Brachial-Ankle Pulse Wave Velocity Is Associated with Kidney Function Decline in Adults Undergoing Cardiovascular Disease Screening

Visual Abstract

Abstract

Key Points

• We studied the link between brachial-ankle pulse wave velocity (baPWV) and CKD risk in adults undergoing cardiovascular disease screening.

• A higher baseline baPWV was predictive of an increased annualized rate of creatinine-based eGFR (eGFR_Cr) decline.

• Compared with normal baPWV, higher baseline baPWV was associated with increased risk of developing CKD and lower CKD-free survival time.

Background

CKD affects 35 million adults in the United States and is associated with high morbidity and mortality. We investigated whether brachial-ankle pulse wave velocity (baPWV), a noninvasive measure of arterial stiffness, was associated with kidney function decline in adults undergoing cardiovascular disease (CVD) screening.

Methods

In 1862 patients referred for an exercise electrocardiogram test between 2007 and 2009, baPWV was measured using an Omron Colin oscillometric device. Creatinine-based GFR was estimated (eGFR_Cr) for participants with available baseline and follow-up creatinine values using the CKD Epidemiology Collaboration formula. We performed multivariable linear regression analyses to assess whether baseline baPWV was associated with annualized change in eGFR_Cr, after adjustment for age, sex, body mass index, baseline eGFR_Cr, and traditional CVD risk factors, including mean arterial pressure, diabetes and hypertension diagnoses, nephroprotective medication use, smoking status, and total cholesterol. To determine risk of developing CKD and CKD-free survival time by arterial stiffness subgroups, we performed covariable-adjusted Cox proportional hazards modeling and a Kaplan-Meier survival analysis, respectively.

Results

After exclusion of participants with missing covariable data and adjustment for covariables, one SD increase in baPWV was associated with a 0.18 ml/min per 1.73 m² BSA greater yearly decrease in eGFR_Cr (P = 0.023) over a median follow-up time of 13.2 years. Participants with higher baseline baPWV had increased risk of CKD (hazard ratio [95% confidence interval], P value; elevated baPWV: 2.4 [1.1 to 5.4], P = 0.037; borderline-elevated baPWV: 1.8 [1.1 to 3.1], P = 0.028) and had a shorter CKD-free survival time (median CKD-free survival time [years]; normal baPWV: 16.2; borderline-elevated baPWV: 16.0; elevated baPWV: 15.6 [P < 0.001]).

Conclusions

Elevated baPWV, a noninvasive measure of arterial stiffness that can be obtained in the office setting, was associated with decline in kidney function, risk of CKD development, and CKD-free survival in individuals undergoing CVD screening.

Introduction

CKD affects 35 million adults in the United States (US) and is associated with high morbidity and mortality.¹ Common risk factors for developing CKD include hypertension, diabetes, cardiovascular disease (CVD), and prior AKI.² Increased arterial stiffness (AS) has been associated with the development and progression of CKD.³

Aortic pulse wave velocity (PWV) is considered the gold standard noninvasive measure of AS.^4,5 The distance (Δd) between two points in the arterial system and the time delay (Δt) between the start of the pressure waveform at those sites correspond to PWV, where PWV=Δd (centimeters)/Δt (seconds).⁴ Carotid-femoral PWV (cfPWV) is a commonly used measure of aortic stiffness representing the contribution of the elastic thoracic and abdominal aorta, which is responsible for buffering high pressures in the arterial system.^4,6 Several studies have demonstrated the association between higher aortic stiffness, measured by cfPWV, and kidney function decline, as defined by rate of change in eGFR (∆eGFR),^7-9 development of CKD,^3,10-15 and development of ESKD.¹⁶

cfPWV measurement typically requires exposure of the groin to place a tonometer at the femoral artery. Attempts have been made to avoid groin exposure by analyzing the pulse waveform at other sites, including the brachial and tibial arteries, yielding brachial-ankle PWV (baPWV). baPWV is commonly used to assess AS in many East Asian countries, but only a few studies have evaluated the relationship between baPWV and kidney outcomes, and the results have been inconsistent.^6,17-20 In addition, whether baPWV is associated with change in kidney function over time in adults from the United States is unknown.⁶ We hypothesized that AS, measured by baPWV, is associated with rate of kidney function decline and risk of developing CKD in US adults. We tested this hypothesis in adults referred for CVD screening in the United States.

Methods

Study Population and Inclusion Criteria

Patients referred to the Mayo Clinic Cardiovascular Health Clinic for CVD screening by exercise electrocardiogram (ECG) between 2007 and 2009 were approached for enrollment in the study. Inclusion criteria for the study included (1) age 18 years or older and (2) consent to enroll at the time of the exercise ECG screening test. In total, 1862 eligible patients consented to enroll in the study. The study was approved by the Mayo Clinic Institutional Review Board.

Cohort Definitions

To ensure that the largest possible number of recruited participants were included in the data analyses, while accounting for missing baseline or follow-up creatinine measures, two separate cohorts were defined. Cohort 1 was included in linear regression modeling of annualized change in creatinine-based eGFR (∆eGFR_Cr). Cohort 2 was included in the Kaplan-Meier survival analysis and Cox hazards proportional modeling. Participants meeting study inclusion criteria were included in one or both cohorts based on each cohort's exclusion criteria, as described in sections Cohort 1—Exclusion Criteria and Cohort 2—Exclusion Criteria.

Cohort 1—Exclusion Criteria

Participants were excluded from cohort 1 if less than two serum creatinine measurements were available, if serum creatinine measurements were less than 3 years apart, if research authorization was withdrawn, or if age, sex, and body mass index (BMI) data were unavailable. Participants were excluded from the fully adjusted linear regression model if baseline covariables, as defined in section Covariables, were unavailable.

Cohort 2—Exclusion Criteria

Participants were excluded from cohort 2 if they had preexisting CKD, defined as eGFR_Cr <60 ml/min per 1.73 m² BSA, urine albumin-to-creatinine-ratio (UACR) >30 mg/g, and/or International Classification of Diseases (ICD) codes for CKD stages 3–5 within 6 months of the baseline ECG. Participants were also excluded if serum creatinine measurements were not available within 6 months before or after the ECG test date or if follow-up eGFR_Cr measurements, UACR measurements, and/or relevant ICD codes were unavailable. Participants included in cohort 2 were excluded from Cox proportional hazards modeling if baseline covariables, as described in section Covariables, were unavailable.

Measurements

baPWV

baPWV was measured before the exercise ECG using an Omron Colin oscillometric device (model BP-203RPE, Omron Colin Co., Ltd., Tokyo, Japan). Participants were fasting and had rested for at least 5 minutes before the test. Occlusion cuffs were placed above the elbow and above the medial malleolus of participants while lying in a supine position, ECG electrodes were attached, and a microphone for phonocardiography was placed at the left fourth intercostal space. Brachial and tibial artery pressures with corresponding waveforms, heart sounds, and electrocardiograms were obtained simultaneously. The arterial path length between the upper arms and ankles was estimated from the height of the participant using standardized height-based formulae.^6,21,22 The time taken for the arterial pressure wave to travel between the brachial and tibial arteries was measured. baPWV was obtained from the ratio of arm to ankle distance and arm to ankle time difference, calculated by the waveform analyzer. Right and left baPWV were measured once and separately calculated, and the mean of the two measures was used for the analyses. The reproducibility and validity of the baPWV measurement using this method have been demonstrated previously.^21,22 Figure 1 depicts a schematic for the measurement of baPWV.

Figure 1.

Measurement of baPWV. baPWV=($\frac{{∆d}_{AB}}{{∆t}_{AB}}$) is measured by calculating the distance traveled by the pulse wave (${∆d}_{AB}$=$d_{SA}$−$d_{SB}$) between two vascular sites (i.e., ankle, brachium) divided by the time delay (${∆t}_{AB}$=${\hspace{0.17em}t}_{SA}$−$t_{SB}$) between the start of the pressure waveform at those sites. The height-based calculations of the distance between the suprasternal notch and brachium and ankle are $d_{SB}=\left[\hspace{0.17em}\left(0.2195\times H\right)-2.0734\hspace{0.17em}\right]$ and $d_{SA}=[\hspace{0.17em}\left(0.8129\times H\right)+12.328$], respectively, where $H$ is height is in centimeters.^21,22 Data from Stone et al.³³ baPWV, brachial-ankle pulse wave velocity.

Kidney Disease Measures

Serum creatinine measurements between participants' study enrollment dates and August 20, 2024, were extracted from the electronic health record using the Mayo Data Explorer.²³ eGFR_Cr was calculated using the 2021 CKD Epidemiology Collaboration creatinine-based formula:

$${eGFR}_{Cr}=142\times \mathit{min}{\left(\frac{S_{Cr}}{ĸ},1\right)}^{\alpha }\times \mathit{max}{\left(\frac{S_{Cr}}{ĸ},1\right)}^{-1.2000}\times {0.9938}^{Age}\times 1.012\hspace{0.17em}\left(if\hspace{0.17em}female\right),$$

where $S_{Cr}$ is a standardized serum creatinine in mg/dl, κ=0.7 (females) or 0.9 (males), α=−0.241 (females) or −0.302 (males), min $\left(\frac{S_{Cr}}{ĸ},1\right)$ is the minimum of $\frac{S_{Cr}}{ĸ}$ or 1, max $\left(\frac{S_{Cr}}{ĸ},1\right)$ is the maximum of $\frac{S_{Cr}}{ĸ}$ or 1, and age is measured in years.²⁴

Cohort 1

∆eGFR_Cr was calculated for participants with at least two creatinine values obtained over a period of at least 3 years. To account for the variable follow-up time of participants, annualized ∆eGFR_Cr was used as the outcome measure. Annualized ∆eGFR_Cr was defined as ∆eGFR_Cr divided by follow-up duration. ∆eGFR_Cr was calculated as the most recent eGFR_Cr value available at the time of data extraction minus the baseline value. Baseline eGFR_Cr was defined based on the creatinine value obtained on or after the exercise ECG date, if the time frame between baseline and follow-up creatinine dates was at least 3 years. Pre-enrollment creatinine measurements were used as baseline values if no baseline creatinine measurements on or after the enrollment date were available or if duration from the creatinine measure obtained on or after the enrollment date to the follow-up creatinine measure was less than 3 years. Follow-up duration was defined as the time between creatinine measures. To evaluate effect modification by baseline kidney function, we stratified cohort 1 participants into two groups: normal baseline eGFR_Cr (≥90 ml/min per 1.73 m² BSA) and reduced baseline eGFR_Cr (<90 ml/min per 1.73 m² BSA).

Cohort 2

For Cox proportional hazards modeling and the Kaplan-Meier survival analysis, each participant's baseline eGFR_Cr was defined as the median eGFR_Cr value within 6 months of (before or after) ECG testing. These participant-specific median eGFR_Cr values were then averaged to determine the mean value for the cohort. For Cox proportional hazards modeling, the primary outcome was the development of CKD. CKD onset was defined as the earliest occurrence of (1) eGFR_Cr <60 ml/min per 1.73 m² BSA for ≥3 months, (2) UACR >30 mg/g for ≥3 months, or (3) an ICD code indicating CKD stages 3–5. If a participant had multiple indicators, the first occurring measure was used to define CKD onset. A Kaplan-Meier survival analysis was performed to determine CKD-free survival time. The event of interest was a composite of CKD development or death prior to CKD onset. Survival time was defined as the time from the baseline creatinine measure to the first occurrence of CKD, death, or the end of study follow-up (August 20, 2024), whichever occurred first.

Covariables

For both models, atherosclerotic vascular disease risk factors, including age, sex, BMI, baseline eGFR_Cr, mean arterial pressure (MAP), diabetes and hypertension diagnoses, nephroprotective medication use, smoking status, and total cholesterol, were defined as covariables. Except baseline eGFR_Cr, all baseline covariable measures were obtained at the time of the exercise ECG. Baseline eGFR_Cr was defined according to the criteria outlined in sections Cohort 1 and Cohort 2 for cohorts 1 and 2, respectively.

Continuous variables included age, BMI, baseline eGFR_Cr, MAP, and total cholesterol. BMI was calculated by dividing the measured weight (kg) by height squared (m²). MAP was calculated as:

$$(\text{DBP}+1/3\hspace{0.17em}\left[\text{SBP}-\text{DBP}\right])$$

Where DBP is diastolic BP, SBP is systolic BP, and pressure is measured in mm Hg.

Binary variables for both models included sex (male versus female), diabetes diagnosis (present versus not present), hypertension diagnosis (present versus not present), smoking status (ever smoker versus never smoker), and nephroprotective medication use (yes versus no). Nephroprotective medication use was defined as baseline use of angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, or mineralocorticoid receptor antagonists.

Statistical Analyses

Analyses were performed using R.²⁵ P values < 0.05 denote statistical significance.

Baseline Characteristics

The distribution of continuous baseline variables was assessed through histograms, calculation of basic statistical measures (i.e., mean and SD, median and interquartile range, mode, range, skew, kurtosis), and the Shapiro-Wilk Test for normality. Baseline characteristics of the study population are summarized using descriptive statistics in Tables 1 and 2 for cohorts 1 and 2, respectively. Continuous variables were presented as mean±SD or median and interquartile range. Categorical variables were reported as percentages. The Wilcoxon rank-sum or Kruskal-Wallis tests and Pearson chi-squared or Fisher exact tests were used to determine statistically significant differences for continuous and categorical variables, respectively.

Table 1.

Baseline characteristics^a of cohort 1 (1271 participants^b), overall and across arterial stiffness categories based on brachial-ankle pulse wave velocity at study enrollment

Baseline Characteristics	Total (n=1271)	Normal baPWV (<14 m/s) (n=752)	Borderline-Elevated baPWV (14–18 m/s) (n=447)	Elevated baPWV (≥18 m/s) (n=72)	P Value
baPWV (m/s)	13.8±0.23	12.3±0.10	15.5±0.10	19.5±0.17	<0.001
eGFRCr (ml/min per 1.73 m2 BSA)	86±15	88±14	84±15	78±18	<0.001
Age (yr)	54±11	50±9	59±9	68±9	<0.001
Sex (% male)	69	67	70	67	0.47
Race (% White)	97	97	97	96	0.66
BMI (kg/m2)	28.9±5.4	28.4±5.2	29.6±5.6	29.9±5.0	<0.001
MAP (mm Hg)	97±12	92±9	103±10	111±14	<0.001
Hypertension diagnosisc (% present)	32	21	44	65	<0.001
Diabetes diagnosisd (% present)	7	4	10	19	<0.001
Nephroprotective medication use (% yes)	6	4	10	11	<0.001
Smoking statuse (% ever-smoker)	40	39	42	51	0.076
Total cholesterolf (mg/dl)	194±37	195±36	193±38	192±40	0.20
Follow-up time (yr)	13.3 (8.1–15.2)	13.2 (7.9–15.1)	13.5 (8.8–15.3)	12.1 (6.1–15.2)	0.18

baPWV, brachial-ankle pulse wave velocity; BMI, body mass index; eGFR_Cr, creatinine–based eGFR; MAP, mean arterial pressure.

^aBaseline characteristic values are reported as mean±SD, median (interquartile range), or percentage.

^bComplete data were not available for 243 participants. The baseline characteristic results within each category were calculated using only participants with available data.

^cHypertension diagnosis data were unavailable for six participants (four in normal brachial-ankle pulse wave velocity group and two in borderline-elevated brachial-ankle pulse wave velocity group).

^dDiabetes diagnosis data were unavailable for nine participants (seven in normal brachial-ankle pulse wave velocity group and two in borderline-elevated brachial-ankle pulse wave velocity group).

^eSmoking status data were unavailable for 11 participants (seven in normal brachial-ankle pulse wave velocity group and four in borderline-elevated brachial-ankle pulse wave velocity group).

^fTotal cholesterol data were unavailable for 231 participants (114 in normal brachial-ankle pulse wave velocity group, 97 in borderline-elevated brachial-ankle pulse wave velocity group, and 20 in elevated brachial-ankle pulse wave velocity group).

Table 2.

Baseline characteristics^a of cohort 2 (1694 participants^b), overall and across arterial stiffness categories based on brachial-ankle pulse wave velocity at study enrollment

Baseline Characteristics	Total (n=1694)	Normal baPWV (<14 m/s) (n=1052)	Borderline-Elevated baPWV (14–18 m/s) (n=552)	Elevated baPWV (≥18 m/s) (n=90)	P Value
baPWV (m/s)	13.7±0.23	12.2±0.11	15.5±0.10	19.6±0.19	<0.001
eGFRCr (ml/min per 1.73 m2 BSA)	88±14	90±14	86±14	83±15	<0.001
Age (yr)	53±11	49±10	59±9	68±8	<0.001
Sex (% male)	67	65	73	60	0.0010
Racec (% White)	97	96	97	96	0.87
BMI (kg/m2)	28.8±5.5	28.2±5.4	29.7±5.7	29.5±5.0	<0.001
MAP (mm Hg)	96±10	92±9	100±10	106±12	<0.001
Hypertension diagnosis (% present)	28	18	42	60	<0.001
Diabetes diagnosis (% present)	6	4	10	17	<0.001
Nephroprotective medication use (% yes)	17	17	18	17	0.75
Smoking statusd (% ever-smoker)	40	37	43	53	0.004
Total cholesterole (mg/dl)	195±40	196±37	195±44	196±49	0.54

baPWV, brachial-ankle pulse wave velocity; BMI, body mass index; eGFR_Cr, creatinine-based eGFR; MAP, mean arterial pressure.

^aBaseline characteristic values are reported as mean±SD, median (interquartile range), or percentage.

^bComplete data were not available for 341 participants. The baseline characteristic results within each category were calculated using only participants with available data.

^cRace data were unavailable in 21 participants (seven in normal brachial-ankle pulse wave velocity group, seven in borderline-elevated brachial-ankle pulse wave velocity group, and seven in elevated brachial-ankle pulse wave velocity group).

^dSmoking status data were unavailable in 54 participants (25 in normal brachial-ankle pulse wave velocity group, 24 in borderline-elevated brachial-ankle pulse wave velocity group, and five in elevated brachial-ankle pulse wave velocity group).

^eTotal cholesterol data were unavailable for 313 participants (174 in normal brachial-ankle pulse wave velocity group, 118 in borderline-elevated brachial-ankle pulse wave velocity group, and 21 in elevated brachial-ankle pulse wave velocity group).

Cohort 1—Linear Regression Modeling with Covariable Adjustment

We performed multivariable linear regression analyses to test whether baPWV was associated with annualized ∆eGFR_Cr in cohort 1. Two adjustment models were applied to the dataset. In model 1, covariables included age, sex, BMI, and baseline eGFR_Cr. Model 2 included these and all additional covariables, as defined in section Covariables. Given the bidirectional relationship between CKD and AS, we stratified our analysis by kidney function. Specifically, we created two models: one for patients with eGFR_Cr ≥90 ml/min per 1.73 m² BSA (model 2a) and another for those with eGFR_Cr <90 ml/min per 1.73 m² BSA (model 2b).

Cohort 2—Cox Proportional Hazards Modeling with Multivariable Adjustment and Kaplan-Meier Survival Analysis

In cohort 2, we performed covariable-adjusted Cox-proportional hazards modeling to depict risk of CKD development and Kaplan-Meier survival analysis to determine CKD-free survival time by baseline AS subgroups. Cox proportional hazards models incorporated all prespecified covariables, as defined in section Covariables, and assessed hazard ratios (HRs) for each predictor. Wilcoxon rank-sum, Kruskal-Wallis, Pearson chi-squared, and Fisher exact tests were used to determine statistical significance of results.

Results

Baseline Characteristics

Cohort 1

After excluding participants who withdrew consent or were missing prespecified baseline and follow-up data (n=591), cohort 1 included 1271 participants. The mean eGFR_Cr was 86 ml/min per 1.73 m² BSA at baseline. Based on previously defined baPWV thresholds, participants were categorized into normal (baPWV <14 m/s), borderline-elevated (14 m/s ≤baPWV <18 m/s), and elevated (baPWV ≥18 m/s) AS groups.^26–29 Baseline characteristics for cohort 1 participants are presented in Table 1.

Cohort 2

After excluding participants who withdrew consent or were missing prespecified baseline and follow-up data (n=168), cohort 2 included 1694 participants. Participants were categorized in the same AS subgroups as in cohort 1. These participants were included in the covariable-adjusted Cox proportional hazards modeling and the Kaplan-Meier survival analysis. Baseline characteristics for cohort 2 are presented in Table 2.

baPWV Predicts eGFR_Cr Decline

In cohort 1, the mean eGFR_Cr at follow-up was 76 ml/min per 1.73 m² BSA, resulting in a mean eGFR_Cr difference from baseline of 10 ml/min per 1.73 m² BSA after a median follow-up of 13.3 years. For each SD increase in baPWV, there was a 0.17 ml/min per 1.73 m² BSA greater annualized decline in eGFR_Cr, independent of age, sex, BMI, and baseline eGFR_Cr (P = 0.0072). After additional adjustment for CVD risk factors in participants with complete data available (n=1028), there was a 0.18 ml/min per 1.73 m² BSA greater annualized decline in eGFR_Cr per SD increase in baPWV (P = 0.023). In the fully adjusted model, elevated baPWV, older age, and unexpectedly, higher baseline eGFR_Cr was significant predictors of annualized ∆eGFR_Cr. In participants with baseline eGFR_Cr ≥90 ml/min per 1.73 m² BSA, there was a 0.27 ml/min per 1.73 m² BSA greater annualized decline in eGFR_Cr for each SD increase in baseline baPWV (P = 0.041). In this cohort 1 subgroup, age, sex, baseline eGFR_Cr, MAP, and diabetes diagnosis were other predictors of the annualized ∆eGFR_Cr. In those with baseline eGFR_Cr <90 ml/min per 1.73 m² BSA, there was no association between baPWV and annualized ∆eGFR_Cr. The results of the covariable-adjusted linear regression models are presented in Table 3 and Supplemental Tables 1-4.

Table 3.

Partially and fully adjusted multivariable linear regression analysis models depicting the association between brachial-ankle pulse wave velocity and annualized change in eGFR_Cr, adjusting for baseline brachial-ankle pulse wave velocity, age, sex, body mass index, and eGFR_Cr (model 1) and additionally, mean arterial pressure, diabetes and hypertension diagnoses, nephroprotective medication use, smoking status, and total cholesterol (model 2) in cohort 1 (1271 participants and 1028 participants, respectively^a-e) over a median follow-up time of 13.3 and 13.2 years, respectively

Model	Covariables	baPWV β	95% CI	P Value	Outcome
Model 1 (n=1271)	Age, sex, BMI, baseline eGFRCr	−0.00072	−0.0012 to −0.00027	0.0072	∆eGFRCr (ml/min per 1.73 m2 BSA)/yr
Model 2 (n=1028)	Age, sex, BMI, baseline eGFRCr, MAP, diabetes diagnosis, hypertension diagnosis, nephroprotective medication use, smoking status, total cholesterol	−0.00083	−0.0015 to −0.00012	0.023
Model 2a (n=397)	Age, sex, BMI, baseline eGFRCr, MAP, diabetes diagnosis, hypertension diagnosis, nephroprotective medication use, smoking status, total cholesterol	−0.0012	−0.0024 to −0.000056	0.041
Model 2b (n=631)	Age, sex, BMI, baseline eGFRCr, MAP, diabetes diagnosis, hypertension diagnosis, nephroprotective medication use, smoking status, total cholesterol	−0.00063	−0.0015 to 0.00027	0.17

The fully adjusted multivariable linear regression analysis was separately applied to cohort 1 participants with baseline eGFR_Cr ≥90 (model 2a) and <90 ml/min per 1.73 m² BSA (model 2b).

Model 1: adjusted R²: 0.13.

Model 2: adjusted R²: 0.14.

Model 2a: adjusted R²: 0.11.

Model 2b: adjusted R²: 0.14.

∆eGFRCr, change in eGFR; baPWV, brachial-ankle pulse wave velocity; b, coefficient; BMI, body mass index; CI, confidence interval; eGFRCr, creatinine-based eGFR; MAP, mean arterial pressure.

^aBaseline diabetes and hypertension diagnoses, smoking status, and total cholesterol data were unavailable for 243 of the 1271 cohort 1 participants. These participants were excluded from the model 2, 2a, and 2b multivariable linear regression analyses due to missing covariables.

^bDiabetes diagnosis data were unavailable for nine participants (seven in normal brachial-ankle pulse wave velocity group and two in borderline-elevated brachial-ankle pulse wave velocity group).

^cHypertension diagnosis data were unavailable for six participants (four in normal brachial-ankle pulse wave velocity group and two in borderline-elevated brachial-ankle pulse wave velocity group).

^dSmoking status data were unavailable for 11 participants (seven in normal brachial-ankle pulse wave velocity group and four in borderline-elevated brachial-ankle pulse wave velocity group).

^eTotal cholesterol data were unavailable for 231 participants (114 in normal brachial-ankle pulse wave velocity group, 97 in borderline-elevated brachial-ankle pulse wave velocity group, and 20 in elevated brachial-ankle pulse wave velocity group).

Higher Baseline baPWV Is Associated with Increased Incidence of CKD

Among the 1694 participants included in the Kaplan-Meier survival analysis, 145 developed CKD. Those with higher baseline AS had a shorter CKD-free survival time (median CKD-free survival time [years]; normal baPWV: 16.2; borderline-elevated baPWV: 16.0; elevated baPWV: 15.6 [P<0.001]; Figure 2 and Supplemental Table 6). Cox-proportional hazards modeling, with adjustment for covariables and exclusion of participants with missing covariable data (n=341), revealed increased risk of CKD in participants with elevated and borderline-elevated baseline baPWV when compared with normal baPWV (comparison groups: HR [95% confidence interval], P value; elevated versus normal baPWV: 2.4 [1.1 to 5.4], P = 0.037; borderline-elevated versus normal baPWV: 1.8 [1.1 to 3.1], P = 0.028). Table 4 presents the results of Cox proportional hazards modeling for the risk of developing CKD, showing HRs for baseline AS categories and relevant covariables. In addition to baseline elevated and borderline-elevated baPWV, predictive factors for developing CKD included older age, elevated BMI, hypertension and diabetes diagnoses, and lower baseline eGFR_Cr (Supplemental Table 5).

Figure 2.

CKD occurred in 145 participants. Cohort 2 participants (n=1694) with elevated baPWV and borderline-elevated baPWV had lower CKD-free survival time compared with participants with a normal baPWV. Median CKD-free survival times (IQR) were 16.1 years (15.6–16.6) for the cohort, 16.2 years (15.7–16.6) for normal baPWV, 16.0 years (15.4–16.6) for borderline-elevated baPWV, and 15.6 years (11.6–16.4) for elevated baPWV (P < 0.001). For this display only, Kaplan-Meier survival curves were truncated at 15.2 years for visualization due to sparse data beyond this point. However, median CKD-free survival times were calculated using the full dataset without censoring at 15.2 years, reflecting the complete follow-up time of participants. IQR, interquartile range.

Table 4.

Cox proportional hazards model evaluating the association between baseline characteristics and incident CKD in 1353 cohort 2 participants with complete covariable data^a

Variable	HR	95% CI	P Value
Elevated versus normal baPWV	2.4	1.1 to 5.4	0.037
Borderline-elevated versus normal baPWV	1.8	1.1 to 3.1	0.028
Age	1.1	1.0 to 1.1	<0.001
Male sex	1.0	0.62 to 1.6	0.98
BMI	1.1	1.0 to 1.1	0.016
MAP	1.0	0.98 to 1.0	0.89
Diabetes diagnosis	2.2	1.2 to 4.0	0.0080
Hypertension diagnosis	1.7	1.0 to 2.67	0.018
Nephroprotective anti-hypertensive medication use	1.3	0.76 to 2.1	0.38
Ever-smoker	0.84	0.55 to 1.3	0.41
Total cholesterol	1.0	0.99 to 1.0	0.34
Baseline eGFRCr	0.96	0.94 to 0.98	<0.001

Event occurrence: 96, Global P value (Log-Rank) 3.5063×10²⁴, Concordance Index 0.82, Akaike Information Criterion: 1250.25. baPWV, brachial-ankle pulse wave velocity; BMI, body mass index; CI, confidence interval; eGFR_Cr, creatinine-based eGFR; HR, hazard ratio; MAP, mean arterial pressure.

^aComplete data were not available for 341 of the 1694 cohort 2 participants. These participants were excluded from covariable-adjusted Cox proportional hazards modeling due to missing covariables.

Discussion

In this retrospective cohort study of adults referred for CVD screening with exercise ECG, we found that higher baseline baPWV was independently associated with a greater annualized decline in eGFR_Cr. This association persisted after adjustment for traditional risk factors and was primarily driven by participants with preserved kidney function at baseline. This suggests that AS may play a more prominent role in early kidney function decline, perhaps reflecting increased susceptibility to subclinical vascular injury or hyperfiltration-mediated damage in individuals with greater kidney reserve. Across all models, older age was a strong and consistent predictor of decline, underscoring aging's central role in annualized ∆eGFR_Cr. Unexpectedly, baseline eGFR_Cr was inversely associated with follow-up kidney function. In this case, our linear regression models likely reflected regression to the mean, where those starting with lower values had less room to decline or may have partially recovered from transient reductions, while those with higher values may have exhibited greater decline due to reversion toward their physiologic average or unrecognized hyperfiltration. Furthermore, including baseline eGFR_Cr as a covariable helped isolate the independent effect of baPWV but may have attenuated associations for comorbidities such as diabetes and hypertension, particularly among participants with reduced baseline eGFR_Cr. In a separate survival analysis, we observed a higher risk of incident CKD and shorter CKD-free survival time among individuals with elevated and borderline-elevated baPWV. After adjusting for relevant covariables, elevated baPWV was associated with a 2.4-fold increased risk of CKD, while borderline-elevated baPWV conferred a 1.8-fold increased risk. These findings suggest a dose-response relationship between AS and CKD risk, reinforcing the prognostic value of baPWV for long-term kidney outcomes. Older age, elevated BMI, diabetes and hypertension diagnoses, and lower baseline eGFR_Cr predicted CKD development in hazards modeling.

The differences observed between the Cox proportional hazards model and the multivariable linear regression model likely reflect key distinctions in outcome definitions, model structure, and data distribution. The Cox model evaluated the time to incident CKD, capturing whether participants crossed a clinically meaningful threshold, whereas the linear model assessed the annualized ∆eGFR_Cr, including all participants regardless of whether CKD developed. Although the Cox model was based on 96 CKD events, the linear model included over 1000 participants, making the former much more sensitive to predictors associated with earlier CKD onset. As a result, predictors showed stronger associations with incident CKD but weaker or nonsignificant associations with the overall rate of decline. Importantly, lower baseline eGFR_Cr was associated with CKD incidence in the Cox model because participants closer to the diagnostic threshold were inherently more likely to cross it over time. Unlike the linear model, which is more susceptible to regression to the mean, particularly in participants with initially high or low values, the Cox model defines a fixed clinical event, reducing the influence of natural fluctuation around a baseline. As a result, the association between lower baseline eGFR_Cr and incident CKD in the Cox model likely reflects true progression rather than a statistical artifact. These distinctions underscore the complementary nature of the two modeling approaches in characterizing CKD risk. Together, these results support a potential role for baPWV as an early marker of kidney vulnerability, particularly in the preclinical stages of kidney dysfunction. To the best of our knowledge, this is the first study to investigate the association between baseline baPWV and kidney function over time in US adults.

The kidney arterial vasculature is a relatively low-resistance bed exposed to the pulsations resulting from left ventricular ejection in systole.³⁰ The elastic aorta is responsible for buffering high arterial pressures, but arterial stiffening, specifically that of the aorta, augments pulsatility.³⁰ The reduction of this buffering and the exposure of kidney vasculature to increased pressure pulsatility may underly the association between increased baPWV and development of CKD. cfPWV, a marker of central AS (i.e., aortic stiffness), has been associated with CKD in multiple studies, and previous studies have shown that baPWV correlates well with aortic PWV.^3,7,16,21 Our study extends these prior reports by demonstrating that baPWV, a marker of both central and peripheral stiffness, is associated with kidney function and CKD-free survival.

Guidelines from the Japanese Society of Hypertension and European Society of Hypertension recommend evaluation for hypertension-mediated organ damage to the vascular system by measuring baPWV.^17,31 However, the evidence for baPWV as a predictor kidney outcomes in US populations was lacking before our study. In a prospective cohort study of 2053 adults in Japan, higher baPWV was associated with a lower follow-up eGFR and higher rate of annualized eGFR decline over a follow-up of 5–6 years.¹⁹ In the same study, changes in baPWV were not associated with increased baseline eGFR.¹⁹ The only previous study in the United States was a cross-sectional analysis evaluating the association of baPWV and kidney function in 3242 adults. Higher baPWV was not associated with lower eGFR.²⁰ By contrast, our study, with a median follow-up of 13.3 years, demonstrated this association.

One strength of our study is its relatively long follow-up duration, which enabled robust evaluation of longitudinal kidney outcomes. To assess longitudinal decline, we modeled annualized ∆eGFR_Cr, allowing for standardized comparisons across individuals with varying follow-up times. We stratified linear regression analyses by baseline eGFR_Cr to reflect clinically relevant subgroups and adjusted for baseline eGFR_Cr within each model to account for differences in kidney reserve. This approach helped clarify the association between baPWV and kidney function decline. In our survival analysis, we further strengthened validity by excluding individuals with preexisting CKD, thereby focusing on time to incident CKD, enhancing temporal clarity and internal validity. The use of a comprehensive CKD outcome definition enabled a broad and clinically meaningful assessment of incident CKD across both Kaplan-Meier and Cox proportional hazards models. Finally, the use of complementary modeling approaches provided a more comprehensive understanding of the relationship between AS and kidney outcomes.

Several limitations of our study should be noted. First, the study cohort had limited representation of non-White and female participants, which may restrict the generalizability of our findings to the broader US population. Second, although the 2024 Kidney Disease Improving Global Outcomes Clinical Practice Guideline for the Evaluation and Management of CKD recommends combining creatinine and cystatin C-based eGFR measures to increase diagnostic accuracy of CKD, cystatin C was not routinely measured during the study period (2007–2009) and thus could not be incorporated into our analysis.³² Only a single baseline measurement of baPWV was available, limiting our ability to assess changes in AS over time or evaluate baPWV as a longitudinal outcome. Although stratifying linear regression models and adjusting for baseline eGFR_Cr helped mitigate confounding by baseline kidney function, this approach does not fully capture the potential bidirectional relationship between vascular stiffness and CKD pathogenesis. In addition, although UACR >30 mg/g was included in the CKD outcome definition, albuminuria was not assessed as a continuous or stratified predictor, limiting our ability to evaluate its interaction with AS. We used diabetes diagnosis as a binary covariate given limited availability of fasting glucose and hemoglobin A1c data, which did not fully capture the spectrum of disease control. By defining antihypertensive medication variable to focus specifically on nephroprotective medications, we aimed to isolate their potential kidney benefit while minimizing, although not fully removing, the confounding influence of overall BP status. In Cox modeling, hypertension diagnosis was significantly associated with CKD risk. Meanwhile, nephroprotective medications were not significantly associated with CKD risk, which may reflect the relatively low use of these therapies in the cohort, limiting statistical power to detect their benefit. Newer nephroprotective agents, including sodium-glucose cotransporter 2 inhibitors and glucagon-like peptide-1 receptor agonists, were not available during the study period and therefore could not be evaluated. Finally, past reviews have detailed potential pitfalls of using baPWV as a marker of AS due to the measure's depiction of central (i.e., large, elastic arteries) and peripheral (i.e., medium-sized peripheral arteries of the lower extremity) AS, rather than central AS (i.e., aortic stiffness) measured by cfPWV.³³ Additional concerns include reduced reliability in patients with certain cardiovascular pathology (i.e., arterial stenosis or aneurysm) and the potential inaccuracy of height-based formulae to estimate arterial path length.³³ Although height-based baPWV measurements are generally valid and reproducible, this approach may introduce individual-level variability due to anatomical differences.^6,21 Despite these limitations, baPWV remains clinically valuable by capturing a broader spectrum of AS and has demonstrated predictive utility for CKD outcomes.

In a prior meta-analysis, the addition of aortic PWV to traditional CVD risk factors in an intermediate risk group improved prediction of a CVD event during a 10-year follow-up.³⁴ Similarly, risk equations exist for predicting CKD development and progression in those at risk for CKD or with early-stage CKD (CKD-Prognosis Consortium, Klinrisk, KindeyLx).² Our study results suggest that AS measured by baPWV is independently associated with kidney function and should be considered for inclusion in such equations. Addition of AS measures predicting CKD development may increase detection of at-risk participants, especially in groups considered to be low risk by traditional risk assessment, and allow for earlier initiation of nephroprotective strategies.²

In conclusion, baPWV, a measure of AS, was predictive of annualized ∆eGFR_Cr, CKD risk, and CKD-free survival in adults referred for CVD screening, particularly for individuals with preserved kidney function at baseline. Higher baPWV at baseline was associated with lower follow-up eGFR_Cr after a median follow-up duration of 13 years, as well as higher risk of developing CKD and lower CKD-free survival time. Overall, our findings contribute to the literature highlighting AS as a prognostic tool for kidney disease.³ Incorporating AS measures into existing CKD prediction tools may help to identify those at risk of developing CKD.² Furthermore, our study motivates investigations assessing consistency in the relationship between baPWV and kidney function decline when other markers, including creatinine and cystatin C-based eGFR and UACR, are used as markers of CKD and covariables. Future studies should incorporate repeated measures of both baPWV and kidney function to better characterize temporal dynamics and clarify causal relationships between AS and CKD progression. Research is needed to inform the implications of targeting AS to prevent development or delay progression of kidney disease.³⁵

Supplementary Material

kidney360-7-291-s001.pdf ^{(2.6MB, pdf)}

kidney360-7-291-s002.pdf ^{(195.6KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/KN9/B325.

Author Contributions

Conceptualization: Iftikhar J. Kullo.

Data curation: Meliksah Arslan, Kylie Van Dyke.

Formal analysis: Meliksah Arslan, Caroline B. Ledet, Kylie Van Dyke.

Funding acquisition: Iftikhar J. Kullo.

Investigation: Meliksah Arslan, A. Rauoof Malik.

Methodology: Meliksah Arslan, Caroline B. Ledet, A. Rauoof Malik, Kylie Van Dyke.

Project administration: Meliksah Arslan, Iftikhar J. Kullo, Caroline B. Ledet.

Resources: Iftikhar J. Kullo.

Software: Meliksah Arslan, Kylie Van Dyke.

Supervision: Iftikhar J. Kullo.

Validation: Meliksah Arslan, Kylie Van Dyke.

Visualization: Caroline B. Ledet, Kylie Van Dyke.

Writing – original draft: Meliksah Arslan, Caroline B. Ledet.

Writing – review & editing: Meliksah Arslan, Iftikhar J. Kullo, Caroline B. Ledet, Kylie Van Dyke.

Funding

I.J. Kullo: National Heart, Lung, and Blood Institute (K24 HL137010 and R01HL 135879).

Declarative Statements

This study includes clinical experimentation and received Institutional Review Board or Ethics Committee approval. All patients provided written informed consent. This study includes clinical experimentation and complies with the Declaration of Helsinki.

Data Availability Statements

Data cannot be shared. Explanation Why Data Cannot Be Shared: The data that support the findings of this study are not publicly available due to institutional policies. Data access is restricted to maintain the confidentiality and privacy of the study participants in compliance with applicable regulations and Institutional Review Board requirements. Consequently, the individual level data cannot be shared or made available for external use. Summary level data are available from the senior author on reasonable request.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/KN9/B326.

Supplemental Table 1. Multivariable linear regression analysis model (model 1) depicting the association between baPWV and annualized ΔeGFR_Cr, adjusting for baseline baPWV, age, sex, BMI, and baseline eGFR_Cr in cohort 1 (1271 participants) over a median follow-up time of 13.3 years.

Supplemental Table 2. Multivariable linear regression analysis model (model 2) depicting the association between baPWV and annualized ΔeGFR_Cr, adjusting for baseline baPWV, age, sex, BMI, baseline eGFR_Cr, MAP, diabetes and hypertension diagnoses, nephroprotective medication use, ever-smoker status, and total cholesterol in cohort 1 (1028 participants^a) over a median follow-up time of 13.2 years.

Supplemental Table 3. Multivariable linear regression analysis model (model 2a) depicting the association between baPWV and annualized ΔeGFR_Cr, adjusting for baseline baPWV, age, sex, BMI, baseline eGFR_Cr, MAP, diabetes and hypertension diagnoses, nephroprotective medication use, ever-smoker status, and total cholesterol in cohort 1 participants with baseline eGFR_Cr >90 ml/min per 1.73 m² BSA (397 participants^a).

Supplemental Table 4. Multivariable linear regression analysis model (model 2b) depicting the association between baPWV and annualized ΔeGFR_Cr, adjusting for baseline baPWV, age, sex, BMI, baseline eGFR_Cr, MAP, diabetes and hypertension diagnoses, nephroprotective medication use, ever-smoker status, and total cholesterol in cohort 1 participants with baseline eGFR_Cr <90 ml/min per 1.73 m² BSA (631 participants^a).

Supplemental Table 5. CKD-free survival time^a and CKD disease status of cohort 2 (1694 or 1353 participants^b) in total and across arterial stiffness categories based on baPWV at study enrollment.

Supplemental Table 6. eGFR_Cr, ΔeGFR_Cr, annualized ΔeGFR_Cr^a in cohort 1 (1271 participants) at study follow-up^b according to baseline arterial stiffness measured by baPWV.