Forward and backward wave morphology and central pressure augmentation in men and women in the Framingham Heart Study

Abstract

Central pressure augmentation is associated with greater backward wave amplitude and shorter transit time and is higher in women for reasons only partially elucidated. Augmentation also is affected by left ventricular function and shapes of the forward and backward waves. The goal of this study was to examine the relative contributions of forward and backward wave morphology to central pressure augmentation in men and women. From noninvasive measurements of central pressure and flow in 7437 participants (4036 women) from 19 to 90 years of age (mean age 51 years), we calculated several variables: augmentation index, backward wave arrival time, reflection factor, forward wave amplitude, forward wave peak width and slope of the backward wave upstroke. Linear regression models for augmentation index, adjusted for height and heart rate, demonstrated non-linear relations with age (age: βx00302; =4.6±0.1%, P<0.001; age²: βx00302;=−4.2±0.1%, P<0.001) and higher augmentation in women (βx00302; =4.5±0.4%, P<0.001, model R²=0.35). Addition of reflection factor and backward wave arrival time improved model fit (R²=0.62) and reduced the age coefficients: age (βx00302; =2.3±0.1%, P<0.001) and age² (βx00302; =−2.2±0.1%, P<0.001). Addition of width of forward wave peak, slope of backward wave upstroke and forward wave amplitude further improved model fit (R²=0.75) and attenuated the sex coefficient (βx00302;=1.9±0.2%, P<0.001). Thus, shape and amplitude of the forward wave may be important correlates of augmentation index, and part of the sex-difference in augmentation index may be explained by forward and backward wave morphology.

The central arterial pressure waveform is the summation of forward traveling waves generated by left ventricular contraction and backward traveling waves returning from the periphery. Backward waves increase arterial pressure pulsatility while reducing distal flow pulsatility.¹ Pulse pressure and proximal aortic stiffness increase substantially with advancing age² and are known predictors of cardiovascular risk.^3–5 Augmentation index (AI), which represents the relative contribution of the backward wave to central pulse pressure, increases from young adulthood to midlife.² Traditionally, pressure augmentation has been attributed to increased aortic stiffness, which increases pulse wave velocity (PWV) and causes premature arrival of the backward wave in systole.^1;6 However, this explanation may be overly simplistic, because carotid-femoral PWV increases dramatically from the 5^th decade of life onward, whereas AI reaches a maximum in the 5^th decade then falls thereafter.² Previous studies have suggested that the forward wave^7;8 and reservoir pressure⁹ may be primary determinants of pressure augmentation and central blood pressure. Furthermore, women have markedly higher AI than men across the lifespan, which has been attributed to the early return of the backward wave because of shorter stature. However, height does not fully account for the sex difference in AI.¹⁰ Marked differences in ventricular contraction in men and women have been demonstrated in previous studies^10–12 and may contribute to sex differences in AI.

The ventricular response to a given backward wave may affect the balance between late systolic pressure augmentation versus flow deceleration.¹³ In addition, alterations in the nature and location of the reflecting site, in association with marked aortic stiffening after midlife, may alter the shape of the backward pressure wave.¹⁴ We hypothesized that variation in the ability of the left ventricle to sustain a high level of ejection in systole, as assessed by the width of the forward wave peak (time above 80% of peak amplitude), would be associated with increased pressure augmentation and that the shape of the backward wave may contribute to age- and sex-related differences in AI. Accordingly, the present investigation assessed the contributions of standard correlates of AI and novel waveform morphology variables to the inter-individual variation in AI across the age spectrum and with sex in a large community-based sample.

Methods

Study participants

The Framingham Offspring, Third Generation and Omni cohorts have been described;^15;16 details can be found in the online-only Data Supplement (please see http://hyper.ahajournals.org). The Boston University Medical Center Institutional Review Board approved the protocol, and all participants gave written informed consent.

Satisfactory evaluation of central pressure-flow relations was obtained in 7470 of 7674 participants. Cases without a carotid pressure inflection point detected before the end of the systolic ejection period (n=31, 0.4% of the cohort) were excluded from AI calculations. Additional exclusions for missing covariate data gave a final sample size of 7437 participants (97% of the original 7674 participants).

Noninvasive hemodynamic data acquisition and analysis

Details of the noninvasive hemodynamic protocol and analyses are summarized in the online-only Data Supplement (please see http://hyper.ahajournals.org). Augmentation pressure (AP) was calculated as the difference in peak pressure and pressure at the inflection point of the carotid waveform if the inflection point preceded peak pressure. If peak pressure preceded the inflection point, AP was 0. AI was calculated as previously described.¹⁷ Briefly, AI was computed by taking the difference between peak pressure and pressure at the inflection point and dividing that difference by central pulse pressure. AI was negative if peak pressure preceded the inflection point and positive if the inflection point preceded peak pressure. An alternate version of AI (AI_BWF) was similarly calculated except that backward wave foot time (see below) was used instead of the carotid inflection point to distinguish the transition point. Central pressure and flow waves were separated into their forward and backward components¹⁸ and forward wave peak width was determined by finding the width (ms) of the forward wave at 80 percent of maximum amplitude (Figure 1A). Width of the forward wave peak measured the duration of near maximal LV ejection and quantified the extent to which the forward wave had a plateau-like (Figure 1A, high AI group) or peaked (Figure 1A, low AI group) shape. Backward wave foot arrival time was found by starting at the maximum derivative between 20% and 80% amplitude of the upstroke of the backward wave and then extrapolating a tangent line to an intersection with the zero pressure baseline (Figure 1B). The indexed maximum slope of the backward wave upstroke measured the steepness of the rising edge of the backward wave, which could be indicative of the LV contraction rate or the nature of the reflecting sites. The maximum slope of the backward wave upstroke was indexed to the amplitude of the backward wave, which effectively created an inverse time constant (i.e., units are Hz) that assessed the relative rate at which the backward wave reached a near maximal value. Such indexing of pressure derivatives was proposed by Mirsky et al to allow for comparison of cardiac function among individuals.¹⁹ Reflection factor was defined as backward wave amplitude divided by forward wave amplitude and served as an indicator of the magnitude of global wave reflection.

Figure 1. Forward and backward wave morphology.

The cohort was divided into groups based on quartiles of augmentation index (AI). The low AI and high AI plots represent the averaged waves for participants in the 1^st and 4^th quartiles, respectively. Panel A shows forward wave morphology. The horizontal line represents 80% of the forward wave pressure amplitude. Panel B shows the maximum slope (time derivative) of the upstroke of the backward wave. The foot of the backward wave, marked by a vertical tick, is determined by extrapolating a tangent line from the point of the maximum derivative to the baseline.

Simulated pressure waves

Simulated pressure waves were created to illustrate relative contributions of forward and backward waves to pressure augmentation by altering the 1^st and 4^th AI quartile backward waves without changing the forward waves (Supplement Figure S1), as described in the online-only Data Supplement (please see http://hyper.ahajournals.org).

Statistical analyses

Characteristics of the sample were tabulated by sex. Key hemodynamic variables were plotted by sex and age decades. Multiple linear regression models were examined to assess correlates of AI, AP and AI_BWF. Only participants with AP greater than zero were included in AP models.

The base model for all multivariable analyses included standard correlates of AI, specifically: age, age², sex, height and heart rate. Values of age were centered (by subtracting mean age) then squared. All continuous variables were standardized (mean=0, SD=1) in order for model coefficients to be interpretable as a difference in AI (in native % units) or AP (mm Hg) per 1 SD difference in the independent variable. The second model added backward wave foot time and reflection factor. The third model also added forward wave amplitude, width of the forward wave peak and log-transformed indexed slope of backward wave upstroke. To identify parsimonious models, the final model for each dependent variable was examined using both stepwise and backward selection on hemodynamic variables. In the final regression models, we confirmed that all independent variables had variance inflation factors less than 3. Additional models for AI were examined in participants who were not taking anti-hypertensive medications. We included 5 hemodynamic variables in multivariable analyses; hence, a two-sided P-value of <0.01 was considered statistically significant.

Results

Clinical characteristics of the study sample are presented in Table 1. The cohort spans broad age limits from 19 to 90 years and consists of 54% women. Average body mass index was in the overweight range. Approximately 1 in 4 participants was taking antihypertensive medication. Hemodynamic variables are presented in Table 2. As expected, AI and AP were substantially higher in women regardless of the method used to define AI. The forward wave peak was wider and inflection point timing and backward wave foot timing were earlier in women (Table 2). Variation in key hemodynamic variables as a function of age is presented in Figure 2. AI in women was higher than in men, peaked between 50 and 60 years of age and was lower with age thereafter. AI in men was higher with age through the sixth decade and plateaued thereafter. AP was higher with age in women and men until midlife followed by a late life plateau in women and steadily higher values with age throughout the lifespan in men. Forward wave amplitude was slightly lower with age before age 50 years and was steadily higher with age thereafter in both sexes. Reflection factor and width of the forward wave peak were higher in women until the eighth decade of life, when they fell to approximately the same mean values as in men. In women, slope of the backward wave upstroke was higher with age through the fifth decade of life then fell, whereas in men this slope fell throughout the adult lifespan. Carotid inflection point timing reached a nadir in midlife and was earlier in women. Backward wave foot arrival was consistently earlier in women as compared with men and, in contrast to inflection point timing, approached a minimum in late life. Age relations for AI_BWF and central pulse pressure are shown in Supplement Figure S2. Age relations for AI_BWF were similar to AI, although the late life decline in AI in women was less pronounced for AI_BWF. In women and men, AI_BWF increased dramatically between 20 and 60 years of age whereas central pulse pressure was unchanged. After 60 years of age, AI_BWF plateaued whereas central pulse pressure increased substantially.

Table 1.

Characteristics of the sample

Clinical Characteristic	Men (n=3401)	Women (n=4036)
Age, years	51 ± 15	51 ± 16
Omni cohort	284 (8)	399 (10)
Height, cm	176 ± 7	162 ± 7
Weight, kg	88 ± 16	71 ± 16
Body mass index, kg/m2	28.2 ± 4.7	26.7 ± 5.9
Systolic blood pressure, mm Hg	124 ± 15	119 ± 18
Diastolic blood pressure, mm Hg	77 ± 10	73 ± 9
Heart rate, min−1	60 ± 10	63 ± 10
Fasting glucose, mg/dL	103 ± 22	97 ± 20
Triglycerides, mg/dL	129 ± 97	106 ± 65
Total/HDL cholesterol	4.1 ± 1.4	3.2 ± 1.0
Diabetes	331 (10)	249 (6)
Hypertension treatment	925 (27)	931 (23)
Lipid lowering treatment	882 (26)	724 (18)
Smoking	465 (14)	505 (13)
Hormone replacement therapy	-	286 (7)
Prevalent CVD	292 (9)	211 (6)

Continuous variables are expressed as mean ± SD, categorical variables as n (%).

Table 2.

Hemodynamic variables

Variable	Men (n=3401)	Women (n=4036)
Augmentation index based on inflection point, %	6.7 ± 12.6	14.2 ± 12.5
Augmentation index based on backward wave foot, %	9.4 ± 15.5	17.4 ± 14.0
Reflection factor	0.34 ± 0.06	0.36 ± 0.07
Forward wave amplitude, mm Hg	51 ± 14	49 ± 16
Augmentation pressure, mm Hg	5.9 ± 6.9	9.5 ± 8.7
Width of the forward wave peak, ms	192 ± 33	206 ± 28
Slope of backward wave upstroke, mm Hg/ms	0.13 ± 0.06	0.13 ± 0.06
Indexed slope of backward wave upstroke, ln(s−1)	1.96 ± −0.28	2.01 ± 0.27
Inflection point timing, ms	144 ± 24	130 ± 26
Backward wave foot timing, ms	125 ± 38	112 ± 31

Data are mean ± 1 standard deviation.

Figure 2. Augmentation index covariates by age decade and sex.

Points centered between two ticks represent the mean value for that decade. Error bars represent 1 standard deviation. AI, augmentation index; AP, augmentation pressure; FW, forward wave; BW, backward wave.

Simulated waveform analysis illustrated contributions of forward and backward waves to AI (Supplement Figure S1). The simulated 4^th quartile pressure waveform (Supplement Figure S1.D) markedly underestimated apparent augmentation in the native 4^th quartile waveform (Supplement Figure S1.B). Similarly, the simulated 1^st quartile pressure waveform (Supplement Figure S1.C) showed substantial persistent augmentation as compared to the native 1^st quartile waveform (Supplement Figure S1.A). The simulated 1^st quartile waveform (Supplement Figure S1.C) had higher apparent augmentation than the simulated 4^th quartile waveform (Supplement Figure S1.D).

Hemodynamic correlates of AI are presented in Table 3 and Supplement Table S1. A base model for AI that included age, age², sex, heart rate, and height explained 35% of the variance in AI (Table 3). When backward wave foot time and reflection factor were added to the base model, the regression coefficients for the age and age² variables were reduced in magnitude by approximately half, yet there was minimal change in the coefficient for sex (Table 3, Model 2). When width of the forward wave peak, forward wave amplitude, and slope of the backward wave upstroke were added to the model, the coefficient of determination of the model improved substantially, the coefficients for age and age² terms were further attenuated, and the coefficient for sex was reduced by more than half (Table 3, Model 3). These models reveal that amplitude, shape and timing of the forward and backward waves contribute to inter-individual variation in AI. Similar models for AI were obtained when only participants not being treated for hypertension were included (n=5581, Supplemental Table S2).

Table 3.

Correlates of augmentation index

Variable	Model 1 (R2 =0.35)			Model 2 (R2 =0.62)			Model 3 (R2 =0.75)
	βx00302;	SE	P	βx00302;	SE	P	βx00302;	SE	P
Age	4.55	0.13	<0.001	2.29	0.12	<0.001	0.38	0.11	<0.001
Age2	−4.20	0.13	<0.001	−2.23	0.10	<0.001	−1.53	0.09	<0.001
Female	4.47	0.36	<0.001	4.34	0.28	<0.001	1.88	0.23	<0.001
Height	−2.79	0.19	<0.001	−0.56	0.15	<0.001	−0.25	0.12	0.033
Heart rate	−3.60	0.13	<0.001	0.15	0.11	0.170	1.12	0.11	<0.001
Backward wave foot time	--	--	--	−1.49	0.13	<0.001	−2.14	0.11	<0.001
Reflection factor	--	--	--	7.62	0.12	<0.001	4.76	0.12	<0.001
Forward wave peak width	--	--	--	--	--	--	5.67	0.13	<0.001
Slope of backward wave upstroke	--	--	--	--	--	--	3.41	0.08	<0.001
Forward wave amplitude	--	--	--	--	--	--	0.78	0.10	<0.001

Slope of backward wave upstroke is indexed to backward wave amplitude and natural log transformed. All continuous covariates are standardized such that βx00302; represents the absolute difference in AI (in units of %) per 1 SD of the covariate. The sex term represents the absolute difference in AI in women compared with men.

Models for AP, including only participants with AP greater than zero (n=5962), are shown in Supplement Table S3. The regression coefficient for forward wave amplitude was substantially larger, while the coefficient for forward wave peak width was substantially smaller, in the AP model as compared with the AI model. Models for AI_BWF are shown in Supplement Table S4 and are largely similar to those for AI (Table 3), except that, as expected, timing of the backward wave foot had a stronger relation with AI_BWF than AI.

Discussion

This study evaluated standard correlates of AI (reflection factor and backward wave foot time), forward wave amplitude, and novel measures of forward and backward waveform morphology (width of the forward wave peak and slope of the backward wave upstroke), and observed that component waveform morphology is an important correlate of inter-individual variation in AI across the age spectrum and between sexes in this large, community-based sample.

Synthesized waveforms in Supplement Figure S1 show the attenuated effect on apparent AI of altering the timing and amplitude of the backward wave alone, without modifying the forward wave. If timing and amplitude of the backward wave were the primary determinants of AI, we would have expected the simulated 1^st and 4^th quartile pressure wave (Supplement Figure S1.C and D) to have the same amount of augmentation as the native 1^st and 4^th quartile pressure waves (Supplement Figure S1.A and B), respectively. Misestimation of augmentation in the simulated pressure waveforms suggests that forward and backward wave shape contribute considerably to the observed augmentation in the ensemble-averaged waveforms, consistent with results presented in Table 3.

Cardiac ejection pattern, assessed by width of the forward wave peak, is an essential component of central pressure augmentation. As illustrated in Figure 1A, participants with lower AI have an early forward wave peak followed by a sharply falling slope, whereas those with higher AI have a broader forward wave peak that allows the backward wave to emerge as the dominant peak. Women have a longer systolic ejection time,¹⁰ more vigorous left ventricular systolic function, and a smaller end-systolic chamber volume,^11;12 consistent with our observation that women have a wider forward wave peak (Figure 2). Table 3, Model 3 shows that the sex difference in AI was reduced by nearly 60% when the shapes of the forward and the backward waves and forward wave amplitude were added as covariates. These observations suggest that the left ventricular ejection pattern and shape of the backward wave, in addition to height, are key components of AI and account for part of the difference in AI observed in men versus women.

In the arterial system, the observed backward wave represents the summation of multiple backward waves from various reflecting sites.¹ As aortic impedance begins to increase after 50 years of age, mismatch at the interface between aorta and muscular arteries is reduced and reflecting sites shift distally.²⁰ The global reflection coefficient becomes progressively less of a real reflection factor at a simple branch point between aorta and muscular arteries and more of a complex reflection factor at the interface with a terminal Windkessel,²¹ which has a complex impedance. The associated complex reflection factor further blunts the upstroke of the backward wave and may contribute to flat or falling AI with advancing age after midlife.

In contrast to AI, AP represents the calibrated extent of pressure augmentation. AP is related to backward wave amplitude, which represents the product of forward wave amplitude and the reflection factor. Therefore, AP is related to forward wave amplitude, which varies considerably over the lifespan, as well as the reflection factor, which varies modestly over the lifespan and actually falls after midlife, particularly in women (Supplement Table S3, Figure 2). We have previously shown that variation in pulse pressure is largely attributable to forward wave amplitude.² The present analyses demonstrate that central AP is also strongly related to forward wave amplitude. Therefore, when evaluating interventions that reduce AP, and thereby reduce pulse pressure, one should consider potential effects of the intervention on forward wave amplitude and morphology in addition to relative wave reflection.

In younger people, mean timing of the inflection point is similar to mean timing of the backward wave foot, but after midlife mean timing of the inflection point is later than that of the backward wave foot (Figure 2). In order to see an inflection point (i.e. second derivative zero-crossing) in the pressure waveform, the (positive) second derivative of the backward wave must equal then exceed the absolute value of the (negative) second derivative of the forward wave. Therefore, as the reflection factor and the indexed slope of the backward wave upstroke begin to decrease late in life, it becomes increasingly difficult for the backward wave to accelerate faster than the forward wave is decelerating. As a result, the inflection point may be detected after the timing of the backward wave foot or not at all. Importantly, regardless of whether AI was calculated based on inflection point timing (Table 3) or backward wave foot timing (Supplement Table S4), relations with hemodynamic covariates were comparable. AI_BWF was more strongly related to timing of the backward wave foot, which was expected given that any misclassification of the timing of the backward wave foot would appear as effect in the relation with calculated AI_BWF.

Limitations

We measured AI using the pressure at the time of the inflection point of the carotid waveform indexed to carotid pulse pressure, as has been the standard method of measurement in previous studies.^1;7;9 However, it is likely that the relative contribution of backward wave arrival time to AI has been over-estimated in previous studies, because using the inflection point both to define AI and as a covariate of AI creates a mathematical dependency of AI upon the arrival time of the backward wave. In order to avoid this collinearity, we used the timing of the backward wave foot rather than the timing of the inflection point in the models in Table 3. Some have suggested that very early arrival of the backward wave may obscure the inflection point altogether in older people,^1;22 resulting in an absent inflection point and an underestimation of AI. We found few individuals with no inflection point (0.4% of the cohort) and excluded them from the present analyses. We analyzed wave reflection by using wave separation based on a simplified transmission line model of the arterial system that assumes the aorta is a tube that is terminated by a Windkessel. An alternative approach separates the backward wave into (static) Windkessel and (traveling) reflected wave components.⁹ In light of this controversy, we have referred to the reflected wave as a backward wave throughout. Our cohorts consisted primarily of white participants of European descent, though a smaller subset of non-white participants was included (9.2%). Therefore, our results may not be generalizable to all races and ethnicities. The cross-sectional, observational design of this study limits our ability to make causal inferences regarding the etiology of hemodynamic variables. Other generational factors may have contributed to the observed differences between age groups. Our study also has several strengths, including large sample size and routine ascertainment of a comprehensive noninvasive panel of arterial function measures and coexistent cardiovascular disease risk factors in a community-based sample, which provides excellent statistical power, facilitates adjustment for multiple covariates, and limits referral bias.

Perspectives

Central pressure augmentation is a complex phenomenon involving the interaction of forward and backward waves. The assumption that higher AI represents the integrated effects of aortic stiffening, increasing PWV and earlier return of a potentially larger backward wave may be overly simplistic. Indeed, we have shown that timing and amplitude of the backward wave only partially explain age- and sex-related differences in AI, whereas characteristics of the left ventricular ejection phase and morphology of the backward wave contributed substantially, particularly to sex differences. Furthermore, prior studies have shown that variability in pulse pressure is primarily determined by variability in the amplitude of the forward wave.² Therefore interventions that attenuate the ventricular response to a backward wave, such as preload reduction, or that address aortic stiffness and mismatch between central aortic diameter and flow,²³ as compared to those focused on reducing wave reflection per se, could be studied as potential options for lowering pulse pressure and augmentation index.

Novelty and Significance.

What is new?

• Variability in augmentation index in men and women is only partially attributable to backward wave timing and amplitude.

• Sex differences in augmentation index are related to left ventricular ejection phase characteristics and backward wave shape rather than only timing and relative amplitude of the backward wave.

What is relevant?

• Treatment of elevated pulse pressure by reducing wave reflection may fail to address ejection phase characteristics and forward wave amplitude, which are important contributors to pressure augmentation and pulse pressure.

Summary

In contrast to the traditional view that attributes augmentation index solely to backward wave timing and amplitude, a sizeable component of variability in augmentation index arises from forward and backward wave morphology and forward wave amplitude. Interventions that alter the ventricular response to a given backward wave, such as preload reduction, may be effective at reducing augmentation.