Introduction

The ambulatory arterial stiffness index (AASI) has been defined as one minus the regression slope of diastolic on systolic blood pressure, as measured at the brachial artery by 24-h ambulatory blood pressure (BP) monitoring.1, 2 AASI reflects the dynamic relation between diastolic and systolic BPs throughout the whole day. Conceptually consistent with a hypothesis put forward in 1914,3 the stiffer the arterial tree, the closer the regression slope and AASI are to zero and one, respectively. Measurement of AASI does not require any other equipment than a validated4 portable monitor to record 24-h blood pressure. To date, cross-sectional analyses5 and at least three prospective cohort studies2, 6, 7 have demonstrated an association of AASI either with signs of target organ damage in never-treated hypertensive patients,5 or with the incidence of cardiovascular mortality and morbidity.2, 6, 7

AASI has been criticized as being merely a surrogate measure8, 9, 10 that may reflect hemodynamic factors such as arterio–ventricular coupling10 rather than arterial stiffness. More direct measurements of arterial stiffness,8 in particular aortic pulse wave velocity (aPWV), have a large amount of epidemiological evidence supporting the predicate that they predict cardiovascular events.11 The aPWV is usually considered as the gold standard for the assessment of arterial stiffness8 and might be helpful in elucidating further the physiological meaning of AASI. For this purpose, we analyzed the REASON trial (Preterax in Regression of Arterial Stiffness in a Controlled Double-Blind Study).12, 13, 14 We compared the effects of antihypertensive treatment with either atenolol (AT) or the combination of perindopril plus indapamide (PER/IND) on AASI, aPWV and pulse pressure (PP).

Methods

Study population

The REASON trial was a multicenter, randomized, double-blind and parallel-group study conducted in 13 countries. It compared the hemodynamic effects of the low-dose combination PER/IND with those of AT. Eligible patients had uncomplicated essential hypertension. Their BP measured in the supine position in the absence of any cardiovascular, antidiabetic or lipid-lowering drugs had to range from 160 to 209 mm Hg systolic or from 95 to 109 mm Hg diastolic. The present article reports on an ancillary study involving 32 of the 52 REASON centers,12, 13, 14 which opted to perform ambulatory BP monitoring (ABPM). Of 471 REASON patients, 269 underwent ABPM. A valid ambulatory recording had to include at least 48 measurements over 24 h and the interval between two successive readings should not be longer than 1 h. We excluded 68 patients, because their ambulatory BP recording at baseline or follow-up was of insufficient quality, leaving 201 patients for the present analysis. All patients gave written informed consent. The study was conducted in accordance with the Declaration of Helsinki after ethical approval had been obtained from the national regulatory authorities of each of the countries.

Treatment

After a 4-week placebo washout period, the patients were randomly assigned to treatment for 1 year, with PER (2 mg per day) plus IND (0.625 mg per day), or AT (50 mg per day) to be taken in the morning. After 3, 6 or 9 months, treatment could be adjusted according to the office BP measurements. If systolic or diastolic BP or both were higher than 160 or 90 mm Hg, the dose of the study medication was increased to two tablets, both to be taken in the morning. During follow-up, patients abstained from taking other BP-lowering drugs.

Blood pressure measurement

Conventional BP was measured at the brachial artery after the patients had rested for 10 min in the supine position, using a standard mercury sphygmomanometer and a suitable cuff size. The devices used for ABPM passed validation in class A or B, according to the protocols of the British Hypertension Society15 or the Association for the Advancement of Medical Instrumentation.16 We programed the recorders to obtain readings at 15-min intervals throughout the whole day, with the first measurement taken between 0800 and 1000 h, just before the intake of the study medication. We standardized the ambulatory recordings; at baseline and after 1 year of follow-up;14 the same technician applied the same monitor to the same arm in each patient and at the same time (±1 h) in the morning.

Pulse pressure, the difference between systolic and diastolic BP, and mean arterial pressure, diastolic BP plus one-third of PP, were computed from the office and the 24-h ambulatory BPs. Blood pressure control on conventional measurement was a systolic BP of less than 140 mm Hg and a diastolic BP below 90 mm Hg. The BP control on ambulatory measurement was a 24-h systolic BP below 125 mm Hg and a 24-h diastolic BP of less than 75 mm Hg.17

Measurement of arterial properties

At baseline and after 1 year of follow-up, we measured aPWV in the morning, after conventional BP measurement, and approximately 24 h after the last drug intake in a controlled environment at a mean (s.d.) temperature of 22±2°C,12 with the Complior (Colson, Paris, France). This device allows online pulse wave recording and the automated calculation of aPWV, and has been validated against the manual method.17 The mean difference between the two methods was −0.20±0.45 m s−1 (manual method, 11.05±2.58 m s−1 vs. automatic device 10.85±2.44 m s−1). The interobserver repeatability coefficient was 0.947 for the manual aPWV measurements and 0.890 for the Complior; for the intraobserver repeatability coefficients, these estimates were 0.938 and 0.935, respectively.

From individual 24-h ambulatory BP recordings, we computed the regression slope of diastolic on systolic BP.1, 2 We defined AASI as one minus the regression slope. We did not force the slope through the origin (intercept=0), because when during diastole the blood flow falls to zero, this does not occur for BP.18 From the individual readings in each 24-h ambulatory BP recording, we also computed the arterio–ventricular coupling index (AVCI) as (T/τ)/(1+2T/3τ),10 where T is the heart period in seconds and τ is the decay time of aortic BP during diastole.10, 19 We computed τ as (60 × mean arterial pressure)/(heart rate × PP).

Other measurements

We measured body weight without shoes, with the subjects wearing light indoor clothing. Body mass index was weight in kilograms divided by height in meters squared. Using <25 kg m−2, 25–30 kg m−2 and ⩾30 kg m−2 as thresholds, subjects were classified into normal weight, overweight and obese, respectively. We applied the NCEP-ATPIII criteria20 to define the metabolic syndrome (MS). Venous blood samples, collected after overnight fasting, were analyzed for blood glucose and the serum concentrations of high-density lipoprotein (HDL) cholesterol and triglycerides by automated enzymatic methods.

Statistical analysis

For database management and statistical analysis, we used SAS software, version 9.1.3 (SAS Institute, Cary, NC, USA). We compared means and proportions by the large sample z-test and by the χ2 statistic, respectively. We compared treatment-induced changes in BP between the two groups, while adjusting for baseline values, sex, age and body mass index. We searched for possible covariables of the arterial outcome variables by a stepwise multiple regression analysis with the P-value for independent variables to enter and stay in the model set at 0.15. The baseline variables considered for entry into the models were sex, age, body weight and height, 24-h mean arterial BP and 24-h pulse rate, smoking (0, 1), intake of alcohol (0, 1), previous antihypertensive treatment (0, 1) and the presence of the MS (0, 1). Analysis of covariance was used to compare the treatment-induced changes between the two treatment groups, while adjusting for baseline and covariables. We used correlation coefficients, unadjusted and adjusted for covariables, to express concordance between the indexes of arterial stiffness.

Results

Baseline characteristics

The 201 patients were enrolled in nine countries (France, n=52; Australia, n=50; Spain, n=38; Ireland, n=22; Germany, n=16; Switzerland, n=7; Belgium, n=6; Austria, n=5; The Netherlands, n=5). The patients randomized to PER/IND (n=107), compared with those allocated to AT (n=94), included a slightly larger proportion of drinkers (60.8 vs. 58.5%; P=0.04), but had lower HDL-cholesterol levels (1.26 vs. 1.37 mmol l−1; P=0.01). For the rest, the two treatment groups were well matched in terms of anthropometrics, smoking habits and metabolic abnormalities (Table 1).

Table 1 Patient characteristics at entry
Full size table

At randomization (Table 2), the two treatment groups had similar mean levels of systolic and diastolic BPs, as measured by a conventional sphygmomanometry (P=0.54 for systolic and 0.88 for diastolic) or by a 24-h ambulatory monitoring (P=0.70 for systolic and 0.35 for diastolic). In 201 patients, pulse rate measured at the office or by 24-h ambulatory monitoring averaged (±s.d.) 72.5±9.8 and 73.8±9.8 beats per minute, respectively, without any difference between the groups (P=0.84 and 0.83 for office and ambulatory measurements, respectively). Similarly, at randomization, there were no between-group differences in the indexes of arterial function, including the conventional and 24-h PP (P=0.66 and 0.74, respectively), AASI (P=0.75), AVCI (P=0.53), τ (P=0.57) and aPWV (P=0.84; Table 3).

Table 2 Blood pressure at randomization and follow-up
Full size table
Table 3 Indexes of arterial function at randomization and follow-up
Full size table

Determinants and concordance of the arterial measurements

In stepwise regression analysis, age, 24-h mean arterial pressure and pulse rate explained 46% of the variation in the 24-h PP. The same covariables plus body height explained 42% of the variance in aPWV. Of the variation in AASI, 25% was explained by age. Age plus the 24-h pulse rate explained 42 and 39% of the variation in AVCI and τ, respectively (Table 4).

Table 4 Determinants of arterial function at baseline in 201 patients
Full size table

Both without adjustment and adjusted for sex, age, body height, 24-h mean arterial pressure and 24-h heart rate, there was significant concordance between PP, AASI and aPWV (Table 5). We did not include AVCI and τ in Table 5, because for purely computational reasons, these measurements showed spuriously high correlation coefficients with one another and with 24-h PP.10

Table 5 Concordance between indexes or arterial function at baseline
Full size table

Effects of treatment on pulse rate and blood pressure

The baseline-adjusted change in the 24-h pulse rate averaged (±s.d.) −11.0±11.3 beats per minute in the AT group and 0.4±7.1 per minute in the PER/IND group. The baseline-adjusted between-group difference (AT minus PER/IND) and the changes in the 24-h pulse rate amounted to 11.4 beats per minute (95% confidence interval, −13.3 to −9.5; P<0.001).

Systolic BP decreased significantly (P=0.005) more in the PER/IND group than in the patients randomized to AT, irrespective of whether systolic BP was measured at the office or by ABPM (Table 2). Adjustment of the changes in systolic BP for baseline, sex, age and body mass index did not alter these findings. In contrast, the treatment-induced changes in diastolic BP (P=0.42) and mean arterial pressure (P=0.30) were similar in both the treatment groups (Table 2). After 1 year of treatment, the control rates of hypertension on conventional BP measurement were 80.8% on AT and 77.6% on PER/IND; on 24-h ABPM, the control rates were 48.9 and 53.3%, respectively. Those control rates of hypertension showed no difference between the two treatment groups.

Effects of on treatment on arterial function

In line with the findings for systolic BP, the PP, as measured by conventional sphygmomanometry or ABPM, decreased significantly more on PER/IND than AT dosage (Table 3). As could be expected on the basis of the decrease in the 24-h pulse rate and the reciprocal lengthening of the heart period, AVCI and τ lengthened on AT, with a significant difference in the treatment-induced changes compared with PER/IND (Table 3). aPWV, but not AASI, decreased slightly but significantly on AT as well as PER/IND. However, for both aPWV and AASI, the between-group differences in the treatment-induced changes were far from statistically significant (Table 3). Adjustment of the indexes of arterial function for the baseline values and entry characteristics, including sex, age, body height, 24-h mean arterial pressure and 24-h pulse rate did not materially alter the aforementioned results (Table 3).

Sensitivity analyses

Sensitivity analyses, which additionally accounted for drinking alcohol and the serum level of HDL-cholesterol, confirmed the results reported in Tables 2 and 3 (data not shown). Our findings also remained consistent when we applied robust regression (least trimmed squares), symmetric regression in the computation of AASI, or excluded influential data points (DFBETA>2/√n).

Because arterial stiffness differs between patients, with and without the MS,21 we tested whether the findings in Tables 2 and 3 were influenced by this condition (Figure 1). With adjustments applied as in Table 2, the 24-h systolic BP decreased to the same extent on AT and PER/IND in patients with the MS (−14.5 vs. −11.5 mm Hg; P=0.39), whereas in patients without the MS, the decrease on AT was smaller than on PER/IND (−7.9 vs. −14.8 mm Hg; P=0.003). Conversely, in patients with the MS, the adjusted 24-h diastolic BP decreased more on AT than PER/IND (−13.2 vs. −5.8 mm Hg; P<0.001), whereas in patients without the MS these decreases were similar (−7.2 vs. −7.7 mm Hg; P=0.58). The P-values for the interaction between treatment and the MS were 0.006, <0.001 and <0.001 for the 24-h systolic, diastolic and mean arterial pressures, respectively (Figure 1). All other interaction terms of treatment with the MS were not statistically significant for BPs or the arterial indexes.

Figure 1
figure 1

Changes in the 24-ambulatory blood pressure in response to 1 year of treatment with atenolol (AT) or perindopril/indapamide (PER/IND). Results are given for systolic and diastolic blood pressures, pulse pressure and mean arterial pressure in patients without (MS−) and with (MS+) the metabolic syndrome, defined according to the NCEP-ATPIII criteria.20 P denotes the significance of the treatment differences, Pint the significance of the interaction between treatment and the MS, and n the number of subjects in each group.

Full size image

Discussion

Both β-blockers and angiotensin-converting enzyme inhibitors reduce BP, but through contrasting hemodynamic mechanisms. β-Blockers without intrinsic sympathomimetic activity, such as AT, decrease heart rate and cardiac output and increase total peripheral resistance, although the latter effect is blunted in the long-run.22 Angiotensin-converting enzyme inhibitors, such as perindopril, cause vasodilatation with inhibition of the reactive neurohumoral activation, increase muscular blood flow and have no negative inotropic effect on the myocardium.23 aPWV, which is determined in the most pulsatile segment of the arterial circulation,24 is the gold standard for measuring arterial stiffness.8 AASI is also determined at an arterial site with pulsatile flow, the brachial artery, but is an indirect measure of arterial function.8 AASI is also influenced by other hemodynamic factors, such as heart rate,1, 25 stroke volume, wave reflections and systolic augmentation, and might also reflect arterio–ventricular coupling.10 However, after 1 year of treatment in the current study, the change in both aPWV and AASI were similar in the two treatment groups. Thus, under treatment with drugs with quite different hemodynamic profiles, AASI behaved similarly as aPWV.

Atenolol lowered systolic BP and PP less than PER/IND. As previously highlighted by other researchers22 and ourselves,12, 13 this difference is largely the consequence of the AT-induced reduction of heart rate, which is responsible for a return of the wave reflections in the central arteries during systole. This timing precludes a lowering of the central systolic blood pressure. Our current results are in keeping with those of the Conduit Artery Function Evaluation ancillary study26 from the Anglo–Scandinavian Cardiac Outcomes Trial.27 This sub-study included 2199 patients randomized to amlodipine with/without PER or to AT with/without bendroflumethiazide.26 Most patients received combination therapy throughout the study. Despite similar brachial systolic BPs between treatment groups (difference (Δ), 0.7 mm Hg; 95% confidence interval, −0.4 to 1.7; P=0.20), there were substantial reductions in the aortic systolic BP (Δ, 4.3 mm Hg; 95% confidence interval, 3.3–5.4; P<0.0001) and aortic PP (Δ, 3.0 mm Hg; 95% confidence interval, 2.1–3.9; P<0.0001) on the newer compared with the older drugs. Furthermore, under treatment with angiotensin-converting enzyme inhibitors, in particular PER, but not under treatment with AT, the structural arteriolar abnormalities associated with hypertension regress.28, 29 The ensuing reduction of the reflection coefficients likely reduces the amplitude of the backward pressure wave and promotes a decrease of systolic BP and PP in the brachial artery.

In patients without MS, the reductions in the 24-h systolic BP and PP were significantly larger on PER/IND than on AT. The opposite trend occurred for diastolic BP in patients with the MS. In untreated hypertensive patients, structural alterations of the arterioles go hand in hand with capillary rarefaction. The MS and a positive sodium balance enhance capillary rarefaction.30, 31, 32 Angiotensin-converting enzyme inhibitors reverse the structural damage at the arteriolar level. Salt depletion by diet or the administration of diuretics restore capillary density.33 We hypothesize that these beneficial effects are more difficult to achieve in patients with the MS than in those without this condition. On the other hand, compared with PER/IND, AT produced a larger decrease in diastolic BP and mean arterial pressure in the patients with the MS than in those without this condition. This might be because of the longer diastole in the presence of vasodilatation, which generally occurs in the presence of obesity,34 one of the hallmarks of the MS.20

AASI reflects the dynamic relation between diastolic and systolic BP throughout the whole day,1, 2 whereas aPWV and the 24-h PP do not account for the diurnal variability in the relation between diastolic and systolic BP. AASI depends on the combined effects of left ventricular ejection, the passive and active components of arterial stiffness and the reflection of the arterial pulse wave. In line with the present findings, we previously demonstrated that in healthy volunteers the unadjusted correlation coefficient between AASI and aPWV was 0.51.1 Furthermore, in randomly recruited Chinese, both before and after adjustment for arterial wave reflections by considering height and heart rate as covariables, AASI correlated more closely with the central and peripheral systolic augmentation indexes than with the 24-h PP.1 In the current study, although accounting for covariables, AASI correlated significantly both with aPWV and the 24-h PP.

Our current study must be interpreted within the context of its potential limitations. First, the original sample size calculations for REASON considered aPWV as the main outcome measure for the comparison between AT and PER/IND. To detect a significant (P=0.05) two-tailed difference of 0.5±1.2 (s.d.) m s−1 with 95% power, the number of subject to be analyzed was estimated to be 300. The REASON trial overall included 471 subjects, but our current report included only 201 patients with an ambulatory BP recording of sufficient quality. To detect a between-group difference of 0.5±1.2 m s−1 with a two-sided P-value of 0.05, our current study had 83% power. In addition, with a sample size of 201, our current study had 95% power to detect a significant (P=0.05) between-group difference in AASI of 0.066±0.13 (s.d.) or more. We cannot exclude that our current analyses were underpowered. Nevertheless, our current findings were consistent with the previously published REASON results.12, 13, 14 Second, data from Framingham Study35 and the Atherosclerosis Risk in Communities study36 showed that lower HDL-cholesterol was a forerunner of vascular disease. Although drinking alcohol increases HDL-cholesterol,37 in our current study HDL-cholesterol at baseline was lower in PER/IND than in AT group. In a sensitivity analysis, we adjusted for HDL-cholesterol and for the small difference in the proportion of drinkers. However, we cannot exclude with certainty that residual confounding biased our results. Notwithstanding these potential limitations, compared with conventional BP measurement, ambulatory monitoring more precisely reflects a subject's usual blood pressure, excludes observer bias and minimizes the white-coat effect.4 These are desired characteristics that might contribute to the validity of AASI as an index reflecting the arterial function.

Our study is the first one to compare the influence of a pharmacological intervention on aPWV, the standard measurement of arterial stiffness8 and AASI. On administration of antihypertensive drugs with quite different hemodynamic profiles, AASI and aPWV behaved similarly. Our current findings, taken together with the prognostic significance of AASI,2, 6, 7 and its concordance with other indexes of arterial function,1 strengthens the position of AASI as a diagnostic instrument reflecting arterial stiffness. A worldwide consortium of investigators are currently enlarging the database of ambulatory BP recordings in relation to cardiovascular outcomes and plan to study the predictive value of AASI in population-based cohorts of different ethnicity.38 In the meantime, manufacturers of devices for ABPM are including the computation of AASI in their software packages. Clinicians might consider AASI in the risk stratification of their patients under the proviso that further clinical and epidemiological validation of this novel prognostic index of arterial function is warranted.