Estimated pulse wave velocity and risk of new‐onset heart failure

Abstract

Aims

As a potential surrogate of carotid‐femoral pulse wave velocity, estimated pulse wave velocity (ePWV) has been confirmed to independently predict the cardiovascular events, but the association between ePWV and heart failure has not been well confirmed. Therefore, we performed this cohort study to evaluate the association between ePWV and risk of new‐onset heart failure.

Methods and results

A total of 98 269 employees (mean age: 51.77 ± 12.56 years, male accounted for 79.9%) without prior heart failure who participated in the 2006–2007 health examination were selected as the observation cohort, with an average follow‐up of 13.85 ± 1.40 years. Area under the receiver operator characteristic curve (AUC) of ePWV was calculated in prediction of heart failure. The adjusted Cox proportional hazard models were used to estimate hazard ratios and 95% confidence intervals. The category‐free net reclassification index (NRI) was calculated to evaluate the reclassification performance of cardiovascular risk models after adding ePWV. The AUC of ePWV was 0.74 in prediction of heart failure. After adjusting for the traditional cardiovascular risk factors except for age and blood pressure, the risk of new‐onset heart failure increased by 35% [hazard ratio (HR): 1.35, 95% confidence interval (CI): 1.33–1.37] for each 1 m/s increase in ePWV. Subgroup analysis showed that ePWV was significantly associated with incident heart failure regardless of THE presence (HR: 1.33, 95% CI: 1.31–1.36, P < 0.01) or absence (HR: 1.59, 95% CI: 1.46–1.73, P < 0.01) of cardiovascular risk factors, male (HR: 1.33, 95% CI: 1.31–1.36, P < 0.01) or female (HR: 1.44, 95% CI: 1.38–1.51, P < 0.01), young and middle‐aged (<52 years) (HR: 1.50, 95% CI: 1.41–1.58, P < 0.01), or middle‐aged and elderly (≥52 years) (HR: 1.23, 95% CI: 1.21–1.26, P < 0.01). The addition of ePWV to the traditional cardiovascular risk model including age and mean arterial pressure could significantly improve the reclassification ability by 31.1% (category‐free NRI = 0.311, P < 0.01).

Conclusions

ePWV was an independent predictor for new‐onset heart failure.

Introduction

It is estimated that there are 65.3 million people worldwide suffering from heart failure (HF), and the prevalence of HF is still rising. The prognosis of HF is very poor, and the 5 year survival rate after HF diagnosis is only 45.5%. ¹ Despite the development of medicines and devices, the mortality rate of HF remains high. ² Therefore, as upstream treatment, effective identification and management of modifiable risk factors is essential to reduce the morbidity and mortality of HF.

Pathophysiological studies showed that arteriosclerosis could lead to diastolic and systolic dysfunction in left ventricular by increasing the preload and afterload of the heart. ³ , ⁴ , ⁵ This was also confirmed by a few of epidemiological studies, which showed a significantly association between increased arterial stiffness and incident HF. ⁶ , ⁷ Carotid‐femoral pulse wave velocity (cfPWV) and brachial‐ankle pulse wave velocity (baPWV) are two non‐invasive and representative indicators of artery stiffness. Both of them have been confirmed to be independent risk factors for cardiovascular events including HF and can be used in cardiovascular risk assessment. ⁸ , ⁹ However, due to the high cost‐effectiveness ratio, poor comfort, and other reasons, the above two indicators have not been fully promoted.

Recent studies showed that estimated pulse wave velocity (ePWV), derived from age and blood pressure, had an excellent consistency with cfPWV, and could also predict cardiovascular events. It is expected to be used as an alternative indicator of PWV in the future. ¹⁰ , ¹¹ , ¹² However, up to now, only the multi‐ethnic study of atherosclerosis (MESA) has confirmed the association between ePWV and HF, ¹³ and there is a lack of research evidence from other countries and regions. Therefore, the current study aims to investigate the association between ePWV and new‐onset HF based on Chinese Kailuan cohort study.

Methods

The Kailuan study (registration number: ChiCTR‐TNC‐1100148) was a prospective community‐based cohort study that was performed in the community of Kailuan in the industrial city of Tangshan. The Ethics Committees of the Kailuan General Hospital confirmed that the study followed the guidelines of the Helsinki Declaration and approved it. All participants signed a written informed consent. Data, analytic methods, and study materials would be made available to other researchers upon request. The participants in the Kailuan study were employees and retirees of the Kailuan Group Company that was a coal mining industry in Tangshan. At baseline of the study between June 2006 and October 2007, we examined the study population of 101 510 individuals (81 110 men) with an age ranging between 18 and 98 years. We performed re‐examinations in 2 year intervals up to the end of the follow‐up on 31 December 2020 or up to the time at which an HF event occurred, whichever came first. In this study, we included those participants with a complete set of baseline data after an exclusion of those with prior HF and those with missing data of age, waist circumference, blood pressure, and fasting serum concentrations of triglycerides, high‐density lipoprotein, and glucose.

All study participants underwent an interview with a standardized questionnaire including questions on demographic, socioeconomic, and clinical parameters. We measured anthropometric parameters such as body height and weight. For the measurement of blood pressure, we obtained three readings at 5 min intervals after the participants had rested in a chair for at least 5 min. We used the average of three measurements for further data analysis. We obtained blood samples under fasting conditions and analysed them biochemically.

The main outcome event was new‐onset HF. All study participants were linked to the Municipal Social Insurance Institution and the Hospital Discharge Register, which allowed the detection of HF. To identify additional study participants with HF, we reviewed the discharge lists from the 11 Kailuan hospitals during the study period from 2006 to 2020, and we also asked the study participants at each re‐examination about a previous HF. For all suspected HF events, three experienced masked physicians reviewed the medical records and adjudicated. The diagnosis of HF was based on the Chinese Guidelines for the Diagnosis and Treatment of Heart Failure 2018, ¹⁴ including (i) clinical symptoms and signs of HF (such as dyspnoea, fatigue, and fluid retention), New York Heart Association cardiac function grade II, III, IV, or Killip cardiac function grade II, III, IV; (ii) echocardiography indicated left ventricular hypertrophy (left ventricular mass index ≥ 115 g/m² in men or ≥95 g/m² in women) and/or left atrial enlargement (left atrial volume index > 34 mL/m²) and/or abnormal diastolic function (E/e′ ≥ 13 or e′ < 9 cm/s); 3) Increased level of natriuretic peptide (NT‐proBNP ≥ 125 ng/mL or BNP ≥ 35 ng/mL). The diagnosis of HF must include condition (i) and one of condition (ii) or (iii).

We measured the baPWV using the BP‐203RPEIII network arteriosclerosis detection device (Omron Health Medical Co. Ltd., Dalian, China). The room temperature was kept between 22°C and 25°C. The participants reclined on a flat bed. The blood pressure cuffs were applied to the upper arms and ankles of the lower limbs. The upper arm cuff airbag marker was aligned to the brachial artery, and the bottom of the cuff was 2 to 3 cm distant from the elbow socket. The lower extremity cuff airbag sign was located on the medial side of the lower extremity, the lower cuff margin was 1 to 2 cm distant from the medial malleolus. After a rest for at least 5 min in the supine position, the measurements were performed. The measurements were repeated and the second value was used as the final result.

As described by Greve et al. in detail, we used for the calculation of the ePWV in individuals with cardiovascular risk factors the equation of ePWV = 9.58748315543126 − 0.402467539733184 × age + 4.56020798207263 × 10⁻³ × age² − 2.6207705511664 × 10⁻⁵ × age² × ;MBP + 3.1762450559276 × 10⁻³ × age × MBP − 1.83215068503821 × 10⁻² × MBP (MBP: mean blood pressure). For individuals without cardiovascular risk factors, we calculated the ePWV as ePWV = 4.62 − 0.13 × age + 0.0018 × age² + 0.0006 × age × MBP + 0.0284 × MBP. Individuals without cardiovascular risk factors were defined as non‐smokers without any components of a metabolic syndrome and without a history of myocardial infarction or stroke. ¹⁰

Statistical analysis was performed using the SAS software (Version 9.2, SAS Institute, Cary, NC, USA). The study population was divided into four groups according to the ePWV quartiles (Q1: <7.57 m/s, Q2: 7.57–8.78 m/s, Q3: 8.79–10.33 m/s, and Q4: ≥10.34 m/s). Baseline data for participants with and without HF were presented. Continuous variables with a normal distribution were expressed as the mean and standard deviation and compared using the independent sample t‐test. Due to the skewed distribution of C‐reactive protein and triglycerides, we performed a logarithmic transformation of these values and expressed them as median and quartiles. Categorical variables were described as percentages and compared using the Chi‐squared test. The Spearman correlation analysis was used to evaluate the correlation between ePWV and age and between ePWV and mean arterial pressure. Collinearity was analysed between ePWV and other covariates in a linear regression model. Cox proportional hazard models were used to examine the association of ePWV with new‐onset HF after adjustment for the following cardiovascular risk factors in the model: sex (male/female), smoking status (never and past, current, ≥1 cigarettes/day), drinking status (never and past, current, ≥1 time/day), education level (elementary school, high school or above), exercise (none, occasionally or frequently, ≥1 times/week), family history of cardiovascular disease (yes/no), history of myocardial infarction (yes/no), history of stroke (yes/no), body mass index, and fasting serum concentrations of total cholesterol, glucose, uric acid, and C‐reactive protein. After being stratified by gender, median age and presence or absence of cardiovascular risk factors, the Cox regression analysis was repeated. Area under the receiver operator characteristic curve (AUC) and the optimal cutpoint (corresponding to the maximum Youden index) of ePWV was calculated in prediction of HF. The study population was further divided into two groups according to the cutpoint of ePWV. Cox regression model was then used to compare the difference in the risk of HF between the two groups after adjustment for the above confounders. The reclassification performance of cardiovascular risk models after adding ePWV was evaluated using category‐free net reclassification index (NRI). In a following sensitivity analysis, the AUC of ePWV and baPWV for predicting HF was calculated and compared in those who completed baPWV measurement. Cox regression models were then used to evaluate the association of each 1 m/s increase in ePWV and baPWV with new‐onset HF after adjustment for the above confounders. All statistical tests were two‐sided, and P < 0.05 was considered statistically significant.

Results

Out of the total of 101 510 participants, the current study included 98 269 participants (mean age: 51.77 ± 12.56 years, male accounted for 79.9%) with a complete set of baseline data after excluding 79 participants with previous HF and 3162 participants with missing data of age, waist circumference, blood pressure, and fasting serum concentrations of triglycerides, high‐density lipoprotein, and glucose.

Table 1 showed the demographic and clinical differences between participants with and without HF development. Compared with participants without HF, age, the level of ePWV, blood pressure, body mass index, blood lipids, fasting blood glucose, uric acid, and C‐reactive protein were significantly higher in participants with HF (P for all <0.01). Also, the proportion of males and the prevalence of myocardial infarction and stroke were significantly higher in participants with HF than that in those without HF (P for all <0.01).

Table 1.

Baseline characteristics in participants with and without heart failure development

Variables	Participants with heart failure		t/χ 2	P
	No	Yes
	N = 95 041	N = 3228
Age, years	51.45 ± 12.49	61.39 ± 10.50	−52.51	<0.001
Men, n (%)	75 774 (79.73)	2761 (85.53)	65.56	<0.001
ePWV, m/s	9.05 ± 1.99	10.76 ± 1.96	−48.03	<0.001
SBP, mmHg	130.7 ± 20.88	142.5 ± 23.01	−28.77	<0.001
DBP, mmHg	83.41 ± 11.75	86.68 ± 12.31	−14.86	<0.001
MAP, mmHg	102.3 ± 14.51	109.0 ± 15.30	−24.46	<0.001
BMI, kg/m2	25.02 ± 3.48	25.89 ± 3.64	−13.29	<0.001
High school or above, n (%)	18 773 (19.75)	376 (11.65)	130.70	<0.001
Exerciser, n (%)	84 070 (88.46)	2849 (88.26)	0.12	0.73
Current drinker, n (%)	35 159 (36.99)	923 (28.59)	94.80	<0.001
Current smoker, n (%)	32 281 (33.97)	997 (30.89)	13.22	<0.001
History of MI, n (%)	1064 (1.12)	176 (5.45)	470.4	<0.001
History of stroke, n (%)	2273 (2.39)	213 (6.60)	224.1	<0.001
FHCVD, n (%)	5722 (6.02)	17 1 (5.30)	2.90	0.089
TG, mmol/L a	1.27 (0.89–1.92)	1.43 (1.01–2.16)	−10.01	<0.001
TC, mmol/L	4.94 ± 1.14	5.12 ± 1.27	−7.76	<0.001
LDL‐C, mmol/L	2.35 ± 0.91	2.39 ± 1.05	−2.22	0.027
HDL‐C, mmol/L	1.55 ± 0.40	1.57 ± 0.44	−2.91	0.004
Fbg, mmol/L	5.46 ± 1.64	6.08 ± 2.55	−13.69	<0.001
CRP, mg/L a	0.81 (0.30–2.30)	1.35 (0.55–3.60)	−19.49	<0.001
SUA, μmol/L	289.2 ± 83.59	314.7 ± 93.78	−15.19	<0.001

BMI, body mass index; CRP, C‐reactive protein; DBP, diastolic blood pressure; ePWV, estimated pulse wave velocity; Fbg, fasting blood glucose; FHCVD, family history of cardiovascular disease; HDL, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; MAP, mean arterial pressure; MI, myocardial infarction; SBP, systolic blood pressure; SUA, serum uric acid; TC, total cholesterol; TG, triglyceride.

^a Expressed as median (interquartile range).

After a mean follow‐up of 13.85 ± 1.40 years, 132, 426, 834, and 1836 participants in Q1, Q2, Q3, and Q4 groups of ePWV developed HF, respectively. With the increase of ePWV quartile groups, the incidence density of HF showed an increasing trend, which were 0.39/1000 person‐years, 1.24/1000 person‐years, 2.45/1000 person‐years, and 5.49/1000 person‐years, respectively (Table  2 ). Figure 1 showed that the cumulative incidence of HF increased with the ePWV quartile group (P for log‐rank <0.01).

Table 2.

Hazard ratios and 95% confidence intervals of heart failure

ePWV	Case/n	Incidence/1000 person‐years	HR (95% CI)	P
Each 1 m/s increase	—	—	1.35 (1.33–1.37)	<0.001
Each 1 m/s increase a	—	—	1.35 (1.33–1.38)	<0.001
Grouped by quartiles
Q1 (<7.57 m/s)	132/24 567	0.39	Ref
Q2 (7.57–8.78 m/s)	426/24 567	1.24	2.79 (2.29–3.40)	<0.001
Q3 (8.79–10.33 m/s)	834/24 568	2.45	5.05 (4.19–6.09)	<0.001
Q4 (≥10.34 m/s)	1836/24 567	5.49	10.30 (8.59–12.34)	<0.001
Grouped by the optimal cut‐off value
<9.42 m/s	880/60 559	1.04	Ref
≥9.42 m/s	2348/37 710	4.54	3.45 (3.18–3.74)	<0.001

CI, confidence interval; ePWV, estimated pulse wave velocity; HR, hazard ratio.

Model adjusted for sex, smoking status, drinking status, education level, exercise, family history of cardiovascular disease, history of MI, history of stroke, body mass index, total cholesterol, fasting blood glucose, serum uric acid and C‐reactive protein.

^a Excluding 1703 participants with atrial fibrillation at baseline and during the follow‐up.

Figure 1.

Kaplan–Meier curves for estimated pulse wave velocity (ePWV) quartile groups.

After adjusting for confounders except age and blood pressure, the risk of new‐onset HF in groups Q2, Q3, and Q4 was 2.79 (95% CI: 2.29–3.40, P < 0.01), 5.05 (95% CI: 4.19–6.09, P < 0.01), and 10.30 (95% CI: 8.59–12.34, P < 0.01) times higher than that in group Q1, respectively. Each 1 m/s increase in ePWV was associated with a 35% increase in the risk of HF (HR: 1.35, 95% CI: 1.33–1.37, P < 0.01). After excluding 1703 participants with atrial fibrillation at baseline and during the follow‐up, each 1 m/s increase in ePWV was still associated with incident HF (HR: 1.35, 95% CI: 1.33–1.38, P < 0.01) (Table  2 ).

The receiver operator characteristic curve analysis showed that the AUC of ePWV for predicting HF was 0.74 and the optimal cutpoint was 9.42 m/s (Fig. 2 ). The incidence density of HF in the group of ePWV ≥9.42 m/s was significantly higher than that in the group of ePWV <9.42 m/s, which was 4.54/1000 person‐years and 1.04/1000 person‐years, respectively. Figure 3 showed that the cumulative incidence of HF in the group of ePWV ≥9.42 m/s was significantly higher than that in the group of ePWV <9.42 m/s (P for log‐rank <0.01). After adjusting for confounders except age and blood pressure, the risk of HF in the group of ePWV ≥9.42 m/s was 3.45 (95% CI: 3.18–3.74, P < 0.01) times higher than that in the group of ePWV <9.42 m/s (Table  2 ).

Figure 2.

Receiver operator characteristic curve with the optimal cut‐off value of estimated pulse wave velocity (ePWV) for predicting heart failure. AUC, area under the receiver operator characteristic curve; ROC, receiver operator characteristic curve.

Figure 3.

Kaplan–Meier curves for group of estimated pulse wave velocity (ePWV) <9.42 m/s and group of ePWV ≥9.42 m/s.

Further stratified by the presence or absence of cardiovascular risk factors, gender, and median age, the results showed that each 1 m/s increase in ePWV was significantly associated with the occurrence of HF regardless of the presence (HR: 1.33, 95% CI: 1.31–1.36) or absence (HR: 1.59, 95% CI: 1.46–1.73) of cardiovascular risk factors, male (HR: 1.33, 95% CI: 1.31–1.36) or female (HR: 1.44, 95% CI: 1.38–1.51), young and middle‐aged (<52 years) (HR: 1.50, 95% CI: 1.41–1.58) or middle‐aged and elderly (≥52 years) (HR: 1.23, 95% CI: 1.21–1.26) (Table  3 ).

Table 3.

Hazard ratios and 95% confidence intervals of heart failure for each 1 m/s increase in ePWV in different subgroups

Subgroups	Case/n	Incidence/1000 person‐years	Each 1 m/s increase in ePWV
			HR (95% CI)	P
Participants without cardiovascular risk factors	149/12 982	0.83	1.59 (1.46–1.73)	<0.001
Participants with cardiovascular risk factors	3079/85 287	2.61	1.33 (1.31–1.36)	<0.001
Men	2761/78 535	2.54	1.33 (1.31–1.36)	<0.001
Women	467/19 734	1.71	1.44 (1.38–1.51)	<0.001
Age < 52 years	645/49 742	0.93	1.50 (1.41–1.58)	<0.001
Age ≥ 52 years	2583/48 527	3.87	1.23 (1.21–1.26)	<0.001

CI, confidence interval; ePWV, estimated pulse wave velocity; HR, hazard ratio.

Model adjusted for sex, smoking status, drinking status, education level, exercise, family history of cardiovascular disease, history of MI, history of stroke, body mass index, total cholesterol, fasting blood glucose, serum uric acid, and C‐reactive protein.

Pearson correlation analysis showed that ePWV was strongly correlated with both age (r = 0.85, P < 0.01) and mean arterial pressure (r = 0.68, P < 0.01) (Supporting information, Table S1 ). In addition, ePWV also showed a significant collinearity with age and mean arterial pressure in a linear regression model (Table  S2 ). Therefore, age and mean arterial pressure were not adjusted in above multivariate analyses in this study. To make up for that, we further evaluated the improvement after adding ePWV to the traditional cardiovascular risk model. The results showed that adding ePWV to the traditional cardiovascular risk model excluding age and mean arterial pressure improved the AUC by 12% and the reclassification ability by 58.3% (category‐free NRI = 0.583, P < 0.01). The addition of ePWV to the traditional cardiovascular risk model including age and mean arterial pressure did not change the AUC significantly, but the reclassification ability was improved by 31.1% (category‐free NRI = 0.311, P < 0.01) (Table  4 ).

Table 4.

Area under the receiver operator characteristic curve and improvement in models with and without estimated pulse wave velocity for heart failure

Index		Heart failure
	ePWV added	No	Yes
AUC (95% CI)	Model 1	0.680 (0.670–0.689)	0.761 (0.754–0.769)**
	Model 2	0.768 (0.760–0.775)	0.767 (0.760–0.775)
Category‐free NRI (95% CI)	Model 1	—	0.583 (0.549–0.617)**
	Model 2	—	0.311 (0.278–0.343)**

AUC, area under the receiver operator characteristic curve; ePWV, estimated pulse wave velocity; NRI, net reclassification index.

Model 1 including sex, smoking status, drinking status, education level, exercise, family history of cardiovascular disease, history of MI, history of stroke, body mass index, total cholesterol, fasting blood glucose, serum uric acid, and C‐reactive protein.

Model 2 including age, mean arterial pressure and all the factors in model 1.

^** P < 0.001.

In order to compare the predictive value of ePWV and baPWV for new‐onset HF, we conducted a sensitivity analysis in which 34 406 participants without previous HF who completed baPWV measurement from 2010 to 2017 and had complete baseline data were included. During an average follow‐up of 6.19 years, 408 participants developed HF, with an incidence density of 1.91/1000 person‐years. The scatter plot showed a significant linear association between baPWV and ePWV (R ² = 0.43) (Fig. 4 ). The AUC of ePWV for predicting HF was significantly higher than that of baPWV (0.83 vs. 0.75, P < 0.01). After adjusting for confounders, each 1 m/s increase in ePWV and baPWV increased the risk of HF by 41% (HR: 1.41, 95% CI: 1.37–1.45, P < 0.01) and 16% (HR: 1.16, 95% CI: 1.14–1.18, P < 0.01), respectively (Table  5 ).

Figure 4.

Scatter plot with brachial‐ankle pulse wave velocity and estimated pulse wave velocity. baPWV, brachial‐ankle pulse wave velocity.

Table 5.

Hazard ratios and 95% confidence intervals of each 1 m/s increase in ePWV or baPWV for heart failure

Parameter	ePWV	baPWV
AUC (95% CI)	0.83 (0.81–0.85)	0.75 (0.72–0.77) b
HR (95% CI) a	1.41 (1.37–1.45) c	1.16 (1.14–1.18) c

AUC, area under the receiver operator characteristic curve; baPWV, brachial‐ankle pulse wave velocity; ePWV, estimated pulse wave velocity.

^a Adjusted for sex, smoking status, drinking status, education level, exercise, family history of cardiovascular disease, history of MI, history of stroke, body mass index, total cholesterol, fasting blood glucose, serum uric acid and C‐reactive protein.

Compared with ePWV.

^b P < 0.01.

^c P < 0.01.

Discussion

The current study found that ePWV had an excellent predictive power for HF. After adjusting for traditional cardiovascular risk factors except age and blood pressure, the risk of new‐onset HF in the second, third and fourth quartile of ePWV was 2.79, 5.05, and 10.30 times higher than that in the first quartile, respectively. For each 1 m/s increase in ePWV, the risk of HF increased by 35%. In addition, subgroup analysis also showed that baseline ePWV was significantly associated with the occurrence of HF regardless of presence or absence of cardiovascular risk factors, male or female, young and middle‐aged, or middle‐aged and elderly. Moreover, the addition of ePWV to the traditional cardiovascular risk model including age and mean arterial pressure could significantly improve the reclassification ability by 31.1%. These results suggested that ePWV was an independent predictor for new‐onset HF.

Consistent with our study results, Heffernan et al. ¹³ followed 6814 middle‐aged and older adults (age: 45–84 years) from MESA for an average of 12.5 years and found that the risk of HF in the second, third and fourth quartile group of ePWV was 2.02 (95% CI: 1.16–3.53), 2.81 (95% CI: 1.55–5.09) and 4.79 (95% CI: 2.43–9.45) times greater than that in the first quartile group, respectively, after adjusting for age, sex, body mass index, smoking, usage of antihypertensive drugs, and waist circumference. After HF was further divided into HF with reduced ejection fraction (HFrEF) and HF with preserved ejection fraction (HFpEF), the results showed that the risk of HFrEF in the second, third, and fourth quartile groups of ePWV was 2.50 (95% CI: 1.22–5.16), 4.35 (95% CI: 2.17–8.73), and 8.37 (95% CI: 4.24–16.52) times greater than that in the first quartile group, respectively. The risk of HFpEF in the fourth quartile group of ePWV was 3.94 (95% CI: 1.39–11.17) times higher than that in the first quartile group. But there was no difference in the risk of HFpEF between the second or the third quartile group and the first quartile group. The results of the above study, which included a multiracial population and analysed HFrEF and HFpEF separately, were more robust and convincing. However, due to the small sample size and the small number of HF events in that study, the results may be underestimated when the analysis of HF subtypes was performed. In addition, the above study adjusted for age in the multivariate analyses, which may cause overadjustment. However, even so, the study still reached a positive conclusion, further indicating that ePWV was a strong predictor of HF.

Because ePWV is derived from age and mean arterial pressure, the collinearity between ePWV and age and blood pressure is inevitable. In most studies ¹⁰ , ¹¹ , ¹² about ePWV, researchers did not adjust for age and blood pressure in multivariate models. Instead, only risk scores (e.g. Framingham risk score and systematic coronary risk evaluation) calculated from traditional cardiovascular risk factors were adjusted. Of those studies, Vishram‐Nielsen ¹² analysed the association of ePWV with cardiovascular events and found that ePWV predicted cardiovascular events independently of Framingham risk score and systematic coronary risk evaluation, but not independently of traditional cardiovascular risk factors including age and systolic blood pressure. This may be explained by the significant collinearity of ePWV with age and blood pressure.

The predictive ability of the model is rarely affected by collinearity; we therefore evaluated the improvement after adding ePWV to the traditional cardiovascular risk model. The results showed that adding ePWV to the traditional cardiovascular risk model including age and mean arterial pressure did not change the AUC significantly, but the reclassification ability was improved by 31.1%. It suggested that ePWV could predict HF independently of the traditional cardiovascular risk factors including age and blood pressure.

Subgroup analysis showed that the association between ePWV and new‐onset HF was more significant in young and middle‐aged, female, and participants without obvious cardiovascular risk factors. It was consistent with the findings of Ohkuma et al. ⁸ in which the association between baPWV and cardiovascular events was more pronounced in the population with a low cardiovascular risk compared with those with a high cardiovascular risk. It may be related to the different pathogenic pathways of cardiovascular events. Endothelial dysfunction and vascular inflammation, rather than arterial stiffness, were the major pathogenic pathways leading to cardiovascular events in individuals at high cardiovascular risk. ¹⁵ Therefore, the risk assessment of HF was equally important for those at high or low cardiovascular risk. The use of simple and effective ePWV to identify high‐risk groups and carry out early intervention may reduce more HF patients and save more medical resources.

In the further sensitivity analysis, baPWV was also found to be an independent risk factor for HF. Each 1 m/s increase in baPWV was associated with a 16% increase in the risk of HF after adjustment for confounders. It was consistent with previous conclusions. ⁷ However, compared with baPWV, ePWV seemed to have a stronger association with HF (HR: 1.41 vs. 1.16) and better predictive power for HF (AUC: 0.83 vs. 0.75). Similar results were found in studies with cardiovascular events as endpoints. ¹⁰ , ¹⁴ The apparent superiority of ePWV over baPWV strongly suggests that age and blood pressure are the main determinants of HF development. In addition, we have reason to suspect that ePWV may carry additional risk information, such as a complex interaction between age and blood pressure.

The pathophysiological mechanisms of arterial stiffness leading to HF have not been fully elucidated. Possible mechanisms were as follows: first, increased arterial stiffness accelerated the conduction velocity of the reflected wave, which led to an earlier return of the wave to the aortic root during mid‐late systole (rather than diastole). Systolic blood pressure and left ventricular afterload were thus elevated. The increase of cardiac afterload could cause the imbalance of myocardial oxygen supply and demand and the thickening of the ventricular wall. Such long‐term changes could lead to impaired diastolic and systolic function of the heart, and eventually caused symptomatic HF. ³ Second, arteriosclerosis could lead to increased peripheral resistance, retention of water and sodium, inflammation and oxidative stress by causing endothelial dysfunction, and excessive activation of the renin–angiotensin–aldosterone system and immune system. These neurohumoral changes could increase the afterload and preload of the heart and promote the development of HF. ⁴ , ⁵

This study had certain clinical significance. First, in contrast to PWV, which require expensive measuring equipment, ePWV could be calculated from age and blood pressure alone. The application of ePWV would therefore greatly reduce the medical and economic burden of the country and individuals. However, further large‐scale multi‐centre studies were needed to establish a population‐based calculation formula of ePWV. Second, as an independent risk factor for HF, ePWV could be used to quickly identify the individuals at high risk. While the risk of HF could be minimized by early active control of blood pressure and other accompanying risk factors, such as overweight or obesity, hyperlipidaemia, hyperglycaemia, hyperuricaemia, and inflammation.

This study also had some limitations. First, the study population was predominantly male, and the results may not be generalizable to other populations. However, gender subgroup analyses were performed, and the results were not significantly different. Second, valvular heart disease and atrial fibrillation are two important confounders affecting cardiac function. However, the results were not changed significantly after excluding participants with atrial fibrillation while we cannot exclude the influence of valvular heart disease on the results due to the lack of relevant data. Third, due to the limited outcome information, the diagnosis of HF was imprecise and separate analysis for HFrEF and HFpEF was not possible. However, only one similar study ¹³ had been conducted so far, and the results of these studies can serve as a basis for further research.

In conclusion, in this adult Chinese community, the ePWV was found to significantly associate with the risk of new‐onset HF, independently of traditional cardiovascular risk factors.

Although there is increasing evidence on ePWV, ePWV is still an estimated value and cannot replace PWV. It is limited to a temporary replacement and assistance in screening people at high cardiovascular risk when PWV detection is not available. Besides the gold standard, but high cost‐effective carotid‐femoral PWV technique recent emerging methodologies exist (cardiac MRI PWV, user‐friendly oscillometric PWV). Rather further studies are needed in HF to investigate the thorough task of arterial stiffness and PWV‐guided diagnostic strategies.

Funding

None.

Conflict of interest

None declared.

Supporting information

Table S1. Correlation analysis of estimated pulse wave velocity with age and blood pressure.

Table S2. Collinearity between the estimated pulse wave velocity and other covariates in a linear regression model with the incidence of heart failure as the dependent variable.

Ji, C. , Wang, G. , Huang, Z. , Zhu, C. , and Liu, Y. (2024) Estimated pulse wave velocity and risk of new‐onset heart failure. ESC Heart Failure, 11: 2120–2128. 10.1002/ehf2.14778.