Author + information
- Received November 19, 2008
- Revision received February 11, 2009
- Accepted February 23, 2009
- Published online August 18, 2009.
- Bryan Williams, MD⁎ (, )
- Peter S. Lacy, PhD,
- CAFE and the ASCOT (Anglo-Scandinavian Cardiac Outcomes Trial) Investigators
- ↵⁎Reprint requests and correspondence:
Dr. Bryan Williams, Department of Cardiovascular Sciences, University of Leicester, Clinical Sciences Building, Leicester Royal Infirmary, P.O. Box 65, Leicester, LE2 7LX, United Kingdom
Objectives The CAFE (Conduit Artery Function Evaluation) study showed less effective central aortic pressure lowering with atenolol-based therapy versus amlodipine-based therapy in people with hypertension. The present study examined the importance of heart rate (HR) as a determinant of this effect.
Background Recent analyses have suggested that beta-blockers are less effective at reducing cardiovascular events than alternative blood pressure (BP)-lowering therapies. There has been much debate about the mechanism for this shortfall in benefit and specifically the role of HR lowering by beta-blockers.
Methods Central pressures were derived from brachial pressure and radial pulse wave analysis in 2,073 patients, and 7,146 measurements were recorded and analyzed over follow-up for up to 4 years.
Results There was no impact of HR on brachial systolic or pulse pressures; however, there was a highly significant inverse relationship between HR and central aortic systolic and pulse pressures (p < 0.001). This was dependent on a strong inverse relationship between HR and augmentation index, indicative of increased wave reflection at lower HRs. Multiple regression, adjusted for brachial BP, showed HR to be the major determinant of central pressures. Moreover, HR and brachial BP accounted for 92% of the variability in central systolic and pulse pressures. Consequently, drug-related differences in central aortic pressures were markedly attenuated after adjustment for HR.
Conclusions When comparing beta-blocker–based treatments with other BP-lowering strategies, HR reduction with beta-blockers is a major mechanism accounting for less effective central aortic pressure reduction per unit change in brachial pressure.
Beta-blockers have been a primary treatment for hypertension for many years. However, recent analyses have suggested that beta-blocker–based therapy may be less effective at preventing cardiovascular events when compared with alternative blood pressure (BP)-lowering treatments in people with hypertension (1–5). The United Kingdom National Treatment Guidelines in 2006 recommended that beta-blockers should no longer be considered a suitable initial therapy for the treatment of hypertension (6). There has been much speculation about mechanisms for this shortfall in cardiovascular protection, especially stroke prevention, with beta-blockers in hypertensive patients. In the CAFE (Conduit Artery Function Evaluation) study, we have previously shown that the beta-blocker atenolol was less effective at lowering central aortic systolic and pulse pressures (PPs) when compared with alternative BP-lowering treatment, despite similar brachial BP control (7). These findings are consistent with data from previous smaller-scale studies of shorter duration (8–11). Further analysis of the CAFE study suggested that central pressures may be an independent predictor of clinical outcomes in hypertensive patients (7). These findings suggest that the shortfall in benefit from beta-blockers could relate to less effective central aortic pressure lowering, despite seemingly similar effects as other drugs treatments on brachial BP. If this is the case, then important questions follow. What is the mechanism for the less effective reduction in central aortic pressures with beta-blockers? Is this mechanism specific to atenolol, or is it more broadly applicable to all beta-blockers?
In the CAFE study, the main difference in central aortic pressures resulted from an increase in pressure wave reflections (augmentation index [AIx]) with atenolol-based therapy, resulting in augmentation of central aortic systolic and PPs. Previous studies have demonstrated that AIx is inversely related to heart rate (HR) (12,13), suggesting that HR reduction may be the main mechanism accounting for less effective central pressure reduction with beta-blocker–based therapies.
These observations prompt further questions. How much of the difference between atenolol- versus amlodipine-based therapy in the CAFE study could be attributed to the differences in HR between treatments? After adjusting for HR differences, was there any residual impact of the 2 BP-lowering regimens on central aortic pressures and hemodynamics?
The answer to these questions clearly has important implications with regard to the potential impact of therapeutic HR manipulation on central aortic pressures and hemodynamics in people with hypertension. The present study thus examined the hypothesis that HR was a major factor accounting for the differential impact of BP-lowering treatments on central aortic pressures and hemodynamics in the CAFE study.
The details of the CAFE study patient population and study design and procedures have been previously published (6) and are briefly summarized below.
CAFE study population and design
The CAFE study was a substudy of the ASCOT (Anglo-Scandinavian Cardiac Outcomes Trial) study (14). Data on central aortic hemodynamics was available from 2,073 participants recruited from 5 ASCOT study centers in the United Kingdom and Ireland. These data form the basis of the present analysis and were collected over a median follow-up of 3 years. At baseline, the patient population was hypertensive, of whom the majority was previously treated (90%). The patients also had 3 additional cardiovascular risk factors to qualify for randomization to 1 of 2 BP-lowering strategies, using a prospective, randomized, open, blinded end point design: 1) a regimen of amlodipine, adding perindopril as required; or 2) a regimen of atenolol, adding bendroflumethiazide-K as required. Additional BP-lowering therapies were common to both treatment arms according to a pre-specified algorithm (14). Antihypertensive treatment was titrated to achieve a target BP (<140/90 mm Hg for people without diabetes and <130/80 mm Hg for people with diabetes). The patient demographics are shown in Table 1.All patients gave written informed consent, and approval for the study was granted by local research ethics committees at each ASCOT study center. Ethical approval was also granted by the United Kingdom Multi-Center Ethics Committee.
Brachial BP, radial pulse wave analysis, and derivation of central aortic pressures and hemodynamic indexes
Brachial BP was measured using a validated semi-automated oscillometric device (Omron 705CP, Omron, Kyoto, Japan) as specified in the ASCOT study protocol (15). The CAFE study used radial artery applanation tonometry and pulse wave analysis (16,17) to derive central BPs and other parameters, as previously described (Online Appendix).
This method generates central aortic pressure waveforms from the radial pressure waveform using a previously validated transfer function (18,19). The central pressure waves were analyzed to identify the outgoing and reflected components and to calculate the AIx (i.e., the proportion of the central PP that is attributable to pulse wave reflection [ΔP], i.e., [AIx = (ΔP/PP) × 100]) (Online Fig. 1). PP amplification was calculated as the ratio of brachial to central PP. An average of 3.4 applanation tonometry measurements per patient were obtained at scheduled ASCOT study follow-up visits. Typical interobserver variability at individual ASCOT centers was 0.3 ± 2.9 mm Hg for central systolic pressure and 1.5 ± 5.9% for AIx. This is consistent with our previously published data using this technique (20).
Statistical analyses were performed in collaboration with the ASCOT Study Coordinating Center at A+ Science, Goteborg, Sweden, using the SAS computer program version 8.2 (SAS Institute Inc., Cary, North Carolina).
Analysis of the impact of HR on brachial BP, central aortic pressures, and hemodynamic indexes
This analysis used 3 complementary strategies: 1) We examined the relationship between HR as a continuous variable and brachial BP, central aortic pressures, and hemodynamic indexes. Data from every CAFE study measurement (n = 7,146) relating HR to these indexes, blinded to treatment allocation, were included in the analysis. 2) Multiple stepwise regression was performed to rank and quantify the impact of HR on brachial and central aortic pressures and hemodynamic indexes. 3) Data from the CAFE study was adjusted for HR to assess the residual impact of drug therapy on central aortic pressures and hemodynamics.
For univariate analyses, data were grouped into deciles of HR, and the relationship between HR and hemodynamic variables was analyzed using linear regression. Regression lines were also fitted to plots of raw data. For multivariate analysis, stepwise multiple linear regression was used. Variables entered into the model were determined by linear correlation analyses. Continuous data variables between BP-lowering regimens were compared using nonpaired Student ttests. Where stated, data were adjusted for HR using general linear modeling before comparisons were made. Data are presented as mean (95% confidence interval) or mean ± SD as stated and a value of p < 0.05 was considered significant.
The baseline characteristics of the CAFE study population according to their randomized BP-lowering treatment allocation are shown in Table 1. The 2 treatment groups were well matched with respect to their demographics, clinical characteristics, and previous medication.
The relationship between HR and brachial versus central pressures
The relationship of HR with brachial and central pressures is shown in Figure 1.The data encompass all measurements performed during the CAFE study follow-up. The data plots were very dense, thus for clarity of presentation, the data were grouped into increments of increasing HR (10 beats/min increments). Importantly, the regressions of the relationships did not differ when comparing the grouped and raw data plots. There was no significant impact of reducing HR on brachial systolic BP (+0.6 mm Hg per 10 beats/min decrease in HR) (Fig. 1A). By contrast, there was a 5-fold greater increase in central systolic pressure per unit change in HR (+3.0 mm Hg per 10 beats/min decrease in HR). A similar dissociation between the impact of HR on brachial and central PP was also observed (Fig. 1B).
Figure 2shows the differences between brachial and central pressures, plotted as a function of HR. Importantly, at lower HRs, the difference between brachial and central pressure progressively decreased, so that central pressure approached brachial pressure at the lowest HRs.
Relationship between HR and components of the central pressure waveform
To investigate the mechanisms involved in the changes in central pressure with variation in HR, we next analyzed the components of the central pressure waveform in relation to HR. There was minimal impact of HR on the amplitude of the outgoing pressure wave (P1 height). However, there was a strong and significant inverse relationship between HR and the amplitude of pressure wave reflections (augmentation), which increased by ∼3 mm Hg per 10 beats/min reduction in HR (Fig. 3A).This finding suggests that the main impact of HR reduction was on the reflected wave, rather than the incident pressure wave. Consistent with this observation, there was a marked increase in AIx with reducing HR: +4.9% per 10 beats/min reduction in HR (Fig. 3B).
The relative contribution of HR to central pressures and hemodynamic variables
To further evaluate the contribution of HR to central pressures and hemodynamics, we performed stepwise multiple linear regression (Table 2).After accounting for brachial BP, HR was the major determinant of central systolic and pulse pressures, accounting for 5% and 9% of the variability in these parameters, respectively. HR was also a major determinant of pressure wave reflections (augmentation and AIx) and PP amplification, accounting for 26%, 34%, and 54%, respectively of the variability in these parameters. Of importance, in this analysis, the BP treatment regimen was a much less powerful determinant of central pressures and wave reflections, accounting for no more than 0.5% of the variability (i.e., at least 10-fold less important than the impact of HR).
Comparison of central pressures and hemodynamic variables between BP-lowering treatment arms before and after adjustment for HR
To further evaluate the relative contribution of HR as a determinant of central pressures, the differential impact of the 2 BP-lowering treatment regimens on central pressures and wave reflections was resolved after adjusting the data for HR differences (Table 3).After HR adjustment, the differences in central systolic and PPs between treatment arms were no longer significant, and the differences in augmentation, AIx, PP amplification, and the brachial-central aortic systolic and pulse pressure changes were markedly attenuated. Taken together, these data suggest that HR is an important determinant of central aortic systolic and PPs and was the main determinant of the difference between central and brachial pressures between treatment arms in the CAFE study.
Within a major clinical outcomes trial, this is the first study to define the impact of drug-related changes in HR, on central aortic pressures and hemodynamics, in hypertensive patients. With over 2,000 patients and over 7,000 measurements, this study had abundant statistical power to test its hypotheses. The data clearly demonstrate the powerful influence of HR, across the physiological range, on central aortic pressures and wave reflections in hypertensive patients, despite minimal effects on brachial pressures.
We show that HR is inversely related to central aortic systolic and PPs. Lower HRs are also associated with reduced PP amplification; thus, at lower HRs, the central aortic systolic pressure becomes closer to the brachial systolic pressure. Importantly, there was minimal impact of HR on the outgoing pressure wave height (P1 height), showing only a minor increase with reduced HR. However, the inverse relationship between HR and indexes of central pressure wave reflection (i.e., augmentation) were much stronger, consistent with increased wave reflection at lower HRs. Remarkably, the slopes for the relationship between HR and central aortic systolic pressure or magnitude of wave reflection (augmentation) were identical (+3 mm Hg per 10 beats/min reduction in HR), suggesting the importance of wave reflection in mediating the HR-related change in central aortic systolic pressure.
This finding that central pressure wave reflection is strongly influenced by HR is supported by data from cross-sectional studies with data stratified by HR (21) and studies of cardiac pacing in humans, which suggested that AIx declines by 4% to 5% per 10 beats/min increase in HR (12,13); the data for AIx from the present study are similar at 4.9% per 10 beats/min reduction in HR. Interestingly, in a recent population study, HR was the most powerful modifiable predictor of AIx, central systolic pressure, and central PP (22).
Impact of HR versus treatment regimen
Multiple regression analysis confirmed the relative importance of HR after BP itself, as a key determinant for all central hemodynamic parameters. Moreover, adjusting the CAFE study data for HR markedly attenuated the difference in central pressures between the 2 BP-lowering treatment regimens. This suggests that this difference was primarily driven by differences in HR, although some residual effects remained. It is conceivable that unmeasured hemodynamic factors such as aortic stiffness and systemic vascular resistance/remodeling, which could be differentially influenced by drug treatments, may have accounted for some of this residual variability (10,23).
Mechanisms for the inverse relationship between HR and central aortic pressures
We suggest 2 mechanisms that could account for the elevation in central pressures with reduced HR: first, reducing HR prolongs cardiac ejection duration, but has no major effect on pulse velocity (7,24). This increases the likelihood of a greater proportion of the reflected wave appearing in late systole for any given pulse wave velocity. Beta-blockade also decreases the dP/dT during ventricular ejection, and this could delay the time to the peak of the outgoing wave (10,25). This could also increase central systolic pressure by increasing the probability of coincidence of the reflected wave with late systole. Our finding of an increased AIx with beta-blockade is consistent with this hypothesis.
Second, the less effective lowering of central aortic systolic and pulse pressures in patients with lower HRs is consistent with basic physiology. According to the derivation of Poiseuille's law, BP is the product of cardiac output × peripheral resistance, where cardiac output is the product of stroke volume and HR. When HR is reduced by drug therapy (e.g., a beta-blocker) mean arterial pressure is maintained by an increase in stroke volume (26)—a phenomenon readily observed in patients with complete atrio-ventricular heart block. In younger patients with compliant conduit arteries, this increase in stroke volume can be accommodated. Indeed, in conditioned athletes, a combination of increased aortic compliance and peripheral vasodilation prevents a marked rise in AIx and central aortic pressure despite very low HRs and markedly increased stroke volumes (27). This represents perfect physiological adaptation to a reduced HR. In contrast, most hypertensive patients are not conditioned athletes, and in the CAFE study were older with stiffened conduit arteries. In this setting, a reduction in HR will result in the increased stroke volume being ejected into a less compliant proximal aorta, resulting in a rise in central aortic systolic and PPs. We suggest that these are the 2 principal mechanisms accounting for the inverse relationship between HR and central aortic systolic and PPs in the CAFE study. Moreover, we suggest that this inverse relationship would be accentuated if HR changes are restricted by drugs (i.e., beta-blockers) during exercise, when the need to increase cardiac output could only be met by an increase in stroke volume. These considerations are of clinical importance given that central PP showed a significant association with clinical outcomes in the CAFE study and other studies (28,29).
Is this data relevant to all beta-blockers?
The present study raises important questions as to whether similar effects would have been observed with beta-blockers other than atenolol, notably vasodilating beta-blockers. Our data suggest that the impact of HR on central aortic pressure is very powerful and consistent across the physiological range, irrespective of treatment allocation in this study. Other studies using invasive monitoring have shown that in patients receiving beta-blockers, the use of powerful vasodilators cannot overcome the impact of HR reduction on wave reflection and central pressures (30). By contrast, a small number of previous studies comparing vasodilating and nonvasodilating beta-blockers have suggested a more beneficial influence of vasodilating beta-blockers on central pressures (31,32). However, these studies were small scale and underpowered, and the differences in central pressures and wave reflections between the different beta-blockers could be accounted for by the lesser reductions in HR with vasodilating beta-blockers and/or differences in brachial BP.
We recruited predominantly white men. However, our regression analysis suggests that the direction of change in central pressures and hemodynamics was the same for women. It is unclear whether similar findings would have been observed in other ethnic groups. Nevertheless, we cannot rationalize why a mechanism that appears to be so dependent on HR would be different in other ethnic groups. Our patient population was also older, with a mean age of 63 years at baseline. It is conceivable that in younger people with more compliant conduit arteries there would be a lesser impact of lower HRs on central aortic pressure. These important considerations need further evaluation.
We used noninvasive methods to derive central aortic pressure from the radial pulse wave, calibrated to brachial BP. It has to be considered whether the mathematical transfer function used to derive central hemodynamic indexes could be sensitive to, or confounded by, changes in HR. The mathematics involved are beyond detailed discussion here but use a transfer function to calculate central pressures from individual radial pressure waveforms that is uninfluenced by the number of waveforms as a function of time. Although, to our knowledge, there have been no specific studies to assess impact of HR on central pressures comparing the methods here with direct invasive measurements, there have been invasive measurements of central aortic pressures in humans in response to changes in HR. In these studies, increasing HR via cardiac pacing has been shown to reduce central aortic pressure (12). Moreover, previous invasive studies have shown that beta-blocker treatment increased (rather than reduced) central aortic pressures (33). These directional changes are consistent with our findings. Furthermore, data from studies directly analyzing carotid or invasively acquired central pressure waves have documented reduced “pressure amplification” with beta-blockade (24,33), consistent with our data. Other studies have also implicated HR as a major factor modulating pressure amplification (13,21,34,35).
Finally, our study examines the association between on-treatment HR and central pressures. It does not directly assess the change in central pressure in response to a treatment-induced change in HR in individual patients. This would have been difficult to do because of confounding due to associated changes in BP per se as a consequence of any treatment changes. Nevertheless, our multiple regression analysis identified HR to be a powerful independent factor influencing the relationship between brachial and central aortic pressures, with the latter being higher at lower HRs.
These data have important clinical implications. There is a well-recognized association between a lower HR and cardiovascular health reported from observational studies (36–38). This is often used as a justification for therapeutic reductions in HR. In the setting of symptomatic ischemic heart disease and in patients with chronic stable heart failure, HR lowering by beta-blockade has been shown to be a very effective treatment strategy. However, the data from the present study question whether extending these assumptions to people with hypertension, especially older people with stiff conduit arteries, is safe and appropriate. Moreover, because the effect of HR on central pressures seems so powerful, our data suggest that there will be less effective central aortic systolic and PP reduction in older hypertensive patients with all beta-blockers, or other drugs that lower HR. In this regard, the newer generation of vasodilating beta-blockers must be shown to be as effective as alternative treatments in preventing cardiovascular events before they can be considered as a suitable routine treatment for older people with hypertension.
In summary, the CAFE-Heart Rate study has demonstrated that a lower HR is associated with higher central aortic systolic and PPs in patients with treated hypertension. We suggest that this is the major reason why beta-blocker–based therapy has been less effective at reducing cardiovascular events, especially stroke, when compared with other treatments in patients with hypertension.
For the text on the measurement of pressure waveforms and definition of variables derived by pulse wave analysis, a list of the CAFE investigators, and a figure on the central arterial pressure wave with derived parameters, please see the online version of this article.
Continuing Medical Education (CME) is available for this article.
- Abbreviations and Acronyms
- augmentation index
- blood pressure
- heart rate
- pulse pressure
- Received November 19, 2008.
- Revision received February 11, 2009.
- Accepted February 23, 2009.
- American College of Cardiology Foundation
- Wiysonge C.,
- Bradley H.,
- Myose B.,
- et al.
- Bangalore S.,
- Wild D.,
- Parkar S.,
- Kukin M.,
- Messerli F.H.
- Bangalore S.,
- Sawhney S.,
- Messerli F.H.
- The CAFE Investigators for the ASCOT Investigators
- London G.M.,
- Asmar R.G.,
- O'Rourke M.,
- Safar M.E.,
- REASON Investigators
- Wilkinson I.B.,
- Mohammed N.H.,
- Tyrrell S.,
- et al.
- Dahlof B.,
- Sever P.S.,
- Poulter N.R.,
- et al.,
- ASCOT Investigators
- Kelly R.,
- Hayward C.S.,
- Ganis J.
- Chen C.-H.,
- Nevo E.,
- Fetics B.,
- et al.
- Laurent P.,
- Albaladejo P.,
- Blacher J.,
- Rudnichi A.,
- Smulyan H.,
- Safar M.E.
- McEniery C.M.,
- Hall I.R.,
- et al.,
- ACCT Investigators
- Asmar R.G.,
- London G.M.,
- O'Rourke M.,
- Safar M.E.,
- REASON Project Coordinators and Investigators
- Nichols W.W.,
- Edwards D.G.
- Edwards D.G.,
- Lang J.T.
- Ting C.T.,
- Chou C.Y.,
- Chang M.S.,
- Wang S.P.,
- Chiang B.N.,
- Yin F.C.P.
- Protogerou A.D.,
- Blacher J.,
- Mavrikakis M.,
- Lekakis J.,
- Safar M.E.
- Dyer A.R.,
- Persky V.,
- Stamler J.,
- et al.