p=0.030) as the overall study population whereas in men the lowest and highest quartiles had higher CVD risk than the middle two quartiles although the differences did not reach statistical significance (log‑rank p=0.150) (Figure S3). The results were largely similar between women and men for cfPWV and cf-fa ratio.

Figure 1. Cumulative probability of a composite cardiovascular disease by quartiles of faPWV, cfPWV, and cf-fa ratio.

Kaplan-Meier curves show the cumulative probability of major adverse cardiac events by quartiles of the faPWV (A), cfPWV (B), and cf-fa ratio (C).

cfPWV, carotid-femoral pulse wave velocity; CVD, cardiovascular disease; and faPWV, femoral-ankle pulse wave velocity.

The inverse association of faPWV with CVD was significant after demographic adjustment (Model 1: hazard ratio [HR] for top vs. bottom quartile 0.80 [95%CI 0.64–1.01], p-for-trend=0.017). (Table 2) The association was borderline significant after further adjustment for CVD risk factors in Model 2 (HR 0.86 [0.68–1.09], p-for-trend=0.096). In contrast, cfPWV was positively associated with CVD in both Models (e.g., HR for top vs. bottom quartile 1.41 [1.11–1.78] and p-for-trend <0.001 in Model 1 and 1.22 [95%CI 0.95–1.56] and 0.043 in Model 2). We found that the association of cf-fa ratio with CVD appeared stronger (e.g., HR for top vs. bottom quartile 1.37 [1.07–1.74] and p-for-trend 0.005 in Model 2) than cfPWV.

Table 2.

Hazard ratios of composite cardiovascular disease by faPWV, cfPWV, and cf-fa ratio

		Q1	Q2	Q3	Q4	p for trend
		HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
faPWV	Model 1	1 (reference)	0.93 (0.74–1.16)	0.75 (0.60–0.95)	0.80 (0.64– 1.01)	0.017
	Model 2	1 (reference)	0.98 (0.78–1.22)	0.81 (0.64–1.03)	0.86 (0.68– 1.09)	0.096
cfPWV	Model 1	1 (reference)	0.99 (0.77–1.27)	1.25 (0.99–1.59)	1.41 (1.11–1.78)	<0.001
	Model 2	1 (reference)	0.93 (0.72–1.20)	1.13 (0.89–1.45)	1.22 (0.95–1.56)	0.043
cf-fa ratio	Model 1	1 (reference)	1.08 (0.84–1.39)	1.27 (1.00–1.62)	1.64 (1.29–2.07)	<0.001
	Model 2	1 (reference)	1.01 (0.79–1.30)	1.15 (0.90–1.47)	1.37 (1.07–1.74)	0.005

Model 1 was adjusted for age, sex, race, education level, and field center.

Model 2 was additionally adjusted for body mass index, systolic blood pressure, estimated glomerular filtration rate, total cholesterol, high-density lipoprotein cholesterol, current smoking status, antihypertensive medication use, and diabetes.

cfPWV, carotid-femoral pulse wave velocity, CI, confidence interval; faPWV, femoral-ankle pulse wave velocity; HR, hazard ratio; and Q, quartile.

When we modeled these PWV measures as restricted cubic splines, for faPWV, the inverse association with CVD risk was evident in its lower range (<~900 cm/s), and then there was no evident risk gradient (Figure 2A). For both cfPWV and cf-fa ratio, CVD risk steadily increased at the range above their medians (Figures 2B and 2C). Again, the risk gradient appeared steeper for cf-fa ratio compared to cfPWV.

Figure 2. The adjusted hazard ratio for composite cardiovascular disease according to faPWV (A), cfPWV (B), and cf-fa ratio (C) modeled as restricted cubic spline.

Solid lines and gray shadings represent the adjusted hazard ratio for CVD and the 95% confidence interval. The histogram represents the number of study participants. The reference values are the median of faPWV (908 cm/s), cfPWV (847 cm/s), and cf-fa ratio (0.75) in quartile 1. The adjusted hazard ratio was estimated using a Cox proportional hazards model adjusted for age, sex, race, education, field center, body mass index, current smoking, systolic blood pressure, antihypertensive medication, diabetes, estimated glomerular filtration rate, total cholesterol, and high-density lipoprotein cholesterol. The x-axes show the estimated values of faPWV, cfPWV, and cf-fa ratio from 0.5 to 99.5 percentiles.

cfPWV, carotid-femoral pulse wave velocity; CI, confidence interval; CVD, cardiovascular disease; and faPWV, femoral-ankle pulse wave velocity.

Subgroup analyses showed that the associations between PWV measures and CVD risk were generally consistent across age, sex, race, and systolic blood pressure categories. However, a significant interaction for sex was observed in cf-fa ratio (p-for-interaction = 0.022), with a stronger association in females (HR 1.28, 95% CI 1.15–1.41) compared with males (HR 1.06, 95% CI 0.95–1.19) (Table 3). Excluding participants with ABI ≤0.9 yielded similar results (Table S2). Furthermore, when we employed Fine and Gray competing risk models, the general patterns remained identical to the primary analysis (Table S3). When we added each PWV to the fully adjusted base model, the likelihood ratio test demonstrated an improvement in model fit (p < 0.001 for both of cfPWV and cf-fa ratio) (Table S4). The improvement in c-statistic was most evident when we added cf-fa ratio (delta c-statistic 0.006 [95%CI 0.001–0.012]).

Table 3.

Associations of faPWV, cfPWV, and cf-fa ratio as a linear term (per 1 SD) and risk of CVD by subgroups

Subgroup	n	faPWV		cfPWV		cf-fa ratio
		Adjusted HR (95%CI)	p for interaction	Adjusted HR (95%CI)	p for interaction	Adjusted HR (95%CI)	p for interaction
Age
≥74 yr	1,823	0.91 (0.82–1.01)	0.780	1.11 (1.00–1.22)	0.080	1.15 (1.05–1.26)	0.196
<74 yr	1,579	0.94 (0.80–1.10)		1.31 (1.12–1.55)		1.29 (1.11–1.49)
Sex
Male	1,229	1.01 (0.88–1.16)	0.097	1.08 (0.95–1.23)	0.216	1.06 (0.95–1.19)	0.022
Female	2,173	0.87 (0.78–0.97)		1.21 (1.08–1.35)		1.28 (1.15–1.41)
Race
White	2,652	0.93 (0.85–1.03)	0.752	1.14 (1.04–1.26)	0.787	1.17 (1.07–1.28)	0.903
Black	750	0.90 (0.74–1.10)		1.17 (0.99–1.40)		1.18 (1.02–1.37)
Systolic blood pressure
≥128.5 mmHg	1,740	0.91 (0.82–1.01)	0.618	1.17 (1.06–1.31)	0.610	1.19 (1.09–1.31)	0.543
<128.5 mmHg	1,662	0.95 (0.83–1.10)		1.12 (0.97–1.29)		1.13 (1.00–1.29)
Overall	3,402	0.93 (0.85–1.01)		1.16 (1.06–1.26)		1.18 (1.09–1.27)

The model was adjusted for age, sex, race, education level, field center, body mass index, systolic blood pressure, estimated glomerular filtration rate, total cholesterol, high-density lipoprotein cholesterol, current smoking status, antihypertensive medication use, and diabetes (same as Model 2). Statistical significance for interactions using a likelihood ratio test.

cfPWV, carotid-femoral pulse wave velocity, CI, confidence interval; CVD, cardiovascular disease; faPWV, femoral-ankle pulse wave velocity; HR, hazard ratio; Q, quartile; and SD, standard deviations.

Individual CVD outcomes

The cumulative incidence of CHD, HF, and stroke did not differ significantly by quartiles of faPWV but did by quartiles of cfPWV and cf-fa ratio (Figures S4–6). Sex-stratified Kaplan–Meier curves for CHD, HF, and stroke are provided in Figures S7–S9. When each CVD subtype was examined separately with these three stiffness measures in Model 2, the only statistically significant contrast was cf-fa ratio and HF (HR of top vs. bottom quartile 1.40 [95%CI: 1.03 – 1.90], p for trend=0.011) (Table 4).

Table 4.

Hazard ratios of coronary heart disease, heart failure, and stroke by faPWV, cfPWV, and cf-fa ratio

	Hazard ratio (95%CI)
	faPWV	cfPWV	cf-fa ratio
Coronary heart disease
Q1	1 (reference)	1 (reference)	1 (reference)
Q2	1.12 (0.76–1.62)	1.10 (0.72–1.69)	1.21 (0.79–1.85)
Q3	0.86 (0.57–1.30)	1.14 (0.75–1.75)	1.21 (0.79–1.84)
Q4	0.97 (0.65–1.45)	1.39 (0.90–2.13)	1.40 (0.92–2.12)
p for trend	0.601	0.131	0.135
Heart failure
Q1	1 (reference)	1 (reference)	1 (reference)
Q2	0.90 (0.68–1.19)	1.03 (0.75–1.43)	0.95 (0.68–1.33)
Q3	0.84 (0.63–1.14)	1.06 (0.77–1.46)	1.17 (0.86–1.60)
Q4	0.87 (0.64–1.17)	1.25 (0.91–1.72)	1.40 (1.03–1.90)
p for trend	0.303	0.147	0.011
Stroke
Q1	1 (reference)	1 (reference)	1 (reference)
Q2	0.77 (0.49–1.20)	0.66 (0.39–1.11)	0.78 (0.47–1.29)
Q3	0.69 (0.44–1.10)	1.06 (0.67–1.70)	0.95 (0.59–1.52)
Q4	0.86 (0.55–1.34)	1.28 (0.80–2.05)	1.30 (0.83–2.05)
p for trend	0.439	0.100	0.140

The model was adjusted for age, sex, race, education level, field center, body mass index, systolic blood pressure, estimated glomerular filtration rate, total cholesterol, high-density lipoprotein cholesterol, current smoking status, antihypertensive medication use, and diabetes (same as Model 2).

cfPWV, carotid-femoral pulse wave velocity, CI, confidence interval; faPWV, femoral-ankle pulse wave velocity; HR, hazard ratio; and Q, quartile.

Discussion

In this large community-based cohort study of older adults in the United States, we observed an inverse association between faPWV and CVD outcomes, while cfPWV showed positive associations as expected. These associations persisted after adjusting for potential demographic and clinical confounders across various subgroups and accounting for competing risk of deaths. We also observed that cf-fa ratio was more strongly associated with CVD outcomes than cfPWV alone, the reference standard measure of central arterial stiffness. When we examined faPWV, cfPWV, and cf-fa ratio and their associations with each CVD subtype separately, the only statistically significant contrast in our fully adjusted model was seen for cf-fa ratio and HF.

A small Japanese study of patients with hypertension showed positive associations between faPWV and CVD outcomes, but none of their Cox analyses was statistically significant, probably due to too few outcomes (n <45).¹² Stone et al. recently reported a non-significant inverse association between faPWV and CVD events using data from the ARIC Study.¹³ The present study with a longer follow-up and a larger number of CVD events extended the literature in a few aspects: the inverse association reaching statistical significance in some analyses and exploring the full-specturm of faPWV using spline terms.

We observed that cf-fa ratio, the ratio of cfPWV over faPWV, was more strongly associated with incident CVD, particularly HF, compared with cfPWV alone. The cf-fa ratio is a reciprocal of “carotid-femoral stiffness gradient” one of the main exposures of interest in a recent paper by Stone et al.¹³ Our study uniquely demonstrated that cf-fa ratio is particularly robustly associated with future HF risk. Of note, a small study with ~300 patients with kidney failure also reported that the ratio of cfPWV to carotid-brachial PWV was a stronger predictor of mortality than each of those PWV parameters alone.²²

In terms of potential mechanisms linking cf-fa ratio to CVD risk, we should acknowledge the concept of central-peripheral arterial stiffness gradient. In healthy individuals, peripheral muscular arteries (e.g., brachial and tibial arteries) are typically stiffer than the central elastic aorta,²³ presumably to attenuate forward pressure waves and ultimately protect end organs from excessive pulsatile pressure.^3,24,25 However, this gradient can reverse with aging as the aorta stiffens while peripheral artery stiffness reportedly remains largely unchanged.²³ Although the end organ for leg arteries is the leg itself, it is possible that faPWV may represent arterial stiffness of other peripheral arteries. According to this concept, it seems reasonable that we observed poorer prognosis related to higher cf-fa ratio, namely higher central PWV and lower peripheral PWV. Nonetheless, we should acknowledge that our findings of cf-fa ratio in our study are hypothesis-generating.Future studies should replicate our results, and if so, we should explore potential mechanisms.

Our study has a few clinical and pathophysiological implications. For example, while cfPWV has been considered the reference standard measure for assessing CVD risk, our data suggest the value of simultaneously assessing faPWV in older adults and calculating cf-fa ratio for better prognostication than cfPWV alone. Importantly, adding faPWV to cfPWV measurements is relatively easy, requiring only the addition of ankle cuffs to existing protocols to measure cfPWV. Also, our results support that pathophysiology in the cardiovascular system is not uniform across vascular beds. Nonetheless, as described above, data on faPWV are sparse, and thus, future studies should replicate our findings.

Whether cf-fa ratio is comparable with other potent CVD risk predictors (e.g., coronary artery calcification or natriuretic peptides) is an important question but beyond the scope of the present study. However, each modality has strengths and challenges (e.g., coronary artery calcium is one the strongest predictors of atherosclerotic CVD but requires computed tomography scanner using radiation), and thus the selection of modalities should be context-specific. In this regard, the measurement of PWVs is totally non-invasive and can simultaneously obtain data on the ABI, a first-line measure of peripheral artery disease. Given that cf-fa ratio showed the strongest association with HF, its measurement could be particularly important in scenarios where HF risk assessment is a priority. Nonetheless, this concept should be evaluated in future studies. Also, cf-fa ratio should be directly compared to other potent predictors of HF such as natriuretic peptides.

This study has a few limitations. First, the study population was limited to White and Black older adults, which may limit the generalizability of the findings to other age and racial/ethnic groups. In this context, nearly two thirds of our study population were women likely reflecting their longevity compared to men although women accounted for ~55% of ARIC participants at visit 1.²⁶ Second, given our exclusion criteria, we could not simply extrapolate our findings to individuals with clinical conditions that could affect PWV data (e.g., morbidly obese, major arrhythmia, or history of vascular procedures). Third, the biological mechanisms underlying the difference between central versus peripheral arterial stiffness remain unclear, warranting future studies to explore other relevant biological factors such as inflammation and endothelial dysfunction. Finally, although we carefully adjusted for several potential confounders, we cannot deny the possibility of residual confounding.

Conclusion

In this large community-based cohort study of older adults without CVD, we found that faPWV tended to be inversely associated with the incidence of CVD, while cfPWV showed a positive association. Notably, cf-fa ratio was more strongly associated with CVD (especially HF) than cfPWV. This study indicates the distinct prognostic implications of central and peripheral arterial stiffness and suggests cf-fa ratio as an alternative stiffness measure for CVD risk assessment.

Abbreviations list

ABI

ankle-brachial index

ARIC

Atherosclerosis Risk in Communities

BMI

body mass index

cfPWV

carotid-femoral pulse wave velocity

cf-fa ratio

carotid-femoral to femoral-ankle pulse wave velocity ratio

CHD

coronary heart disease

CI

confidence interval

CVD

cardiovascular disease

eGFR

estimated glomerular filtration rate

faPWV

femoral-ankle pulse wave velocity

HDL

high-density lipoprotein

HF

heart failure

HR

hazard ratio

PWV

pulse wave velocity

SD

standard deviation

Data availability statement

The ARIC datasets are accessible through BioLINCC upon obtaining study approvals in accordance with NIH policies. Data request forms are available at https://biolincc.nhlbi.nih.gov/studies/aric/.