From short‐term blood pressure variability to atherosclerosis: Relative roles of vascular stiffness and endothelial dysfunction

Abstract

Both arterial blood pressure (BP) average levels and short‐term BP variability (BPV) relate to hypertension‐mediated organ damage, in particular increased carotid artery intima‐media thickness (IMT) and carotid‐femoral pulse wave velocity (PWV). Endothelial dysfunction possibly mediates such damage. The authors aimed at further investigating such role in hypertensive patients. In 189 recently diagnosed, untreated hypertensive patients the authors evaluated, in a cross‐sectional design, the relationships of BP average levels and short‐term systolic (S) BPV (standard deviation of awake SBP or of 24‐hour‐weighted SBP) with IMT and PWV, and how much these relationships are explained by endothelial function parameters—brachial artery flow‐mediated dilation (FMD) and digital reactive hyperemia index (RHI). Multivariable models assessed the strength of these relationships to derive a plausible pathogenetic sequence. Both average SBP values and our measures of SBPV were significantly related to IMT (24‐hour mean SBP: r = .156, P = .034; 24‐hour‐weighted SBPV: r = .157, P = .033) and to PWV (24‐hour mean SBP: r = .179, P = .015; 24‐hour‐weighted SBPV: r = .175; P = .018), but only poorly related to FMD or RHI (P > .05 for all). At univariable regression analysis, FMD and RHI were both related to IMT, (P < .001), but not to PWV. When FMD and RHI were added to average SBP and SBPV parameters in a multivariable model, both significantly (P < .005) contributed to predict IMT, but not PWV. Thus, endothelial dysfunction relates to IMT independently of BP parameters, but appears to play a minor role in the association between BP variability‐related variables and arterial stiffening.

1. INTRODUCTION

Hypertension causally determines an increased risk of cardiovascular events, ¹ , ² which may be preceded—and therefore predicted—by the occurrence of hypertension‐mediated organ damage (HMOD, previously referred as target‐organ damage). One type of HMOD is manifested as an increased carotid artery intima‐media thickness (IMT), commonly considered a marker of early atherosclerosis ³ and shown to be a relevant predictor of cardiovascular events, ⁴ , ⁵ although its additional value on top of classical risk factors has been questioned. ⁶ In addition to average blood pressure (BP) levels, also BP variability (BPV) correlates with—and has been reported to be a determinant of—HMOD, including carotid artery IMT, independent of average BP levels. ⁷ , ⁸ Hypertension is also associated with increased arterial stiffness, commonly assessed by the pulse wave velocity (PWV), ⁹ , ¹⁰ which is itself an important predictor of outcomes. ¹¹ , ¹² , ¹³ An increased arterial stiffness was also found to be associated with a greater systolic (S) BPV. ¹⁴ , ¹⁵ , ¹⁶

Endothelial dysfunction, physiologically characterized by a reduced bioavailability of endothelium‐derived nitric oxide (NO), has a putative role in the very early phases of atherosclerosis, ¹⁷ , ¹⁸ and hypertension is also associated with systemic endothelial dysfunction, ¹⁹ which may be considered a general transducing mechanism for vascular disease, predicting progression of preclinical carotid arterial disease and possibly more closely related to atherosclerotic changes than conventional risk factors. ²⁰

Whether the effects of BPV on vascular HMOD (specifically IMT and PWV) are mediated by changes in endothelial function or are rather due to more direct effects on the smooth muscle cells or the fibrous component of the arterial wall is still unclear. We attempted to gain insight into such relationships by testing the strength of associations of short‐term BPV with in vivo parameters of endothelial function (the brachial artery vascular reactivity and digital pulse amplitude in response to hyperemia) on the one hand, and with IMT and vascular stiffness (assessed through the PWV) on the other hand, in order to construct a plausible sequence of events from high BP and high BPV to vascular damage. We studied this in a population of previously untreated, uncomplicated hypertensive patients.

2. METHODS

An extended version of the Methods is provided in the Supporting information.

2.1. Study population

We analyzed data from 239 consecutive community‐dwelling hypertensive patients, recently diagnosed (1‐24 months, median 6 months) by clinic BP measurement according to standard criteria (see Extended Methods). We then excluded patients with conditions associated with changes in autonomic nervous system activity potentially able to influence BPV over 24 hours, as well as endothelial function, an effect specifically relevant for this study (see Extended Methods in the Supporting information). The final study population consisted therefore of 189 patients, in whom we assessed traditional risk factors, such as family history of hypertension, tobacco smoking, coffee or alcohol consumption and level of physical activity (regular physical activity being defined as aerobic exercise on a regular basis 3‐to‐4 times per week ²¹ ).

2.2. Ambulatory blood pressure monitoring and assessment of blood pressure variability

All patients underwent ambulatory BP monitoring (ABPM) with a validated oscillometric device (SpaceLabs 90207; Monitor Inc). ABPM was performed on a working day, with the subjects involved in their usual daily activities. We obtained BP and heart rate readings every 15 minutes during daytime (awake period, between 7 AM and 11 PM) and every 30 minutes during nighttime (asleep period, between 11 PM and 7 AM). We instructed subjects to take note of their activities and of the time of retiring to bed in a diary.

Short‐term BPV was estimated separately for the daytime and the nighttime as the standard deviation (SD) of mean systolic (S) BP and diastolic (D) BP over the awake or asleep time periods (which, based on patients’ diaries, coincided with the predefined daytime and nighttime subperiods, respectively). Short‐term BPV was also assessed as weighted 24‐hour SBP SD and weighted 24‐hour DBP SD, that is, as the average of daytime and nighttime SBP or DBP SD, separately computed and weighted for the duration of the daytime and the nighttime period, respectively, as previously reported. ²² Because of our previous reports showing better correlations of SBP and SBPV—as compared to DBP and DBPV—with HMOD, ⁷ , ⁸ also confirmed in our present study, we have here focused our analyses and discussion of BPV parameters mainly on daytime and 24‐hour‐weighted SBP and 24‐hour‐weighted SBPV.

2.3. Carotid ultrasonography

We performed ultrasound examinations of the common carotid artery, bulb, and internal carotid artery bilaterally, using a Philips EnVisor C apparatus (Philips Electronics) equipped with a 10‐MHz linear‐array transducer, according to our previous study ⁷ (see Extended Methods for further details).

2.4. Assessment of brachial artery flow‐mediated dilation

We performed measurements of brachial artery diameters and blood flow velocity using a Doppler ultrasound machine (Philips EnVisor C) equipped with a high‐resolution (10 MHz) linear‐array transducer with subjects at rest in a supine position. We used a customized holding device to secure the transducer in place. We acquired a longitudinal image of the brachial artery 5‐10 cm proximal to the antecubital fossa, and here we performed measurements of baseline blood flow and diameter. After the acquisition of baseline measurements, we interrupted blood flow with an occluding cuff placed on the forearm and inflated for 5 minutes at supra‐systolic (200‐250 mm Hg) BP values. After cuff deflation, we obtained ultrasound‐derived measurements of the brachial artery diameters and blood flow for 3 minutes. We measured the distance between the leading edges of interfaces with ultrasonic calipers. The transition between different acquisition modes was instantaneous and facilitated by the built‐in pre‐set software in our ultrasound machine. The same two investigators (AT and RT) recorded and analyzed all ultrasound brachial images. We calculated flow‐mediated dilation (FMD) using the following formula:

$$FMD=\left(maximumdiameter-baselinediameter\right)/baselinediameter\times 100.$$

2.5. Determination of digital pulse amplitude

Digital pulse vascular amplitude (PVA) induced by reactive hyperemia (RH) is here defined as the RH index (RHI), and was measured with a fingertip pulse amplitude tonometry (PAT) device, as a relatively newer method for assessing vascular function. ²³ , ²⁴ We measured digital PVA in the fasting state with a PAT device (Endo‐PAT2000; Itamar Medical) placed on the tip of each index finger. Further details of the related methodology are reported in the Supporting information.

2.6. Arterial stiffness analysis

We assessed arterial stiffness using the PulsePen^® tonometer (DiaTecne Srl.), which is a validated device for the non‐invasive automatic evaluation of the PWV. ²⁵ For details of methods here used, see the Extended Methods in the Supporting information. We here set the distance of measurement as the suprasternal‐notch‐to‐femoral‐artery distance minus the carotid‐artery‐to‐suprasternal‐notch distance, using a tape measure of the distance between the suprasternal notch and each of the two sites where the tonometer probe was applied.

2.7. Protocol

We performed all studies in a temperature‐controlled (22‐26°C) vascular laboratory by two trained operators (AT and RT) after a 15‐minute run‐in rest. We assessed the brachial artery FMD and digital PVA‐RH measurements in two separate days, in comparable conditions to obtain independent values. To this purpose, we performed the baseline assessment after a 12 hours overnight fast at 9 AM, in resting recording conditions, with subjects refraining from physical exercise for the previous 24 hours. ²⁶ We asked patients to withhold vasoactive medications (angiotensin‐converting enzyme inhibitors, angiotensin receptor blockers, calcium channel blockers, beta‐blockers, and nitrates), ²⁷ as well as alcohol, for 48 hours prior to the visit. ²⁸ We asked subjects to refrain from both tobacco smoking and tobacco smoke exposure, ²⁹ , ³⁰ and not to assume caffeine‐containing drinks ³¹ for at least 12 hours before measurements. We also asked subjects to abstain from vitamin supplementation for ≥72 hours before the assessments. ³² Finally, when studying pre‐menopausal women we had care to perform all measurements at the same time of the menstrual cycle, as previously described. ³³ All vascular measurements here investigated were performed within 30 days of blood pressure monitoring. Informed consent was obtained from each patient, and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki. For its correlative nature and the use of non‐invasive measurements also considered useful for the clinical assessment and management of patients, the protocol was not, however, submitted to a priori approval by the institution's human research committee.

2.8. Statistical methods

We expressed continuous variables as mean ± standard deviation (SD) and qualitative variables as percentages. We performed a preliminar power analysis for testing the relationship between short‐term BPV with parameters of endothelial function and with IMT and vascular stiffness. The power of the study was 90% considering 188 patients, with four predictors, an effect size of 0.4 and an alpha error rate of 0.05.

We performed univariable linear regression analysis between baseline clinical, demographic and conventional BP monitoring variables (the latter including awake, asleep, and 24 hours average SBP and DBP—as independent variables), and indices of HMOD (IMT, FMD, PWV, and RHI—as dependent variables); results are expressed as Pearson's correlation coefficient (r) and relative 95% confidence interval (95% CI). To assess the incremental contribution of BPV to determining HMOD, we next performed multivariable linear regression analysis between measures of vascular HMOD (IMT and PWV) as dependent variables, and BP variables resulting significantly related to IMT and PWV at univariable analysis (independent variables). By uni‐ and multivariable linear regression analysis, we further analyzed the relationship between indices of endothelial dysfunction (FMD, RHI, putative intermediate end points in the pathogenetic chain of events linking BPV with HMOD, as independent variables) with those of vascular stiffness (PWV) or IMT, as dependent variables. We assessed the potential presence of collinearity in the final models and excluded variables affected by collinearity (tolerance: 0.721‐0.990; or variance inflation factor [VIF]: 1.010‐1.387). We visually inspected each model for linearity, heteroscedasticity, and normality of the residuals.

Finally, we estimated the contribution of parameters of endothelial dysfunction (RHI and FMD) to 24‐hour‐weighted SBP variability‐related parameters of vascular organ damage (IMT and PWV) using mediation analysis. ³⁴ Here, a mediation model is any causal system in which at least one causal antecedent X variable is proposed as influencing an outcome Y through a single intervening variable M. ³⁴ We here used linear regression due to the continuous nature of the dependent variable. We performed this specific analysis using the SPSS PROCESS computing tool version 3.4 (model 6), with two mediators. We estimated unstandardized regression coefficients (B) using the bootstrapping (BCa) procedure (5000 resamples) that results in a 95% corrected bias and direct and indirect effect confidence intervals (BCa). We computed the proportion caused by the mediator with the formula: indirect effect/total effect.

For all tests, we set the threshold for statistical significance at P < .05. We performed statistical analyses with the aid of the SPSS release 18.0 statistical software (SPSS Inc Headquarters), unless otherwise specified.

3. RESULTS

Patients’ age (mean ± SD) was 58.4 ± 9.0 years; 60% of patients were male, and the mean BMI was 27.0 ± 4.1 kg/m². Demographic, anthropometric, ambulatory BP monitoring‐derived and HMOD‐related variables are listed in Table 1.

TABLE 1.

Baseline demographic, blood pressure parameters, and indices of hypertension‐mediated vascular organ damage in the overall study population

Variable grouping	Variable
Demographic and anthropometric	Age (y)	58 ± 9
	Sex (M) (%)	60
	BMI (kg/m2)	27.0 ± 4.1
	Family history of hypertension (%)	50
	Dyslipidemia (%)	48
	Smoking (%)	19
	Physical activity (%)	11
	Coffee drinking (%)	78
Blood pressure monitoring‐derived	Awake systolic BP (mm Hg)	131.4 ± 10.8
	Awake diastolic BP (mm Hg)	80.7 ± 8.1
	Awake Mean BP (mm Hg)	97.8 ± 8.4
	Awake systolic BP SD (mm Hg)	11.1 ± 2.5
	Awake diastolic BP SD (mm Hg)	9.1 ± 2.2
	HR (bpm)	75.4 ± 0.10.3
	Asleep systolic BP (mm Hg)	120.0 ± 11.6
	Asleep diastolic BP (mm Hg)	69.8 ± 8.7
	Asleep systolic BP SD (mm Hg)	10.4 ± 3.7
	Asleep diastolic BP SD (mm Hg)	8.9 ± 2.9
	24‐h systolic BP (mm Hg)	127.8 ± 10.8
	24‐h diastolic BP (mm Hg)	77.2 ± 7.8
	24‐h systolic BP SD (mm Hg)	12.5 ± 2.9
	24‐h diastolic BP SD (mm Hg)	10.5 ± 2.4
	24‐h‐weighted systolic BP (mm Hg)	128 ± 10.6
	24‐h‐weighted diastolic BP (mm Hg)	78.5 ± 8
	24‐h‐weighted systolic BP SD (mm Hg)	10.8 ± 2.9
	24‐h‐weighted diastolic BP SD (mm Hg)	8.6 ± 1.9
Hypertension‐mediated vascular organ damage‐related	FMD (%)	11.24 ± 9.72
	RHI	1.735 ± 0.549
	PWV (m/s)	9.28 ± 3.66
	IMT (mm)	0.076 ± 0.013

Variables are expressed as mean ± SD for continuous variables and as percentage for categorical variables.

Abbreviations: BMI, body mass index; BP, blood pressure; FMD, flow‐mediated dilation; HR, heart rate; IMT, intima‐media thickness; M, male; PWV, pulse wave velocity; RHI, reactive hyperemia index; SD, standard deviation; y, years.

3.1. Impact of systolic blood pressure average values and systolic blood pressure variability on organ damage

Table 2 summarizes the relationships between all ABPM‐derived variables and all HMOD variables here explored (IMT, FMD, RHI, and PWV) through univariable linear regression analysis.

TABLE 2.

Univariable linear regression between ambulatory blood pressure variables (as independent variables) and parameters of hypertension‐mediated vascular organ damage (IMT and PWV) or endothelial dysfunction (FMD and RHI)

	Parameters of hypertension‐mediated vascular organ damage				Parameters of endothelial dysfunction
	PWV (m/s)		IMT (mm)		FMD (%)		RHI
	r (95% CI)	P‐value	r (95% CI)	P‐value	r (95% CI)	P‐value	r (95% CI)	P‐value
Parameters of Mean BP (mm Hg)
Awake systolic	0.175 (0.033; 0.310)	.017	0.208 (0.067; 0.341)	.004	−0.083 (−0.223; 0.061)	.254	−0.024 (−0.167; 0.120)	.747
Awake diastolic BP	0.077 (−0.067; 0.218)	.297	−0.094 (−0.234; 0.050)	.201	−0.064 (−0.205; 0.080)	.381	−0.125 (−0.263; 0.018)	.089
Asleep systolic BP	0.204 (0.063; 0.337)	.006	0.089 (−0.055; 0.229)	.229	−0.019 (−0.162; 0.124)	.793	−0.156 (−0.293; 0.013)	.033
Asleep diastolic BP	0.054 (−0.090; 0.196)	.471	−0.086 (−0.226; 0.058)	.244	−0.009 (−0.152; 0.134)	.898	−0.180 (−0.315; −0.038)	.014
24‐h systolic BP	0.179 (0.037; 0.314)	.015	0.156 (0.013; 0.293)	.034	−0.076 (−0.127; 0.068)	.304	−0.053 (−0.195; 0.091)	.473
24‐h diastolic BP	0.046 (−0.098; 0.188)	.537	−0.106 (−0.245; 0.038)	.148	−0.058 (−0.199; 0.086)	.429	−0.159 (−0.295; −0.016)	.030
24‐h‐weighted SBP (mm Hg)	0.262 (0.124; 0.390)	<.001	0.169 (0.027; 0.305)	.020	−0.017 (−0.160; 0.126)	.82	−0.033 (−0.175; 0.111)	.654
24‐h‐weighted DBP (mm Hg)	0.259 (0.120; 0.388)	.001	0.003 (−0.140; 0.146)	.967	0.040 (−0.104; 0.182)	.601	−0.028 (−0.170; 0.116)	.714
Parameters of BP variability (mm Hg)
Awake systolic BP variability	0.236 (0.096;0.367)	.001	0.150 (0.007; 0.287)	.041	−0.087 (−0.227;0.057)	.236	0.091 (−0.053; 0.231)	.213
Awake diastolic BP variability	0.152 (0.009; 0.289)	.040	0.019 (−0.124; 0.162)	.797	−0.074 (−0.215; 0.070)	.313	0.118 (−0.026; 0.257)	.106
Asleep systolic BP variability	0.149 (0.006; 0.286)	.045	0.182 (0.040; 0.317)	.013	0.043 (−0.101; 0.185)	.561	0.089 (−0.055; 0.229)	.225
Asleep diastolic BP variability	00.149 (0.006; 0.286)	.045	0.148 (0.005; 0.285)	.044	−0.113 (−0.252; 0.031)	.122	0.065 (−0.079; 0.206)	.377
24‐h systolic BP variability	0.117 (−0.027; 0.256)	.115	0.115 (−0.029; 0.254)	.117	−0.004 (−0.147; 0.139)	.957	0.181 (0.039; 0.316)	.013
24‐h diastolic BP variability	0.060 (−0.084; 0.201)	.419	−0.002 (−0.145; 0.141)	.984	−0.134 (−0.272; 0.009)	.067	0.129 (−0.014; 0.267)	.079
24‐h‐weighted SBP variability	0.175 (0.033; 0.310)	.018	0.157 (0.014; 0.294)	.033	−0.013 (−0.156; 0.130)	.859	0.004 (−0.139; 0.147)	.961
24‐h‐weighted DBP variability	0.211 (0.070; 0.344)	.005	−0.053 (−0.195; 0.091)	.468	−0.052 (−0.194; 0.092)	.492	−0.028 (−0.170; 0.116)	.732

Significant P‐values are evidenced in bold.

Abbreviations: 95% CI, 95% confidence interval; BP, blood pressure; D, diastolic; FMD, flow mediate dilation; IMT, intima‐media thickness; PWV, pulse wave velocity; r, Pearson's correlation coefficient; RHI, reactive hyperemia index; S, systolic.

Awake SBP values were significantly correlated with PWV and IMT. Asleep SBP was significantly correlated with PWV. More robust correlations with those parameters of HMOD were found for 24‐hour‐weighted SBP than for asleep SBP (Table 2).

More parameters of BP variability than of average BP values were found significantly related to PWV and IMT (Table 2). In particular, both awake SBPV and 24‐hour‐weighed SBPV were significantly related to IMT (P = .041 and P = .033, respectively) and PWV (P = .001 and P = .018, respectively). Also, asleep SBPV was related to both PWV and IMT (Table 2).

Neither average BP values nor BPV values were related to FMD, while they showed weak relationships with RHI. Notably, among all BP parameters, only 24‐hour‐weighted SBP variability was weakly related to RHI (P = .013).

3.2. Impact of diastolic blood pressure variability on organ damage

Also, awake DBP variability and 24‐hour‐weighted DBP variability appeared to be significantly related with PWV, but not with IMT. Asleep DBP variability was significantly related to both parameters of HMOD. Because of the overall weaker relationships with parameters of HMOD than for corresponding systolic values, such DBP variability parameters were not further tested in multivariable analyses.

3.3. Multivariable analysis models to postulate the pathogenetic sequence of events

In multivariable models, we focused on awake SBPV and on 24‐hour‐weighted SBPV only, disregarding asleep SBPV, in line with the results of a previous study of ours ⁷ showing a stronger correlation of the former two variables than of the latter with HMOD, and because also in the present cohort the association of asleep BPV estimates with organ damage was overall weaker than for awake BPV, and in any case incorporated in the 24‐hour‐weighted SBP‐SD estimates. Since parameters of SBP variability appeared to be related to both IMT and PWV—both markers of HMOD—but not, or only to a marginal extent, to indices of endothelial dysfunction (FMD and RHI), we tested the hypothesis of a direct relationship of FMD and RHI with IMT, as well as that of a direct relationship of PWV—a marker of arterial stiffness—with IMT. Results of univariable and multivariable linear regression analyses related to these hypotheses are reported in Tables 3 and 4. Univariable regression models showed that both parameters of endothelial dysfunction (FMD and RHI) appeared significantly and negatively related to IMT (P = .001 for both) (Table 3, Models 1 and 3). We therefore next entered all such parameters in multivariable regression models aimed at improving the prediction of IMT (dependent variable), further adding the 24‐hour‐weighted SBPV (Table 3, models 2, 4, and 5). The added contribution of the 24‐hour‐weighted SBPV in predicting IMT was significant in models that included parameters of endothelial function (P = .031, P = .030, and P = .027, respectively).

TABLE 3.

Univariable and multivariable regression models of the relationship of IMT (as dependent variable) as a function of 24‐h‐weighted SBP variability, FMD, and RHI, alone or in combination (independent variables)

	Independent factor	Standardized b (95% CI)	P‐value	Adjusted R 2 for the model
Model 1	FMD	−0.251 (−0.380; −0.112)	.001	.063
Model 2	FMD	−0.257 (−0.385; −0.119)	<.001	.091
	24wSBP	0.153 (−0.289; −0.011)	.031
Model 3	RHI	−0.241 (−0.371; −0.102)	.001	.058
Model 4	RHI	0.242 (0.103; 0.372)	.001	.083
	24wSBP	0.156 (0.014; 0.292)	.030
Model 5	RHI	−0.203 (−0.336; −0.062)	.005	.117
	FMD	−0.223 (−0.354; −0.083)	.002
	24wSBP	0.155 (0.013; 0.291)	.027

All models are adjusted for 24‐h‐weighted SBP. Statistically significant values in bold.

Abbreviations: 24wSBP, 24‐h blood pressure variability as standard deviation of mean 24‐h‐weighted systolic blood pressure; 95% CI, 95% confidence interval; b, standardized regression coefficient; FMD, flow‐mediated dilation; IMT, intima‐media thickness; PWV, pulse wave velocity; R ², coefficient of determination; RHI, reactive hyperemia index.

TABLE 4.

Univariable and multivariable regression models of the relationship of PWV (as dependent variable) as a function of 24‐h‐weighted SBP variability, FMD, and RHI, alone or in combination

	Independent factor	Standardized beta (95% CI)	P‐value	Adjusted R 2 for the model
Model 1	FMD	−0.009 (−0.151; 0.134)	.772	0.005
Model 2	FMD	−0.008 (−0.150; 0.135)	.799	0.060
	24wSBP	0.224 (0.084; 0.355)	.019
Model 3	RHI	0.060 (−0.083; 0.201)	.422	0.024
Model 4	RHI	0.062 (−0.081; 0.203)	.404	0.065
	24wSBP	0.175 (0.033; 0.310)	.015
Model 5	RHI	0.061 (−0.082; 0.202)	.831	0.035
	FMD	−0.016 (−0.158; 0.127)	.412
	24wSBP	0.175 (0.033; 0.310)	.017

All models are adjusted for 24‐h‐weighted SBP. Statistically significant values in bold.

Abbreviations: 24wSBP, 24‐h blood pressure variability as standard deviation of mean 24‐h‐weighted systolic blood pressure; 95% CI, 95% confidence interval; b, standardized regression coefficient; FMD, flow‐mediated dilation; IMT, intima‐media thickness; PWV, pulse wave velocity; R ², coefficient of determination; RHI, reactive hyperemia index.

Conversely, univariable regression models showed that both parameters of endothelial function (FMD and RHI) did not appear to be significantly related to PWV (Table 4, models 1 and 3). We therefore next entered all such parameters in multivariable regression models aimed at improving the prediction of PWV (dependent variable), further adding the 24‐hour‐weighted SBPV (Table 4, models 2, 4, and 5). The added contribution of the 24‐hour‐weighted SBPV in predicting PWV was significant in models that included parameters of endothelial function (P = .019, P = .015, and P = .017, respectively).

Among these models, all the linear models predicting IMT had a better linear fitting than models predicting PWV.

To explore the impact on the results of our study of using only the awake time (daytime) SBPV as a simpler index of the short‐term SBPV as compared to weighted 24‐hour SBP‐SD, we conducted a sensitivity analysis by focusing only on the impact on HMOD of awake SBP SD. FMD and RHI parameters alone or in combination maintained a significant association with IMT, while BP variability measured with indices of awake SBP (awake SBP SD) was associated with IMT to a lesser extent compared with the 24‐hour‐weighted SBP SD (Table S1 compared with Table 3).

3.4. Mediation analysis to explain the effect of the various blood pressure variables on hypertension‐mediated vascular damage

The full mediation model of all variables (SPBV, RHI, and FMD) for IMT was significant (P < .001), accounting for 4.2% (0.0003/0.0072) of the mediating effect. The direct effect of SBPV on RHI and FMD (b = 0.001 and −0.016, respectively) was not significant, while the effects of RHI and FMD on IMT (b = 0.039 and −0.003, respectively) were significant. Moreover, the direct effect of SBPV on IMT (b = 0.007, P = .027) was also significant. The BCa results proved the significant effect of SBPV on IMT through endothelial dysfunction parameters (B = 0.0003, bias‐corrected and accelerated 5000 BCa 95% CI: 0.0001‐0.018). This part of the analysis indicates that SBPV influences IMT to some—limited—extent and that parameters of endothelial dysfunction (FMD and IMT) appear to mediate little of SBPV effects on IMT.

The full mediation model of all variables (SPBV, RHI, and FMD) for PWV was also significant (P < .001), accounting for 0.2% (0.0006/0.2441) of the mediating effect. Both direct effect of SBPV on RHI and FMD (b = −0.030 and −0.006, respectively) and of the RHI and FMD on PWV (b = 0.344 and −0.006, respectively) were not significant. Conversely, the direct effect of SBPV on PWV (b = 0.243, P = .027) was statistically significant. Finally, the bootstrapping results proved the non‐significant effect of SBPV on PWV through endothelial dysfunction (B = 0.0006, bias‐corrected and accelerated 5000 BCa 95% CI = −0.0246‐0.0198].

These results suggest that parameters of endothelial dysfunction (RHI and FMD) partially mediate the relationship between SBPV and IMT, but do not appear to mediate the effects on PWV, which mostly appear as direct (Figure 1, also reporting numerical outcomes of the mediation analysis).

FIGURE 1.

Mediation analysis to determine the relationship between parameters of 24‐h‐weighted systolic blood pressure variability (SBPV) and atherosclerotic target‐organ damage, as marked by the carotid artery intima‐media thickness (IMT) (Panel A) and the pulse wave velocity (PWV) (Panel B). SBPV appears to directly affect arterial stiffness (as indicated by PWV). SBPV does not appear, however, to affect parameters of endothelial function (as measured by the brachial artery flow‐mediated dilation (FMD) or the reactive hyperemia index (RHI)), which are, however, strongly related to IMT. SBPV might also affect carotid atherosclerosis (IMT) by other, still unexplored mechanisms. Data are reported as unstandardized regression coefficients (b) (in general, the higher this coefficient, the stronger the relationship) and relative 95% confidence interval (CI) in parentheses

4. DISCUSSION

In our study, short‐term SBP variability correlates significantly with two indices of vascular HMOD, namely vascular stiffness (PWV) and IMT, the latter taken as an index of subclinical atherosclerosis. Moreover, our study also shows that two well‐known and validated indices of endothelial function, FMD and RHI, predicted IMT (P < .001 for both), but not PWV. When one or both parameters of endothelial function considered in our study—namely FMD or RHI—were added to average BP and short‐term SBPV parameters in a multivariable model, they significantly contributed to predict IMT (P < .005 for both), but not PWV. We can reasonably infer from this analysis that, in a cohort of recent‐onset, previously untreated, hypertensive patients, endothelial dysfunction contributes to IMT largely independent of the direct contribution of BP average levels or of SBP variability. Moreover, endothelial dysfunction seems to play only a minor role—if any—in relating BP variables with arterial stiffening. Finally, our data indicate that short‐term SBPV might contribute more to arterial stiffening than to IMT.

In a previous paper of ours, ⁷ we had provided evidence that short‐term BPV, in particular when assessed as awake SBP SD and as 24‐hour‐weighted SBP SD—the latter calculated to remove the influence of the awake/asleep BP difference on short‐term BPV ²² —is positively related to IMT in patients similar to those included in our current study, that is, in recently diagnosed, previously untreated hypertensive patients. This relation remained significant independent of differences in average SBP values. ⁷ In that study, we also showed ⁸ that BPV is positively related to inflammatory markers, in turn known to be markers and perhaps mediators of vascular—including carotid artery—disease. Although our current data on the relationship of BPV with IMT are cross‐sectional and only based on a correlation analysis, being thus by definition unable to prove a cause‐effect relationship, it is tempting to speculate that BPV might cause vascular damage, which appears as a likely possibility. Our current study sheds some light on possible mechanisms by which this might occur.

Intima‐media thickness is a marker of subclinical atherosclerosis, ⁴ but involves changes in two components of the vessel wall, possibly coexisting in the genesis of atherosclerosis and not distinguishable by the carotid ultrasound assessment. These components are an expansion of the intima, due to early atherosclerotic changes, and increased thickness of the smooth muscle cell layer of the media, with the additional contribution of increased fibrosis. These two components likely have separate pathogenetic mechanisms, the former related to the intimal accumulation of macrophage/foam cells as a consequence of endothelial dysfunctions ²⁰ , ³⁵ ; the latter related to the medial hypertrophy and fibrosis caused by increased arterial strain. ³⁶ In the present study, we used in vivo techniques to evaluate endothelial vasodilatory function and vascular stiffness separately. Indeed, the brachial artery FMD ²⁶ , ³⁷ and the fingertip RHI ²³ appear to reflect large artery and arteriolar endothelial function, respectively. On the other hand, vascular stiffness is here investigated with measurement of the carotid‐femoral PWV and is a means for assessing mostly medial arterial wall changes occurring in hypertension. ¹⁰ , ³⁸ Both techniques have been found to predict vascular events, ³⁹ , ⁴⁰ , ⁴¹ their results being interrelated. ⁴²

Short‐term BP variability, assessed as both awake SBP SD and as 24‐hour‐weighted SBP SD, was in our study highly significantly related to IMT, and even more so to PWV. Conversely, these SBPV indices did not appear to correlate with indices of endothelial dysfunction (Table 2). Awake diastolic BPV was also related to PWV, but not to IMT, reinforcing the concept of a higher impact of BPV on arterial stiffness, more than on intimal atheroma formation. The significant relationships of both awake SBPV and 24‐hour‐weighed SBPV with PWV, but not with IMT, were confirmed at multivariable regression analysis after taking into account other determinants of HMOD. This is consistent with a model in which BP variability directly affects arterial stiffness, while the latter, in turn and indirectly, may affect IMT.

It has been shown that short‐term BPV has an adverse prognostic impact, while, on the contrary, the degree of BP reduction at night (dipping) carries favorable clinical implications. ⁴³ Therefore, the estimate of the 24‐hour BPV with the global 24‐hour BP SD, which is mathematically determined both by short‐term BP fluctuations occurring during the awake and the asleep times separately considered, as well as by the degree of BP changes between day and night (dipping), might not faithfully reflect the impact of short‐term BPV only on cardiovascular outcomes. In order to overcome this problem, one proposed method is the calculation of the 24‐hour‐weighted BP SD. This approach to the short‐term BPV assessment is based on the calculation of the average of awake and asleep time BP SD, each of them “weighted” for the duration of the awake (usually daytime) and asleep (usually nighttime) subperiods, respectively, thus allowing to exclude the impact of day‐night BP changes on the calculation of such a 24‐hour‐weighted BP SD. ⁴³ In our sensitivity analysis of the relationships of this simplified measure of daytime SBP variability, with parameters of HMOD (Table S1), FMD and RHI parameters alone or in combination maintained a significant association with IMT. Conversely, awake SBP SD was associated with IMT to a lesser extent than 24‐hour‐weighted SBP SD (Table 3). These findings thus appear to further support the value of the 24‐hour‐weighted approach to obtain a more accurate and comprehensive estimate of the short‐term BPV, including SBP fluctuations occurring not only during the awake but also during the asleep period.

Another finding of our study that deserves discussion is that alterations in the two indices of endothelial dysfunction here considered, the brachial artery FMD and the digital RHI, contributed to determining changes in IMT, but not in PWV, at multivariable analysis, taking all the various indices of HMOD into account (Table 4) and in our mediation analysis. Altogether, these findings are consistent with a model by which most of the effect of SBPV on vascular stiffness are independent of endothelial dysfunction, while endothelial dysfunction appears to be a contributor to the effects of BPV on IMT. This model also postulates a minor role of SBPV in determining endothelial dysfunction, while not excluding other possible contributions of BPV to IMT changes through other, still unknown, mechanisms (Figure 1).

4.1. Study limitations

We acknowledge the limitations of causality inference from correlation analyses. We cannot indeed rule out the hypothesis that increased arterial stiffness affecting the carotid arteries may result in decreased baroreceptor responsiveness, thus contributing to increased BPV. ⁴⁴ , ⁴⁵ We also acknowledge the weakness of some relationships here found, supporting the hypothesis that there other factors, beyond those here considered, altogether better explain the HMOD. On the other hand, we find the suggestion that BPV might affect changes in the arterial media, largely determining vascular stiffness, more directly than IMT, and largely without changes in endothelial function, to be a logic and attractive hypothesis, a finding that contributes to the ongoing discussion on the clinical relevance of BPV estimates. According to the same hypothesis, both increased BPV and endothelial dysfunction might eventually determine carotid atherosclerosis. To support such a possibility, intervention trials aimed at exploring changes in outcome induced by selective manipulation of BPV independent of changes in average BP levels would be needed, as no solid data are currently available on this issue. We also acknowledge that our inferences are based on a special selection of entry and exclusion criteria for our study population, and thus might be less applicable to the entire population of hypertensive patients. Further studies on broader and less selected populations are therefore warranted.

4.2. Study strengths

Our paper has also, however, important strengths and merits. Ours is—to the best of our knowledge—the first study addressing the possible role of intermediate determinants of IMT in relation to the effects of changes in BPV on them, also with the support of mediation analysis. We focused on a population with favorable characteristics for the achievement of our study aims, including subjects with recent (<2 years) onset of hypertension and no long‐lasting previous treatment, therefore largely not confounded by the duration of the hypertension condition and by the effects of drugs. While other investigations have proven the existence of a relationship of BPV with IMT ⁷ , ⁴⁶ and arterial stiffness, ¹⁴ our present one is the first report providing clues to identify a logical sequence in the pathogenetic chain of events leading from BPV to vascular damage.

5. CONCLUSION

SBPV is associated with both carotid artery IMT and carotid‐femoral PWV, but this latter appears more to be an effect on the medial (smooth muscle cell) component of the arterial vessels than mediated through changes in endothelial function. Carotid atherosclerosis is, however, also in turn affected by changes in endothelial function, measurable by the brachial artery FMD and the digital RHI, probably as the result of the influence of other risk factors, which appear to be even stronger determinants of IMT. Our results therefore provide elements of knowledge for a deeper understanding of the pathways through which hypertension causes vascular disease, and reinforce the concept that BPV is significantly and independently associated with HMOD. Given the cross‐sectional nature of our observations, however, the question of whether an increased BPV might be the cause or the consequence of the appearance of HMOD remains to be properly answered in studies with a longitudinal design.

CONFLICT OF INTEREST

None by any of the authors.

AUTHOR CONTRIBUTIONS

Alfonso Tatasciore, MD, PhD: Conception of the work; Recruitment of patients; Collection of data with Ambulatory Blood Pressure Monitoring, Carotid Artery ultrasound and Pulse Wave Velocity; Production of the database; Initial drafting of the manuscript. Marta Di Nicola, PhD: Statistical analysis. Roberto Tommasi, MD: Collection of data with Ambulatory Blood Pressure Monitoring, Carotid Artery ultrasound and Pulse Wave Velocity; Inclusion of data in the database. Francesco Santarelli, MD: Collection of data with Ambulatory Blood Pressure Monitoring, Carotid Artery ultrasound and Pulse Wave Velocity; Inclusion of data in the database. Carlo Palombo, MD: Critical assessment of the data and contribution to the Discussion. Gianfranco Parati, MD: Critical assessment of the data and contribution to the Discussion. Raffaele DE Caterina, MD, PhD: Conception of the work; Analysis of the data; Response to Reviewers’ critiques; Writing of the final manuscript.

Supporting information

Supporting Information

Tatasciore A, Di Nicola M, Tommasi R, et al. From short‐term blood pressure variability to atherosclerosis: Relative roles of vascular stiffness and endothelial dysfunction. J Clin Hypertens. 2020;22:1218–1227. 10.1111/jch.13871