Increased Wave Reflection Rather Than Central Arterial Stiffness Is the Main Determinant of Raised Pulse Pressure in Women and Relates to Mismatch in Arterial Dimensions

A Twin Study

Hemodynamic measurements

Brachial blood pressure was measured in duplicate, using a validated oscillometric device (Omron 705CP, Omron, Tokyo, Japan), after subjects had been seated in a quiet room for at least 10 min. Radial pulse waveforms and measurements of central arterial stiffness (carotid-femoral pulse wave velocity [PWV]) were obtained with the subject in a supine position using the SphygmoCor system (Atcor, West Ryde, Australia). Applanation tonometry of the radial artery with a high-fidelity transducer (Millar Instruments, Houston, Texas) was used to obtain an ensemble averaged radial pulse. The radial artery pressure waveform was calibrated to supine brachial blood pressure. The inbuilt transfer function in the SphygmoCor system provided a corresponding aortic pulse waveform from which cPP, P1, ΔP_aug, and central systolic blood pressure were identified (16). Carotid-femoral PWV was calculated from sequential recordings of electrocardiogram-referenced carotid and femoral pressure waveforms obtained by tonometry using the same device and transducer. The path distance between the carotid and femoral sites was estimated from the distance between the sternal notch and femoral artery at the point of applanation. This reduces errors introduced by multiple measurements and is closer to the true path length than the carotid to femoral distance (17). PWV and cPP measurements were made in triplicate, and mean values were used for analysis. Only waveforms that passed the automatic quality control criteria of the SphygmoCor system were used.

Ultrasonography of the aortic root, abdominal aorta, and femoral arteries

The aortic root was visualized using 2-dimensional echocardiography from a long-axis parasternal view using a Siemens CV70 with a 4-MHz cardiac transducer, and its diameter measured 1 cm from the aortic valve between the anterior and posterior aortic root wall at end diastole. SV was measured in a subgroup of 186 subjects. The cross-sectional area of the aortic valve was estimated from diameter measurements and SV determined by multiplying the cross-sectional area by the velocity time integral obtained from Doppler flow velocity at the level of the aortic valve. To visualize the abdominal aorta, the epigastrium was scanned vertically at the xiphoid process using the same probe. The diameter of the abdominal aorta was measured 1 to 2 cm below the diaphragm at end-diastole. Left and right femoral arteries were visualized with a 13-MHz vascular probe 1 cm proximal to the bifurcation. Automated wall tracking software (Medical Imaging Applications, Coralville, Iowa) was used to measure diameter at end-diastole.

Statistical analysis

Data analysis was performed with SPSS version 14.0 (SPSS Inc., Chicago, Illinois). Genetic modeling was performed using Mx software (Mx statistical modeling, Medical College of Virginia, Richmond, Virginia). Subject characteristics are presented as mean ± SD unless otherwise stated. Student unpaired tand chi-square tests were used to test for differences in characteristics between MZ and DZ twins. Univariate regression analysis was first used to examine relationships of cPP, pPP, P1, and ΔP_augas dependent variables to PWV, mean arterial pressure (MAP), HR, arterial diameters, and potential confounding factors: age, height, weight, smoking status, total cholesterol, high-density lipoprotein cholesterol, and the presence of diabetes. Forward stepwise multiple regression analysis was then performed to examine potential independent determinants of cPP, pPP, P1, and ΔP_aug. Variables included in each analysis were age, HR, MAP, PWV, arterial dimensions (aortic root diameter, abdominal aortic diameter, the ratio of femoral to abdominal aortic diameter), and any other confounding variables found to significantly correlate with the dependent variable in the initial univariate analysis. The same analysis was also applied to the time of arrival of the reflected wave (T1) (Fig. 1). The analysis was repeated separately in women age <60 and ≥60 years, because the prognostic importance of PP changes at approximately 60 years in women (1), and age category (<60 and ≥60 years) interactions with variables entering the final model were tested. All statistical tests were done at the 5% level of significance.

Heritability analysis was performed using the classical twin model in which a greater similarity in MZ compared with DZ twins suggests a genetic influence, based on the fact that MZ twins are genetically 100% identical and DZ twins share approximately 50% of their segregating genes. This model assumes that MZ and DZ twins share their common environment to the same extent. Twin resemblance for each of the phenotypes was examined using the intraclass correlation coefficient for each zygocity. A higher intraclass correlation coefficient for MZ twins compared with DZ twins suggests a genetic influence on the phenotype. For further analysis of genetic and environmental contributions, the variance of each phenotype was assumed to derive from an additive genetic component (sum of individual effects of all loci that influence the trait), shared or unique environment components (ACE model). Measurement error is represented under unique environmental influences. Structural equation modeling implemented in Mx was used to estimate the parameters of the ACE model and corresponding confidence intervals.

Relationship of pPP and cPP to PWV and arterial dimensions

Mean cPP was less than pPP by 9 and 7 mm Hg in subjects age <60 and ≥60 years, respectively (each p < 0.001). Both cPP and pPP correlated with MAP, HR, age, PWV, high-density lipoprotein cholesterol, aortic root diameter, and smoking but not with height or weight by univariate analysis. However, in multiple regression analysis including all variables correlated with PP on univariate analysis, the only significant correlations were with MAP, HR, age, and PWV (each p < 0.0001) (Table 2).

Contribution of P1 and ΔP_augto cPP

Mean values of P1 were greater than those of ΔP_augfor subjects in each age group and in each tertile of the distribution of cPP (Fig. 2).However, ΔP_augcontributed proportionately more to the increase in cPP across the distribution than P1. Thus, in the whole cohort, P1 increased by a mean of 12.0 mm Hg (52% increase) when comparing women in the first and third tertiles of cPP but ΔP_augincreased by 14.1 mm Hg (196% increase). In women meeting a definition of isolated systolic hypertension (ISH) with peripheral systolic blood pressure >140 mm Hg and diastolic blood pressure <90 mm Hg, P1 was 37.5 compared with 26.7 mm Hg in women without ISH. Corresponding values of ΔP_augwere 21.6 and 12.7 mm Hg in women with and without ISH, respectively. In regression analysis (including ΔP_augand P1 as predictors of cPP), ΔP_augand P1 explained 76% and 22% of the variance in cPP. In women age <60 years, this was 73% and 24%, respectively, and in women age ≥60 years was 76% and 21%, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Central Pulse Pressure

Mean values of P1 and augmentation pressure in relation to central pulse pressure in women <60 and ≥60 years of age. P1 = height of first systolic shoulder.

Relationship of P1 and ΔP_augto PWV and arterial dimensions

In multiple regression analysis (incorporating all variables significantly correlated with P1 on univariate analysis), in the whole cohort and in each age group, P1 strongly positively correlated with PWV and MAP but less strongly correlated with HR, arterial diameters, and other variables (Table 3).There were significant interactions between age category and MAP and between age category and HR.

ΔP_augstrongly positively correlated with MAP and age and negatively correlated with HR. ΔP_augdid not significantly correlate with PWV. T1 did not independently correlate with PWV but independently correlated (standardized regression coefficient, β = 0.23, p < 0.001) with the ratio of femoral to abdominal aortic diameter. T1 also significantly correlated with height and (negatively) with MAP. There were significant interactions between age category and HR and between age category and the ratio of femoral to abdominal aortic diameter. In subjects age <60 years (as in the whole cohort) but not in subjects age ≥60 years, ΔP_augnegatively correlated with the ratio of femoral to abdominal aortic diameter. In the subsample of subjects in whom SV was assessed (Table 4),SV was weakly or nonsignificantly related to P1 and ΔP_aug, and the main findings with respect to correlations with PWV were unaltered: P1 remained strongly correlated with PWV, whereas ΔP_augdid not significantly correlate with PWV. The strength of the correlation between ΔP_augand the ratio of femoral to abdominal aortic diameter in subjects age <60 years was stronger, but that of P1 to the ratio of femoral to abdominal aortic diameter was no longer significant.

Heritability analysis

Intraclass correlations for all measures were greater for MZ compared with DZ twin pairs (Table 5),indicating a genetic influence on all phenotypic traits. Heritability estimates are shown in Table 6.In the full ACE model, heritability of PWV was modest, with an estimated genetic component of 34%. The estimated genetic components of cPP, P1, and ΔP_augwere 43%, 31%, and 62%, respectively. Heritability of ΔP_augwas reduced to 43% when corrected for height in addition to HR and MAP. PWV was the only measure for which shared environment was significant.

PP and PWV

PP is often regarded as a surrogate for PWV, ejection of blood into a stiff aorta resulting in a high systolic and pulse pressure. A correlation between PP (both peripheral and central) and carotid-femoral PWV is a universal finding in large epidemiological studies (18–22), but much of the variation in PP is not explained by PWV or other measures of central artery stiffness (18–22). Other factors invoked to explain PP include left ventricular ejection volume, pressure wave reflection, and arterial diameters (23). As in the present study, previous studies have found only a weak relation between left ventricular ejection volume and PP (20,21).

Contribution of P1 and ΔP_augto cPP

To our knowledge, this is the first study to examine separately the contribution of the 2 major components of PP, P1 and ΔP_aug, to cPP. Although P1, the component of PP generated by the outgoing pressure wave, was greater than ΔP_aug,the variance of ΔP_augwas greater than that of P1. Thus, most of the variability in cPP (and in pPP) was accounted for by variation in ΔP_augrather than P1. Partitioning of cPP into ΔP_augand P1 provides physiological insight into the mechanisms governing cPP. P1, the outgoing component of the pressure wave, might be expected to be mainly determined by PWV, which determines the Windkessel function of the central vessels. By contrast, arterial geometry might be expected to influence pressure wave reflection more strongly than arterial stiffness (although stiffness could also influence this through earlier return of reflected waves). In the present study, we found that P1 indeed highly correlated with PWV. However, ΔP_augdid not independently correlate with carotid-femoral PWV. This suggests that PWV is not a major determinant of ΔP_augand is consistent with other studies showing dissociation between measures of pressure wave reflection and PWV during interventions that influence vasomotor tone (24,25).

Our findings differ from those of Mitchell et al. (26) in the Framingham offspring cohort of men and women, where although ΔP_aug(obtained from carotid tonometry) was seen to account for a considerable proportion of variance in cPP, PWV accounted for a greater proportion of the variance in cPP than ΔP_aug. However, subjects with hypertension (47% of the potential study population) were excluded. The present study shows that ΔP_augpositively correlated with MAP and so would be expected to provide a greater contribution to cPP in subjects representative of the general population, including those with hypertension. The conclusions of our study also differ from those of Segers et al. (27) who examined wave reflection in middle-age men and women and found wave reflection to explain at most 26% of the variance in carotid PP. Segers et al. (27) used the augmentation index (AIx) (equal to ΔP_aug/cPP) as a measure of reflection in their regression models, which also included measures of input impedance. The proportion of variance in cPP explained by AIx compared with ΔP_augwill be reduced by cPP being itself a denominator of AIx. Our simple approach, where we have portioned cPP into the sum of P1 + ΔP_aug, provides less insight into impedance but provides an estimation of the relative contributions of P1 and ΔP_aug, which is limited only by the accuracy of the measurements.

Physiological and structural determinants of P1 and ΔP_aug

Both P1 and ΔP_augcould be influenced by arterial geometry. For a given intrinsic elasticity of the aortic wall (to which PWV is related by the Moens-Korteweg equation [28]), the functional compliance of the aorta is directly related to its diameter (29). Although this is offset by the fact that SV and HR are related to body size and aortic root area, P1 might be expected to be inversely related to aortic diameters, particularly those beyond the aortic root. However, one hypothesis of age-related stiffening is that elastin degeneration is associated with aortic dilation (30), and this might blunt or over-ride the expected inverse relation. In the present study, we found a weakly positive correlation of P1 with abdominal aortic diameter in subjects <60 years of age, a negative correlation in subjects ≥60 years of age, and no significant relation in the overall cohort. This result is consistent with recent studies that have found a modest inverse correlation of cPP with aortic root diameter in older women (20,21).

Wave reflections are thought to occur at discontinuities in impedance and thus to depend on the serial distribution of arterial diameter and elasticity, with most reflection arising distal to the aortic bifurcation. In the present study we measured abdominal aortic diameter as a measure of proximal artery dimensions and femoral artery diameter as a measure of distal arterial dimensions. Of all arterial dimensions, ΔP_augwas best predicted by the ratio of femoral to abdominal aorta diameter, with an inverse relationship between ΔP_augand femoral/aortic diameter. This is consistent with the amount and/or timing of reflection being determined by the mismatch between femoral and aortic dimensions or between other distal and proximal arterial dimension, with femoral and aortic dimensions acting as a surrogate for these. The highly significant independent association of T1, the time of arrival of the reflected wave, with femoral/aortic diameter suggests that arterial dimensions have a greater influence on the site of reflection than on the amount of reflection. Furthermore, the lack of association of T1 with PWV suggests that it is primarily the site of reflection that determines the time of arrival of reflected waves and over-rides the importance of transit time to and from the site of reflection. The association between ΔP_augand femoral/aortic diameter was driven by subjects <60 years of age. In subjects ≥60 years of age, in whom PWV is higher, it may be that serial distribution of elasticity becomes more important than that of diameter. Further studies including prospective studies will be required to explore the complex relation between ΔP_augand arterial diameter.

Heritability of P1 and ΔP_aug

To our knowledge, this is the first study to report heritability of the different components of cPP, P1, and ΔP_augin twins. Heritability of P1 (31%) was modest as was that of PWV (34%), consistent with P1 being determined in large part by PWV. The estimate for PWV is consistent with previous estimates of the heritability of PWV from family and twin studies between 0.26 and 0.40 (31,32). The heritability of ΔP_aug(62%) is greater than that observed for most phenotypic traits including systolic blood pressure, and suggests that if other factors such as measurement error are similar, factors additional to PWV that determine reflection have a high genetic component. The findings with respect to ΔP_augare similar to those of AIx, which is known to be highly heritable (33). Although our estimates of the heritability of aortic dimensions were relatively low, these estimates were relatively crude measures of the detailed geometry that determines wave reflection. The high heritability of ΔP_augcould be explained by high heritability of detailed arterial geometry, which is, in turn, determined by intrinsic geometry and smooth muscle tone.

Implications for systolic hypertension in women

The finding that ΔP_augaccounts for the majority of the variation in cPP and that it is highly heritable and independent of PWV has a number of important implications for systolic hypertension in women. It challenges the view that systolic hypertension arises primarily from irreversible stiffening of the aorta caused, for example, by collagen cross-linking and/or calcification. It raises the potential importance of ΔP_augas a risk factor in addition to the well-established risk associated with elevated PWV in hypertension (6,34). It suggests that drugs with a specific action to dilate muscular arteries (and therefore to increase distal/proximal dimensions and reduce wave reflection) could be effective in reducing PP and systolic pressure. Organic nitrates have such specificity for large muscular arteries (35), but their efficacy may be limited by tolerance, increase in oxidative stress, and possibly other nonhemodynamic adverse effects (36). The high heritability of ΔP_augsuggests that ΔP_augis an important trait to examine in relation to potential genes involved in hypertension and CVD.

Study limitations

The study was limited to female twins. The results are likely to be representative of women in the general population, since characteristics and CVD risk profiles of the Twins UK cohort are similar to that of the general population (37), and the mortality of twins is similar to that of the general population (38). The results cannot, however, be generalized to men. We estimated central arterial pressure waveforms from radial waveforms using the SphygmoCor transfer function. Most of the error inherent in this process, at least when computing cPP, is attributable to the limited accuracy with which peripheral blood pressure can be measured (16). When calibration is performed from brachial blood pressures (as in this study), the approach also ignores amplification from the brachial to the radial artery, which may lead to an overestimation of peripheral amplification (39). Although this affects absolute values of cPP and pPP, it does not affect the relative values of cPP and pPP (40) or those of ΔP_augand P1, and thus is unlikely to affect interpretation of the study. We made limited measurements of arterial diameters and cannot exclude the possibility that other relations between ΔP_augand arterial dimensions would emerge if dimensions of smaller arteries were available.