Introduction

Arterial afterload is influenced by the physical properties of the aortic root including its size. Both aging and disease states such as hypertension have been shown to be related to increasing aortic root size in cross-sectional (Vasan et al. 1995), longitudinal (Lam et al. 2010) and post-mortem studies (Virmani et al. 1991). This is associated at molecular level by inflammatory activity (Lindeman et al. 2009) and degradation of elastin fibers in the vessel wall with replacement by collagen (Wang and Lakatta 2002). As this causes central arteries to lose elasticity, the concomitantly observed aortic dilation has been attributed to the exhaustion of strain reserve: the repetitive pulsatile hemodynamic stress is no longer accommodated by distensibility (AlGhatrif and Lakatta 2014; O’Rourke and Nichols 2005; Safar et al. 2003). Hemodynamic stress may also contribute to aortic root dilatation. This has been shown in athletes and is reportedly more common not only in high-intensity sports (Pelliccia et al. 2010) but also in children with complete heart block where aortic dilation appears to regress after pacemaker insertion, in parallel with a reduction in stroke volume achieved by the increase in heart rate (Altit et al. 2014). Whether this process is driven purely by altered hemodynamic load, or if aortic wall inflammation plays a role, is unknown (Davey et al. 2011; Hans et al. 2012).

Windkessel models are applied to conceptualize afterload as divided into its constituent components in the frequency domain (Frank 1899). A widely used model is the 3-element windkessel which constitutes an electrical analog to the circulation and incorporates (a) a capacitor with electric charge storage properties to account for the distensibility of elastic arteries, (b) a peripheral resistive element which is the resistance at steady state flow, i.e., the zero harmonic, localized to the distal circulation and (c) characteristic impedance (ZC), a proximal resistive force acting on the high-frequency impedance spectrum. Based on non-windkessel theories, an analytical argument can be made that a larger aortic root should contribute to a lower afterload: first, the Moens-Korteweg equation posits that the diameter of the aorta is positively related to the aortic elastance-wall thickness product (i.e., a measure of aortic stiffness) and inversely related to the square of the local pulse wave velocity. Second, the water hammer equation posits that pulse wave velocity is positively related to ZC. Taken together, these equations connect a larger aortic root size with lower ZC, which should translate into lower afterload and larger stroke volume (SV), especially when preload and/or contractile reserves are limited. However, this view runs contrary to the traditional paradigm that large artery dilation is a feature of adverse remodeling, and there are few reports from the real world to support this notion (Gardin et al. 2006; Lam et al. 2013; Mitchell et al. 2008).

Different classes of antihypertensive drugs exhibit differences in their effect on central blood pressure, an important finding considering the role central pressure appears to play for the clinical course of the patient (Mancia et al. 2014). While the exact physiology of central pressure remains controversial, its determinants are likely to include not only cardiac output and systemic vascular resistance (SVR) but also factors controlled by properties of the proximal aorta including its stiffness, the pulse wave velocity and the magnitude of reflected arterial waves (Izzo 2004; O’Rourke 1990). Whether the size of the aortic root plays a role for the effects of antihypertensive drugs on central hemodynamics is unknown.

We wished to test the hypothesis that a larger aortic root is associated with lower LV afterload and consequently a larger SV. We postulated that the association between aortic root size and LV afterload is influenced by antihypertensive drugs.

Methods

Patients and data sources

Individuals were identified for inclusion into the present retrospective analysis from a registry of patients undergoing coronary arteriography for acute coronary syndrome at National Heart Centre Singapore, using unique personal national identification numbers to link records to echocardiographic data available in the local clinical echocardiography database. Inclusion criteria were (a) registry enrollment between 3 August 2007 and 30 December 2010 and (b) echocardiography having been performed in close conjunction with the angiography (±1 week). Exclusion criteria were (a) having neither Singapore citizenship nor permanent resident status and (b) any valvular heart disease that invalidates subsequent hemodynamic analysis: specifically, regurgitation of either mitral or aortic valve greater than mild in severity or aortic stenosis of any degree. We included only the first study for patients that had undergone multiple angiograms or ultrasound scans. Individual case sheets and electronic health records were reviewed in cases with missing data including clinical characteristics and blood test results. The registry, which contains demographic and clinical information, is managed by Singapore Cardiac Databank and was audited in November 2015 against source documents and found to be 97.5 % accurate. As this was a retrospective study utilizing existing data, the local institutional ethics review board approved the protocol with waiver of informed consent.

Echocardiograms

Echocardiograms were performed using routine clinical image acquisition (GE Vivid System E5, E7 and Vivid i, Chalfont, United Kingdom; Philips IE33 and CX50, Eindhoven, The Netherlands; Siemens Acuson Sequoia 512 and Aspen, Malvern PA, USA; Toshiba Artida, Otawara, Japan; all transducers 1–5 MHz; lab staffed by 25 sonographers, 10 physicians; Agfa HealthCare IMPAX Cardiovascular Suite, Greenville SC, USA), adhering to applicable guidelines (Lang et al. 2005). The size of the aortic root was recorded at three locations: (a) as the widest diameter at the level of the sinus of Valsalva, (b) at the sinotubular junction and (c) and as the largest diameter of the tubular portion of the ascending aorta as measured from inner edge to inner edge in the parasternal long-axis acoustic window (Fig. 1; Lang et al. 2005). It was decided pre-hoc to use ascending aortic diameter for the purpose of all data analyses in this study and aortic dilation was defined as an indexed diameter ≥2.1 cm/m2 as recommended elsewhere (Evangelista et al. 2010). LV outflow tract diameter was measured 0.5–1.0 cm below the aortic valve in a focused and zoomed-in parasternal long-axis view in mid-systole as the distance between basal septum and anterior mitral valve leaflet, and SV was obtained by multiplying LV outflow tract area (Lang et al. 2005) by the pulsed-wave Dopper velocity–time integral. LV end-diastolic volume (LVEDV) was obtained by dividing SV by LV ejection fraction (LVEF) and LV end-systolic volume (LVESV) was calculated by subsequently subtracting SV from LVEDV, a method validated in a porcine model against 3D echocardiography and post-mortem LV volumes (Hart et al. 2002). Arterial elastance (EA), which was introduced as an estimate of global afterload (Sunagawa et al. 1985) and subsequently validated against impedance measurements in the time domain (Kelly et al. 1992), was obtained by dividing LV end-systolic pressure [LVESP; estimated as 0.9 brachial systolic blood pressure (SBP)] by SV. End-systolic LV elastance (ELV) was used as a pre-load independent measure of LV contractility and calculated by dividing LVESP by LVESV (Sagawa et al. 1977), under the assumptions that end-systolic volume is linearly related to pressure and that V0 is negligible, both of which are reasonable under physiological loading conditions (Burkhoff et al. 2005; Kass and Kelly 1992). Arterial compliance (CA) was calculated by dividing SV by brachial pulse pressure, assuming a linear association between arterial volume and pressure (Chemla et al. 1998). Mean arterial pressure (MAP) was derived from brachial blood pressure as diastolic blood pressure (DBP) +PP/3. SVR was calculated as MAP divided by cardiac output. All chamber size data were indexed to body surface area (BSA) prior to any analyses (Asanoi et al. 1989; Najjar et al. 2004) and are shown in their indexed form throughout the article. Heart rate and brachial artery blood pressure were recorded immediately before echocardiography. The coefficient of variation for measurements of aortic root size, LVEF and SV, analyzed in 50 randomly selected studies as the standard deviation of differences divided by the mean, was 4.3, 7.6 and 8.6 %, respectively.

Fig. 1
figure 1

Representative screenshot showing the aortic root and measurements made

Full size image

Statistics

Logarithmic transformation was applied to skewed variables as appropriate. Correlations were analyzed using Pearson’s or Spearman’s correlation coefficients; T tests or Mann–Whitney U tests were applied for un-paired comparisons, as appropriate. We used multiple regression to analyze central hemodynamics as done elsewhere (Sahlén et al. 2012), using multivariable models to analyze predictors of SV, including a priori the principal determinants of SV as covariates: heart rate, LVEDV (preload), EA (afterload) and ELV (contractility). In similar analysis of predictors of EA (i.e., lumped afterload), we included a priori the elements from a 2-element windkessel model of the arterial circulation as covariates (SVR, CA) as well as root size which acted as a proxy for proximal arterial function (i.e., the third element of a 3-element windkessel model). The impact of vasoactive drugs (antihypertensive or statins) (Koh 2000) was investigated by interaction analyses applied to pre-hoc defined subgroups [treatment with calcium channel blockers (CCB), β-adrenergic receptor blockers, diuretics, long-acting nitrates, angiotensin-converting enzyme (ACE) inhibitors and statins] using centered interaction variables while adjusting for arterial blood pressure. Assumptions for linear regression were checked as follows: linearity of relationships between covariates and the dependent variable were analyzed in component residual plots, residuals were inspected for normality, homoskedasticity and the independence of errors was assessed using the Durbin-Watson statistic, multicollinearity was excluded by calculating variance inflation factors, outliers were identified by calculating leverage values and double-checked against original data post hoc to ensure data veracity. Overall regression model predictive accuracy was assessed as adjusted R 2. The relative importance of regressors was estimated by quantifying the variance attributable to each one using an established method for partitioning the R 2 statistic by averaging sequential sums of squares over all orderings of regressors (Chevan and Sutherland 1991; Grömping 2006). Cohen’s d statistic was calculated as an index of effect size, as the difference between subgroup means divided by pooled standard deviation. Statistical significance was considered present for a two-tailed p value of <0.05. All statistics were analyzed using R v. 3.1.2. Data are presented as mean ± standard deviation (SD) or median (inter-quartile range) as appropriate.

Results

Study population

The original dataset comprised n = 1411 patients of whom n = 161 patients were excluded due to valvular heart disease, leading to a final dataset for analyses comprising n = 1250 patients. Basic characteristics are shown in Table 1. In a majority of patients, the indication for coronary angiography was ST-elevation myocardial infarction (n = 711; 57 %; Table 2). Table 3 shows hemodynamic parameters for the entire study population including indices of LV preload (LVEDV), afterload (EA) and contractility (ELV). There were 133 patients on CCB of whom 90.2 % were documented to be hypertensive.

Table 1 Basic characteristics of study population
Full size table
Table 2 Blood tests, angiograms and echocardiograms
Full size table
Table 3 Hemodynamic data
Full size table

Clinical and pathophysiological correlates of aortic root dilation

Patients with aortic root dilation (n = 1129 %) were older and more commonly hypertensive as shown in Table 1. Among 793 hypertensives, 11.3 % had aortic root dilation (n = 90) vs. 4.6 % (n = 21) of 457 non-hypertensives (Chi2 = 15.2, p < 0.001). As shown in Fig. 2, patients with aortic root dilation had lower EA, larger LVEDV and larger SV than patients with normal aortic root size. They had similar number of diseased vessels on angiography (Table 2) and similar ELV to patients with normal aortic root size (Fig. 2), indicating preserved contractility. Patients with aortic root dilation had higher pulse pressure (56.6 vs. 51.6 mmHg, p < 0.01) but similar CA to patients with normal aortic root size (0.8 vs. 0.8 mmHg mL m−2, p = ns), in keeping with their SV being larger (Fig. 2). Their LVEF was higher (52.4 vs. 49.6 %, p = 0.03). There was a trend to greater use of CCB in patients with aortic root dilation (16 vs. 10 %) which was statistically non-significant (p = 0.07).

Fig. 2
figure 2

Favorable associations between aortic root dilation (x-axes) and both stroke volume and its principal determinants (y-axes). Panels show stroke volume (a), effective arterial elastance (LV afterload; b) end-systolic LV elastance (LV contractility; c) end-diastolic volume (LV preload; d) and heart rate (e). Patients with normal vs. dilated aortic root are shown in blue vs. red (color figure online)

Full size image

Correlations with SV are shown in Table 4 (Model 1). In multivariable analysis, aortic root size remained a significant predictor (p = 0.002), accounting for 2.4 % of the total explained variance of SV after adjusting for HR, LVEDV, ELV, SVR and CA (Table 4; Model 2). Effect size of aortic dilatation on SV was small–moderate in magnitude (Cohen’s d = 0.39).

Table 4 Multivariable regression models of determinants of left ventricular stroke volume
Full size table

Effect of antihypertensive therapy on the association between aortic root size and EA

Determinants of EA in multivariable analysis are shown in Table 5 (Model 3). There were significant interactions between the association between aortic root size and EA and the use of CCBs (p < 0.001) as well as use of β-adrenergic receptor blockers (p = 0.04), but not diuretics (p = 0.10), long-acting nitrates (p = 0.10), ACE-inhibitors (p = 0.41) or statins (p = 0.11).

Table 5 Multivariable regression models of determinants of effective arterial elastance
Full size table

There was a strong association between aortic root size and EA in patients taking a CCB (n = 133, p < 0.001) in whom increased aortic root size accounted for almost 8 % of the overall explanatory power of the model (total R 2 = 0.86). In contrast, in patients not taking a CCB (n = 1117) the association between aortic root size and EA was statistically non-significant in subgroup analysis (p = 0.44; Table 5; Models 4a, 4b). In contrast, the association with aortic root size was weak in patients on β-adrenergic receptor blockers (0.9 % of overall explanatory power attributable to aortic root size). Further subgroup analysis showed that patients taking vs. not taking a CCB had similar SVR (2.07 ± 0.54 vs. 2.13 ± 0.54 mmHg s m2 mL−1, p = 0.51) and CA (0.69 ± 0.24 vs. 0.76 ± 0.27 mL m−2 mmHg−1, p = 0.07) but larger SV (40.5 vs. 36.9 mL m−2, p = 0.003).

Discussion

The present study aimed to characterize the role of aortic root enlargement for LV afterload and SV. We found that a larger aortic root was associated with lower LV afterload (EA) and preload (LVEDV), similar contractility (ELV) and consequently larger SV. Moreover, we found a significant interaction effect indicating heterogeneity of the association between aortic root size and afterload as patients on CCB therapy exhibited a particularly strong association: 8 % of total explained EA was attributable to aortic root size in this subgroup.

Aortic root size, afterload and stroke volume

A principal end-point of the current analysis was EA, effective arterial elastance, which represents global afterload, i.e., a ‘lumped’ measure of the input impedance that opposes ventricular outflow (Kelly et al. 1992; Sunagawa et al. 1985). Aortic root size was positively associated with higher age but inversely associated with EA. Patients with aortic dilation had similar contractility (ELV) but greater preload (LVEDV) and larger SV. This is in keeping with analytical considerations suggesting that aortic dilation should have a beneficial effect on afterload and SV. As expected, the proportion of afterload attributable to aortic properties was small in numerical terms: the amount of variance explained by the regression model that was attributable to aortic diameter, after adjusting for other determinants of afterload, was 3.6 % (Table 5). This figure is analogous to previously published in vivo determinations of afterload by members of our group where ZC (derived from central flow and non-invasively determined central blood pressure waveforms) accounted for 3.5–5.0 % of total arterial resistance (Johnson et al. 2013; Sahlén et al. 2011). Invasive estimates of the proportion attributable to ZC have been performed mainly in individuals that were healthy or at least free of vascular disease and have been in the range 4–7 % (Kelly et al. 1992; Murgo et al. 1981). Non-invasive measurements have yielded estimates in the range 8–11 % in apparently healthy middle-aged subjects (Segers et al. 2007) and approximately 14–17 % in patients with hypertension and heart failure (Desai et al. 2009).

Mitchell et al. recorded ZC in 167 clinical trial participants with systolic hypertension studied off antihypertensive therapy. Patients with high pulse pressure were found to have higher pulse wave velocity, smaller aortic root diameter, greater aortic wall stiffness and higher ZC, suggesting a mismatch between pulsatile flow and diameter in patients with inappropriately small aortic root (Mitchell et al. 2008). While the population studied by Mitchell et al. was different to that of the present study (mean MAP >40 mmHg higher, 100 % were hypertensive), we are able to confirm and extend their findings by demonstrating that the aortic root diameter is indeed an independent determinant of global LV afterload and forward stroke volume. In contrast to what Mitchell et al. found, we found a small but significant increase in pulse pressure in patients with aortic root dilatation (+5 mmHg) which is likely to be attributable to a larger SV, which in turn is determined by several important differences in central hemodynamics of which the aortic root diameter is in fact one. Future research will need to demonstrate how this differs from the hyperdynamic circulatory state that causes systolic blood pressure elevation mainly in younger hypertensives (Fouad et al. 1978).

Contrary to the view that pulsatile load drives aortic root dilation, longitudinal data from the Framingham Heart Study have shown that individuals that develop aortic root dilation actually have lower baseline pulse pressure and higher diastolic blood pressure (Lam et al. 2010). This suggests aortic remodeling may conceivably be an adaptive response to an age-related increase in pulse pressure and that individuals with blunting of this response may exhibit a smaller aortic root and associated higher LV afterload (Mitchell et al. 2003). A positive association between SV and aortic root size has also been observed in other populations including the Losartan Intervention For Endpoint reduction (LIFE) trial (Bella et al. 2002; Cipolli et al. 2009; Palmieri et al. 2001).

Mechanistically, aortic root dilation occurs either due to hemodynamic stress (Altit et al. 2014) or due to local enzymatic processes in the aortic wall including inflammation and activation of neutrophil-derived proteases (Lindeman et al. 2009). It is conceivable that the role of hemodynamically driven, ‘appropriate’ dilation in elderly individuals may receive insufficient recognition, possibly due to the association between atherosclerotic vessel inflammation and dilation of other arterial segments in the body (Ito et al. 2008). A strength of the present report is, therefore, the fact that the population under study did consist almost exclusively of patients with established coronary artery disease and presumably poor endothelial function with ongoing enzymatic activation of arterial walls. Longitudinal investigations in this field do exist, as based on Framingham Heart Study and Cardiovascular Health Study, but were limited by the fact that they predominantly included community-dwelling individuals. As such, they may not have been able to conclusively demonstrate whether hypertensive subjects with unfavorable aortic root characteristics are at increased risk of heart failure (Gardin et al. 2006; Lam et al. 2013). Future research will need to collect longitudinal data on both characteristic impedance and aortic wall stiffness in an appropriately selected group of patients at risk of heart failure to answer this important question.

Impact of aortic root size on antihypertensive therapy

In a predefined interaction analysis, we noted that the association between aortic root size and afterload was heterogeneous and larger among patients taking CCB therapy, the proportion of explained variance attributable to aortic size in that subgroup being 8 %. Other antihypertensive drugs did not demonstrate an interaction of this magnitude although a directionally similar association was found for β-adrenergic receptor blockers.

Central arterial stiffness is a strong predictor of cardiovascular events (28) and there is growing awareness that antihypertensive drugs differ between classes in their ability to alter central hemodynamics and reduce aortic blood pressure (Cushman et al. 2001; Millar et al. 1999). Previous data have shown that CCB therapy reduces central blood pressure more than diuretics (Matsui et al. 2009) and that the vasodilating β-adrenergic receptor blocker nebivolol produces a greater reduction in central pressure (Dhakam et al. 2008) than the β1-receptor-selective agent atenolol (Mackenzie et al. 2009). The largest published study in this field was the Conduit Artery Functional Evaluation (CAFE) study which compared central blood pressure between patients on a CCB-based vs. a β-adrenergic receptor blocker-based antihypertensive regimen. Patients on CCB therapy were found have a central systolic pressure 12 mmHg lower than brachial pressure vs. a peripheral-to-central difference of only 8 mmHg in β-adrenergic receptor blocker-treated patients (Williams et al. 2006). While the present study did not evaluate the exact physiological basis for an effect of CCB therapy, it is known that the main mechanism for blood pressure reduction with CCB therapy is inhibition of voltage-dependent L-type calcium channels of vascular smooth muscle cells (SMC) (Berkels et al. 2003). Animal experiments specifically examining aortic tissue have shown that depolarization-induced contraction does control both basal and active tension of aortic SMCs via activation of L-type calcium channels (Fransen et al. 2012) which in turn may be inhibited by administering CCBs (Michiels et al. 2014). Murine aortic SMCs have also been demonstrated to be more sensitive to the vasodilator effect of nitric oxide in the presence of the L-type CCB nifedipine (Van Hove et al. 2009). Whether the production of nitric oxide per se is also favorably influenced by CCBs is still unknown, however, as the impact of antihypertensive drugs on endothelial function varies significantly between different vascular territories (Virdis et al. 2011). A putative mechanism to explain the significantly greater role played by aortic root size in patients on CCB therapy may be alterations in aortic root resistive or viscoelastic properties over and above the impact on blood pressure (which we adjusted for in multivariable analysis). However, in contrast to other arterial segments where viscoelasticity has been shown to change with drug administration (Safar et al. 1983; Safar et al. 1987; Vatner et al. 1980), animal experiments evaluating aortic properties have yielded conflicting results with an effect of aortic SMC tone on vessel elasticity being observed in some (Armentano et al. 1995; Barra et al. 1993) but not all reports (Niederhoffer et al. 1997; Van Gorp et al. 1996).

Limitations

This study has several limitations: (a) all patients had acute coronary syndrome, i.e., there was no normal control group; (b) some important but less common cardiac co-morbidities (congenital heart disease, cardiomyopathy) are not available through our data sources which remains a potential source of error; (c) CCBs were treated as a class; while the large majority of patients can be assumed to take non-dihydropyridine CCBs, the exact drugs used as well as duration were unfortunately not available to us. (d) While 90 % of CCB-treated patients were documented to have hypertension, it is conceivable that CCB therapy was prescribed for another indication (coronary spasm, refractory chest pain) in the remainder. (e) While we did find the association between aortic root size and afterload heterogeneous across subgroups taking vs. not taking CCB therapy, we did not record central pressure and flow nor directly analyze aortic wall stiffness. (f) We used a method for estimating ELV that assumes that aortic volume and pressure are linearly related and that V0 is zero. However, this is likely to play a minor role under normal physiological circumstances (Burkhoff et al. 2005; Kass and Kelly 1992) and obtaining true ELV based on invasively determined pressure–volume loop data is unlikely to alter the findings of the study which mainly focused on SV and EA. (g) The data in the present study were acquired in a predominantly Chinese population which may influence the external validity of findings (Virmani et al. 1991). (h) Differences between the multiple echocardiographs and transducers used may represent a source of error in image acquisition and (i) scaling was done to BSA in a ratiometric, non-allometric, fashion.

Conclusions

We used non-invasive hemodynamic analyses to investigate a large group of well-characterized patients. The size of the aortic root was independently associated with SV (positively) and EA (inversely). This suggests that dilation of the aortic root, as seen in elderly individuals and hypertensive patients, may in fact be favorable, possibly by mitigating the adverse impact of concomitant aortic wall stiffening. The association between aortic root size and afterload was heterogeneous, and stronger in patients taking CCBs. Future research will need to elucidate the nature of this association and externally validate findings outside ACS patients.