Use of the Frank-Starling mechanism during exercise is linked to exercise-induced changes in arterial load

Abstract

Effective arterial elastance(E_A) is a measure of the net arterial load imposed on the heart that integrates the effects of heart rate(HR), peripheral vascular resistance(PVR), and total arterial compliance(TAC) and is a modulator of cardiac performance. To what extent the change in E_A during exercise impacts on cardiac performance and aerobic capacity is unknown. We examined E_A and its relationship with cardiovascular performance in 352 healthy subjects. Subjects underwent rest and exercise gated scans to measure cardiac volumes and to derive E_A[end-systolic pressure/stroke volume index(SV)], PVR[MAP/(SV*HR)], and TAC(SV/pulse pressure). E_A varied with exercise intensity: the ΔE_A between rest and peak exercise along with its determinants, differed among individuals and ranged from −44% to +149%, and was independent of age and sex. Individuals were separated into 3 groups based on their ΔE_AI. Individuals with the largest increase in ΔE_A(group 3;ΔE_A≥0.98 mmHg.m²/ml) had the smallest reduction in PVR, the greatest reduction in TAC and a similar increase in HR vs. group 1(ΔE_A<0.22 mmHg.m²/ml). Furthermore, group 3 had a reduction in end-diastolic volume, and a blunted increase in SV(80%), and cardiac output(27%), during exercise vs. group 1. Despite limitations in the Frank-Starling mechanism and cardiac function, peak aerobic capacity did not differ by group because arterial-venous oxygen difference was greater in group 3 vs. 1. Thus the change in arterial load during exercise has important effects on the Frank-Starling mechanism and cardiac performance but not on exercise capacity. These findings provide interesting insights into the dynamic cardiovascular alterations during exercise.

the arterial system is increasingly recognized as an important modulator of cardiac performance (6), and a potent predictor of cardiovascular (CV) outcomes (30, 52). The arterial system imposes a load on the heart, both at rest and during exercise. A higher arterial load increases the heart's energetic costs to eject a given amount of blood (10, 24).

Effective arterial elastance (E_A) characterizes the net arterial load that is imposed on the heart (51). E_A was originally derived from a 3-element Windkessel model, which characterizes the arterial system in terms of resistive and pulsatile components (51). Operationally, E_A is defined as end systolic arterial pressure/ stroke volume index. A higher E_A reflects a greater arterial load imposed on the left ventricle (LV) and is directly related to heart rate (HR) and peripheral vascular resistance (which is determined, in large part, by the small arteries), and is inversely related to total arterial compliance (which is determined, in large part, by the central elastic arteries) (11). At rest, the resistive component is the dominant determinant of E_A (11). During exercise both arterial resistance and compliance usually decrease, and the relative contribution of the pulsatile component to E_A increases, so that by 80% of peak exercise the resistive and the pulsatile components provide nearly equal contributions to E_A(37). In various studies, E_A has been shown to increase (34, 37), decline (3) or remain unchanged (16) during exercise. The pattern of change in E_A may vary with age, sex and the intensity of exercise (37, 45). Such marked inter-individual differences in the change in E_A during exercise likely reflect the specific details of the population under study and varying combinations in the determinants of E_A (resistance, compliance and heart rate) among study subjects during exercise (21, 36). How the exercise-induced change in E_A relates to a change in LV filling and ejection has not been established, nor has its relationship with aerobic capacity, a clinically important marker of overall health and mortality (33) . We therefore evaluated whether the variability in the change in E_A during exercise is associated with a distinct pattern of change in arterial and cardiac function during exercise that may have important effects on exercise capacity.

METHODS

Study population.

The study population consisted of community dwelling volunteers with a broad age-range from the Baltimore Longitudinal Study of Aging (46), who underwent multi-gated blood pool scans at rest and during exercise. All subjects in the current investigation had a resting LV ejection fraction (EF) greater than 50% without wall motion abnormalities at rest or during exercise. All subjects were free of CV disease as determined by detailed history and physical examination, normal resting and treadmill exercise electrocardiograms, and absence of perfusion abnormality on thallium scintigraphy during treadmill stress testing in men over 40 years, and in women over 50 years of age. From baseline SBP values, 35% of the study population presented with hypertension based on the JNC 7 guidelines (14). However, no subject was taking any cardiac or antihypertensive medications. The study was approved by the institutional review boards of the National Institute on Aging and Johns Hopkins University. All subjects provided written informed consent to participate.

Evaluations.

All subjects underwent an upright seated bicycle graded exercise test. Pedal speed was maintained constant at 60 rpm, and workloads were increased by increments of 25W every 3 min until exhaustion. Maximum workload (MWL) was defined as the maximum wattage attained during the exercise test. Systolic and diastolic blood pressures (BP) were measured with cuff sphygmomanometry at seated upright rest, and during the final minute of each stage of exercise. End-systolic pressure (ESP) was noninvasively approximated by 0.9 (25) Despite the validation of the estimation of ESP as 0.9 × SBP (13, 25), some studies have suggested that the multiplier value is closer to 0.73 (39). When we repeated the analyses using ESP estimated as 0.73 × SBP the results were not affected; thus, only values derived from ESP estimated as 0.9 × SBP are shown. Noninvasive measurements of cardiac volumes at rest and during each stage of exercise were determined with multi-gated blood technetium 99m pool scans as previously described (21). This method has been validated both at rest (18) and during exercise (47). Adequate scaling of physiological measures for body size is essential for correct interpretation, and often the relationship between body size and physiological function may not be linear, a major assumption for the ratiometrically scaling approach. The allometric scaling approach accounts for this potential nonlinear relationship by normalizing physiological measures using exponential powers that linearize the relationship. Thus all cardiac volumes were ratiometrically or allometrically scaled to body surface area (5), yielding their respective volume indexes: end-diastolic volume index (EDVI), end-systolic volume index (ESVI), and stroke volume index (SVI). Of note, as the types of scaling for body size did not affect the findings, only the values of the ratiometrically scaled parameters are shown. Cardiac index (CI; l/min/m²) was calculated as SVI × HR. Peripheral vascular resistance index (PVRI; dyne.sec/cm⁵/m²) was calculated as mean arterial pressure (MAP; mmHg) × 80/CI, with MAP calculated as (1/3 × systolic BP + 2/3 × diastolic BP). To convert from Wood units to dyn·s·cm⁻⁵ MAP must be multiplied by 80. A crude estimate of total arterial compliance index (TACI) was calculated as SVI/pulse pressure (PP) (37). Ejection fraction (EF) was calculated as SVI/EDVI, and in the absence of Vo is inversely related to E_AI/E_LVI [E_AI/E_LVI = (1-EF)/EF]. Peak oxygen consumption (VO_2peak) was measured using the Medgraphics CPX-D system (Medgraphics, St Paul, Minnesota, USA). Measurement of VO_2peak was introduced in this study in 1986, and data on VO_2peak were only available on a subset (49%) of the cohort. Peak arterio-venous oxygen extraction was calculated from the Fick equation (VO_2peak/CI).Effective arterial elastance index (E_AI) was calculated as ESP/SVI (25), which was previously shown to closely approximate the arterial load directly measured invasively from aortic input impedance and arterial compliance data (25). LV end-systolic elastance, a measure of LV chamber performance was calculated as E_LVI = ESP/ESV-Vo (51) wherein V0, the volume-axis intercept of the end-systolic pressure volume relationship, was assumed to be zero, as previously reported (15). Stroke work index was used as an indirect measure of LV systolic performance (4), and was calculated as SWI = SVIxMAP. To adjust for the known effects of chamber size on E_LVI and SWI (5), EDVI was added as a covariate in the statistical models, and parameter estimates were used as the EDVI adjusted values for E_LVI and SWI. The interaction between the heart and arterial system was depicted by the arterial-ventricular coupling ratio (E_AI/E_LVI) = ESVI/SVI (9). Of note, when Vo is assumed to be negligible, E_AI/E_LVI is related to EF (15). The changes from rest to peak exercise in E_AI (ΔE_AI), and in other arterial and cardiac variables, were calculated by subtracting resting from peak values.

Statistical analysis.

All analyses were performed using the statistical packages SPSS version 13 (SPSS Inc, Chicago, Illinois). Grouped data are reported as mean ± SD unless otherwise stated.

ΔE_AI for the entire population exhibited a Gaussian distribution (Fig. 1A). One analytical approach was to compare the relationship of ΔE_AI in three groups stratified by the lowest, highest, and middle two quartiles to other arterial and cardiac variables. The descriptive characteristics of the three groups, stratified by ΔE_AI, were compared using analysis of variance (ANOVA) or chi-square. The arterial and cardiac parameters measured at rest, at peak exercise, and their change from rest to peak exercise were compared among the 3 groups by ANOVA after adjusting for age, sex and maximal workload, and by a repeated measures ANOVA across relative workloads of 25%, 50%, and 100%of maximum, with an E_AI group by workload interaction term.. The analyses were also adjusted for multiple comparisons using Bonferroni's method. We also examined the relationship between ΔE_AI and CV performance by BP group (normotensive vs. hypertensives) and the relationship between a hypertensive SBP response to the exercise stress test and CV performance. A hypertensive response to exercise was defined as an SBP>220 mmHg for men and >190 mmHg for women (26). A two-tailed P < 0.05 was required for significance. A second analytical approach was to examine the bivariate relationships among the changes from rest to peak exercise in the arterial and cardiac variables as continuous functions in the entire cohort via Pearson's correlation coefficients, and with linear regression analyses that were adjusted for age, sex and maximal workload. To evaluate the contributions of TACI, PVRI and HR to E_AI, multiple regression analysis was performed, and the standardized regression coefficient and the semi-partial correlations were calculated as an index of contribution to E_AI.

Fig. 1.

The distribution of the change in E_AI during exercise A: The frequency distribution of the change in effective arterial elastance index (ΔE_AI) from rest to peak exercise. B: The relationship of ΔE_AI to age in men (closed symbols) and women (open symbols).

RESULTS

The descriptive characteristics of the 3 groups in the resting state are shown in Table 1. The 3 E_AI groups did not differ significantly in age, sex, or body size.

Table 1.

Descriptive characteristics of the study population in the resting state

		EAI
	Total Cohort	Group 1	Group 2	Group 3
n	352	88	176	88
Age, years	53 ± 17	54 ± 19	51 ± 17	53 ± 15
Sex, % men	56	50	53	66
Height, cm	171 ± 10	170 ± 9	171 ± 10	172 ± 10
Weight, kg	74 ± 14	72 ± 15	73 ± 15	77 ± 13
Body mass index, kg/m2	25 ± 4	25 ± 4	25 ± 4	26 ± 4
Body surface area, m2	1.85 ± 0.21	1.83 ± 0.22	1.84 ± 0.21	1.89 ± 0.20
Maximal workload, watts	127 ± 42	118 ± 41	128 ± 40	134 ± 45*
Aerobic capacity, ml·kg−1·min−1	23.8 ± 7.2	23.6 ± 7.4	23.8 ± 7.1	24.1 ± 8.0

Values are means ± SD, or percentage. Effective arterial elastance index (E_AI) groups are defined according to quartiles of ΔE_AI: E_AI group 1, lowest quartile (ΔE_AI < 0.22 mmHg·m²⁻¹·ml⁻¹); E_AI group 2, middle two quartiles (0.22 ≤ ΔE_AI < 0.98 mmHg·m²⁻¹·ml⁻¹); and E_AI group 3, highest quartile (ΔE_AI ≥ 0.98 mmHg·m²⁻¹·ml⁻¹).

^*P < 0.05 compared with E_AI group 1 by ANOVA followed by Bonferroni correction.

ΔE_AI and its determinants during graded exercise.

During exercise, the observed changes in E_AI ranged from a 44% decline to a 149% increase (Fig. 1A). In ∼17% of individuals, E_AI was lower at peak exercise than at rest (i.e., ΔE_AI ≤ 0 mmHg.m²/ml). Conversely, in 19% of individuals, E_AI increased by 50% or more from baseline to peak exercise. The interaction term between age and sex was for ΔE_AI was nonsignificant (Fig. 1B). Furthermore, ΔE_AI was also not associated with body composition. The change in E_AI from rest to peak exercise (ΔE_AI) was stratified into three groups: E_AI group 1, included individuals whose ΔE_AI was in the lowest quartile i.e., ΔE_AI < 0.22 mmHg.m²/ml, (n = 88); E_AI group 2 included individuals whose ΔE_AI was in the middle two quartiles i.e., 0.22 ≤ ΔE_AI < 0.98 mmHg.m²/ml (n = 176); and E_AI group 3 included individuals whose ΔE_AI was in the highest quartile i.e., ΔE_AI ≥ 0.98 mmHg.m²/ml (n = 88). Fig. 2A shows that the 3 groups differed not only in their ΔE_AI, but dramatically differed in the pattern of change in E_AI as the intensity of exercise increased. Group 1, i.e., the group with the lowest ΔE_AI during exercise, had the highest E_AI at rest, associated with a higher resting HR, a lower TACI, and a nonsignificant trend for a higher PVRI (Fig. 2).

Fig. 2.

The change in E_AI and its components across relative exercise workloads. Comparisons of E_AI(A), heart rate (HR: B), peripheral vascular resistance (PVRI: C), total arterial compliance (TACI: D), and systolic blood pressure (SBP: E) among the 3 E_AI groups at rest, at 25% and 50% of peak, and at peak exercise (data shown as means ± SEM).*P < 0.05 E_AI group 1 vs. 2, ‡P < 0.05 E_AI group 1 vs. 3, ^†P < 0.05 E_AI group 2 vs. 3, adjusted for age, sex and maximal workload.

At rest, the components of E_AI, namely TACI (β= −0.25; semi-partial correlations = −0.04) PVRI (β = 0.86; semi-partial correlations = 0.44) and HR (β = 0.54; semi-partial correlations = 0.20) significantly (P < 0.01) accounted for 98% of the variance in E_AI. Similarly, at peak exercise TACI (β = −0.33; semi-partial correlations = −0.10), PVRI (β = 0.93; semi-partial correlations = 0.57), and HR (β = 0.52; semi-partial correlations = 0.18) significantly (P < 0.01) accounted for 90% of the variance in E_AI. When considered as a continuous variable across the entire cohort, ΔE_AI was positively associated with ΔHR (r = 0.12, P < 0.05), ΔPVRI(r = 0.41, P < 0.01), and inversely associated with ΔTACI(r = −0.46, P < 0.01). But the manner and magnitude in which these E_AI determinants (PVRI, TACI, and HR) change from rest to max effort differs according to the relative intensity of the exercise workload and among the three E_AI groups. HR changes uniformly during exercise in all three groups (Fig. 2B). Thus group differences in HR acceleration do not account for the group differences in E_AI during vigorous exercise. Changes in PVRI and TACI during exercise differed by E_AI group (Figures 2C & 2D) (significant E_AI group by workload interaction term) whereby E_AI group 1 had a blunted reduction in TACI, and a greater reduction in PVRI compared with E_AI group 3. Further, despite similar SBP values at rest, E_AI group 1 exhibited a smaller increase in SBP during exercise, and E_AI group 3 a greater SBP response to exercise such that, SBP at peak exercise was significantly higher in E_AI group 3 > groups 2 > group 1 (Fig. 2E).

It should be noted that the above comparisons among the 3 groups, and the associations with ΔE_AI persisted even after adjusting for age, sex and maximal workload (Fig. 2; Tables 2 and 3). Further, similar results were found when we removed the hypertensive individuals from the analyses (see online supplement). The only significant E_AI group by BP group (normotensive vs. hypertensive) interaction was found for ΔE_AI whereby hypertensive individuals in E_AI group 1 had a smaller ΔE_AI response, whereas hypertensive individuals in E_AI group 3 had a greater ΔE_AI response compared with E_AI group 3.

Table 2.

Comparison of resting cardiovascular parameters assessed among the three E_AI groups

	EAI
	Group 1	Group 2	Group 3
EAI, mmHg. m2/ml	2.72 ± 0.66	2.38 ± 0.53***	2.38 ± 0.54***
Blood pressure, mmHg
Systolic	129 ± 17.7	124 ± 16.1	125 ± 16.8
Diastolic	81 ± 11.1	78 ± 9.5*	79 ± 10.5
Pressure, mmHg
End systolic	116 ± 16	111 ± 15	113 ± 15
Mean arterial	97 ± 12	93 ± 11**	95 ± 11
Pulse	48 ± 14	46 ± 12	46 ± 13
Peripheral vascular resistance index, dyn·s/cm5/m2	2,486 ± 629	2,361 ± 592	2,394 ± 596
Total arterial compliance index, ml/m2/mmHg	1.01 ± 0.34	1.12 ± 0.37	1.13 ± 0.36*
Heart rate, beats/min	74 ± 11	69 ± 10***	68 ± 11***
Cardiac index, l/min/m2	3.2 ± 0.7	3.3 ± 0.7	3.3 ± 0.7
Volume index, ml/m2
Stroke	45 ± 9	48 ± 9**	49 ± 10**
End-diastolc	69 ± 14	74 ± 13*	74 ± 14*
End-systolic	25 ± 7	26 ± 7	25 ± 8
Ejection fraction, %	64 ± 7	65 ± 7	67 ± 7*
End-systolic elastance index, mmHg·m2/ml#	5.2 ± 1.1	4.7 ± 1.1**	5.1 ± 1.1†
Stroke work index, mmHg·ml/m2#	5,151 ± 875	5,471 ± 873*	5,449 ± 876
Arterial-ventricular coupling ratio	0.57 ± 0.18	0.54 ± 0.15	0.52 ± 0.17*

Values are means ± SD. Statistics are adjusted for age, sex, and maximal workload. However, similar statistical findings are present for the unadjusted comparisons.

^*P < 0.05,

^**P < 0.01,

^***P < 0.001, compared with E_AI group 1;

^†P < 0.05, compared with E_AI group 2 by ANOVA followed by Bonferroni correction;

^#end-diastolic volume index-adjusted values.

Table 3.

Comparisons of the change in cardiovascular parameters from rest to peak exercise assessed among the three E_AI groups

	EAI group 1
	Group 1	Group 2	Group 3
ΔArterial elastance index, mmHg. m2/ml	−0.14 ± 0.37	0.60 ± 0.20***	1.40 ± 0.49***†††
Blood pressure, mmHg
ΔSystolic	52 ± 26	73 ± 19***	88 ± 22***†††
ΔDiastolic	12 ± 16	12 ± 15	19 ± 17**††
Pressure, mmHg
ΔEnd systolic	47 ± 24	65 ± 17***	79 ± 20***†††
ΔMean arterial	25 ± 13	32 ± 13***	42 ± 14***†††
ΔPulse	41 ± 32	60 ± 23***	70 ± 25***
ΔPeripheral vascular resistance index (dyne.sec/cm5/m2)	−1459 ± 490	−1220 ± 478***	−949 ± 515***†††
ΔTotal arterial compliance index (ml/m2/mmHg)	−0.13 ± 0.51	−0.52 ± 0.30***	−0.67 ± 0.33***††
ΔHeart rate, beats/min	78 ± 23	84 ± 22	86 ± 24
ΔCardiac index (l/min/m2)	6.52 ± 1.91	5.99 ± 1.87***	4.76 ± 1.83***†††
ΔStroke volume index (ml/m2)	20 ± 8	13 ± 7***	4 ± 8***†††
Volume index, ml/ m2
ΔEnd-diastolc	14 ± 11	3 ± 9***	−4 ± 9***†††
ΔEnd-systolic	−7 ± 10	−10 ± 8	−8 ± 8
ΔEjection fraction (%)	15 ± 10	15 ± 8	10 ± 9***†††
ΔEnd-systolic elastance index (mmHg·m2/ml)#	14 ± 18	12 ± 16	7 ± 17*†
ΔStroke work index (mmHg·ml/m2)#	2942 ± 1433	3192 ± 1269	3215 ± 1372
ΔArterial-ventricular coupling ratio	−0.29 ± 0.21	−0.28 ± 0.17	−0.19 ± .018***†††

Values are means ± SD. Changes (Δ) in variables are calculated by subtracting basal from postexercise values. Statistics are adjusted for age, sex, and maximal workload. However, similar statistical findings are present for the unadjusted comparisons.

^*P < 0.05,

^**P < 0.01,

^***P < 0.001 compared with E_AI group 1;

^†P < 0.05,

^††P < 0.01.

^†††P < 0.001 compared with E_AI group 2 by ANOVA followed by Bonferroni correction.

^#EDVI adjusted values.

Cardiac volumes at rest and during exercise.

Resting EDVI and SVI were lower in group 1 vs. the other two groups whereas resting ESVI did not differ among the 3 groups (Fig. 3). The ΔESVI during exercise did not parallel ΔE_AI, whereas the change in EDVI and SVI across the relative workloads differed by E_AI group (Fig. 3)(significant E_AI group by workload interaction term). However, despite an initial increase in EDVI and SVI in all three groups E_AI group 3 did not utilize the Frank-Starling mechanism above 25% of peak effort and thus no further increase in SVI was noted. Thus the 3 groups exhibited marked differences from rest to peak exercise in ΔEDVI (20% in group 1, 4% group 2, −5% group 3) and in ΔSVI (44%, 27%, 8% respectively). An unexpected finding is that the relationship of EDVI to exercise intensity was an inverse mirror image of the relationship of ΔE_AI with exercise intensity (Fig. 2A). Indeed ΔE_AI, expressed as a continuous function of EDVI in the entire cohort, was significantly and inversely correlated with ΔEDVI (r = −0.58, P < 0.01). Of note, these relationships were independent of age, sex and maximal workload. In addition, similar results were found when we examined normotensive and hypertensive individuals separately (see online supplement).

Fig. 3.

The change in cardiac volumes across relative exercise workloads Comparisons of ESVI(A), EDVI(B), and SVI(C), among the 3 E_AI groups at rest, at 25% and 50% of peak, and at peak exercise (data shown as means ± SEM). *P < 0.05 vs. E_AI group 1 vs. 2, ‡P < 0.05 E_AI group 1 vs. 3, ^†P < 0.05 E_AI group 2 vs. 3, adjusted for age, sex and maximal workload.

Cardiac performance at rest and during exercise.

Resting CI did not differ among the 3 groups, but CI significantly higher in group 1 vs. group 3 at peak exercise (Fig. 4A). Further ΔCI as a continuous function among all subjects was inversely associated with ΔE_AI (r=−0.37, P < 0.05).

Fig. 4.

The change in cardiac performance across relative exercise workloads. Comparisons of cardiac index (CI: A), ejection fraction (EF: B), EDVI adjusted values of stroke work (SWI: C), and end-systolic elastance (E_LVI: D) among the 3 E_AI groups at rest, at 25% and 50% of peak, and at peak exercise (data shown as means ± SEM). *P < 0.05 vs. E_AI group 1 vs. 2, ‡P < 0.05 E_AI group 1 vs. 3, ^†P < 0.05 E_AI group 2 vs. 3, adjusted for age, sex and MWL.

At rest, group 3 had the highest EF; note, however, that at peak exercise group 3 had the lowest EF (Fig. 4B). ΔEF was inversely associated with ΔE_AI (r = −0.20, P < 0.05) and thus differed between groups. Because EF is influenced by the loading conditions of the ventricle, we also examined SWI and E_LVI, measures of LV pump performance; since both SWI and E_LVI are affected by chamber size, we therefore adjusted SWI and E_LVI for EDVI (a surrogate marker of chamber in size) (5, 40) in the ANCOVA (Fig. 4C and 4D). At rest, neither EDVI adjusted statistically for SWI nor E_LVI differed between E_AI groups. . Further SWI did not differ among groups during rest or at peak exercise. In contrast, E_AI group 3 demonstrated a blunted increase in E_LVI from 50% of peak to peak exercise compared with E_AI groups 1 and 2. A significant interaction between E_AI group and exercise workload for E_LVI was noted which further suggests that E_AI group 3 demonstrated a deficit in peak LV performance (i.e., limited Frank-Starling mechanism leading to reduced SVI and CI)(Fig. 4D). The above relationships persisted when we examined normotensives and hypertensives separately (see online supplementation).

In spite of a lower peak exercise CI in group 3 (Fig. 4A), peak aerobic capacity did not differ between E_AI groups because calculated peak arterial-venous oxygen extraction which was highest in E_AI group 3 (12.06 ± 3.5 ml/100ml), vs. group 1 (9.58 ± 2.4 ml/100ml), and vs. group 2 (10.02 ± 2.5 ml/100ml), and compensated for the lower peak CI. Further, the differences in arterial-venous oxygen difference between E_AI groups were not due to differences in hemoglobin concentrations (E_AI group 1 = 14.3 ± 1.3; E_AI group 2 = 14.3 ± 1.3; E_AI group 3 = 14.5 ± 1.3). These results were also observed when normotensives and hypertensives were analyzed separately (data not shown).

DISCUSSION

Effective arterial elastance is a measure of the net arterial load that is imposed on the LV and correlates well with aortic input impedance assessed invasively (25). In the current study, we found that the changes in E_AI during exercise in a large cohort of men and women of a broad age range encompassed a spectrum of values, with 17% of the cohort experiencing a decline in E_AI during exercise, and the rest showing varying increases up to 149%. Further the change in E_AI during exercise was accompanied by a distinct pattern of change in cardiovascular dynamics.

The arterial load has an important effect on LV pump function (20), whereby a higher resting E_AI increases the myocardial energetic costs to provide a given SVI (10, 24). Those which have examined E_AI during exercise in humans have reported that E_AI declines (3), remains the same (16), or increases (34, 37). The mechanisms behind the diverse change in E_AI with exercise, reflects the manner and magnitude of the changes in the components of E_AI, namely TACI, and PVRI, which vary within an individual and as a function of exercise intensity. Further, the variable changes in E_AI during exercise may also reflect subclinical arterial disease, such as insufficient vasodilatory response due to autonomic or endothelial dysfunction. Our study is the first to examine the exercise-induced changes in E_AI and the concomitant patterns of exercise induced-changes in ventricular volumes and function. We found a striking inverse relationship between ΔE_AI and the recruitment of EDVI, and the enhancement of SVI and CI with exercise, suggesting that individuals expressing a large increase in E_AI during exercise demonstrate a blunted utilization of the Frank-Starling mechanism, irrespective of age, sex, and body size.

Patterns of arterial alterations.

In this study ΔE_AI was not associated with age or sex, which differs from previous studies. Park et al. (38) found that older hypertensive women (59 ± 8 yrs) expressed a greater change in E_AI from rest to 75W compared with older (58 ± 9 yrs) hypertensive men. Further, Najjar et al. (34) found that during exercise E_AI was similar in young and older normotensive men, but was higher in older vs. younger normotensive women. These differences may be related to our inclusion of normotensive (65%) and hypertensives. Indeed when we separated our cohort into BP groups we find that ΔE_AI is associated with age and sex (data not shown). Further, Park et al. (38) examined the change in E_AI in the supine position and only reported data up to 75watts of exercise, which may further explain the differences between studies, as our figures show.

E_AI was originally derived from a 3-element Windkessel model which characterizes the hemodynamics of the arterial system as TACI, PVRI along with characteristic impedance (51). TACI provides information about the viscoelastic properties of the arteries. Several methods have been used to estimate TACI derived from the Windkessel model (28), and we direct readers to Westerhof et al. (53) for a detail review. The ratio of SV/PP (which is used in this study) is a simple index of TACI and has a physiological basis in the 2-element Windkessel model (12, 48–49). Theoretically, if we assume the periphery is closed, an increase in pressure (PP) resulting from a single stroke volume (SV), TACI relates to compliance as SV/PP. However, because part of the SV outflow leaves the arterial system through the microcirculation while cardiac ejection occurs the SV/PP ratio can overestimates TACI by as much as 60% (43). Thus, to reproduce the 2-element Windkessel model, only the part of blood ejected within the time of peak pressure should be measured. Despite this overestimation, close relationships have been noted between the SV/PP ratio and more direct measures of Windkessel compliance i.e., the area (r = 0.98 and r = 0.58) (12, 43), and the PP method (r = 0.97, and 0.80) (23, 43). Further, Resnick et al. (41) demonstrated that the SV/PP ratio was correlated with both capacitive (reflecting the TACI component associated with the conduit, capacitive function of larger arteries) and oscillatory (depicting the reflective component of TACI) compliance of the arterial tree using a modified Windkessel model (r = 0.92 and 0.68, respectively). However, despite this finding, the SV/PP ratio does not fully take into account the effects of characteristic impedance. It has therefore been suggested that SV/PP is an index of normal/abnormal compliance rather than a calculation or estimation of TACI (12).

Previous studies (11, 37, 42) have demonstrated that the resistive component is the dominant determinant of E_AI, at rest and that during exercise the contribution of the pulsatile component (TACI) to E_AI gradually increases. These results were confirmed in our study whereby SVRI accounted for most of the variance in E_AI at rest (44%; β = 0.86) and peak exercise (57%; β = 0.93). With the contribution of TACI to E_AI increasing from rest (4%; β = −0.25) to peak exercise (10%; β = −0.33). However, the contribution of HR to E_AI both at rest (20%; β = 0.54) and peak exercise (18%; β = 0.52) was similar. Furthermore, the variance in E_AI at rest (r² = 0.98) and peak exercise (r² = 0.90) were almost entirely accounted for by TACI, PVRI and HR. However, as with ΔE_AI, the manner and magnitude by which the components of ΔE_AI change during exercise differ among individuals and by the intensity of exercise and within a given individual, and therefore cannot be predicted a priori. This was highlighted in E_AI group 1 (minimal change in ΔE_AI) in which a blunted reduction in TACI, and the largest reduction in PVRI was noted. The larger reduction in PVRI during exercise in E_AI group 1, denotes an augmented peripheral vasodilatation possibly due to an enhanced endothelial reactivity or reduced sympathetic tone. Further, the ability to preserve TACI during exercise in this group, in spite of the exercise-induced increase in BP (and SVI) suggests an attenuated reduction in arterial distensibility. In E_AI group 3, the greater reduction in TACI than group 1 could reflect an increased vascular smooth muscle cell tone which would limit the vasodilatory response of central aorta and increase arterial impedance (32), and thus the greater BP response noted in E_AI group 3. However, all groups had a similar HR response to exercise, and therefore HR was unlikely to be the cause of the varied TACI response to exercise. In contrast, at rest HR was lower in E_AI group 3 than group 1, and therefore may have contributed to the difference in TACI at rest between groups (27). Further, in E_AI group 3, the excessive increase in MAP than group 1, a key determinant of arterial stiffness, may have contributed to the increased arterial load because at higher MAP the less compliant collagen fibers predominate in the maintenance of vessel-wall stresses (2). In contrast, at lower MAP (noted in E_AI group 1) the compliant elastin fibers are predominately recruited which may contribute to lower arterial stiffness and reduced arterial load.

Some of the components of ΔE_AI were themselves related to each other, whereby a greater preservation of TACI during exercise is associated with a greater reduction in PVRI (and a smaller increase in MAP and PP). This suggests that the tandem changes in resistance and compliance appear to be linked, and raises the possibility of a cross-talk between central and peripheral arteries. However, TACI is influenced, in part, by the compliance of the resistive vessels, and the resistive vessels have compliance properties. Further, central arterial compliance is also influenced by the timing and magnitude of the reflective waves, which are not fully accounted for by TACI (originally from the Windkessel model). Further, given that both parameters are derived noninvasively from pressure and flow, more direct and independent measurements of resistance and compliance are required to further explore this interesting theoretical interaction.

Patterns of ventricular alterations.

Another striking and novel finding of the present study is that the changes in E_AI during exercise were also coupled to correlated patterns of change in LV preload during exercise. Further, changes in LV preload were also associated with changes in PVRI and TACI suggesting that the cross-talk between the large and small arteries may extend to involve the heart. Indeed, Guyton postulated that cardiovascular performance is simply regulated by the rate of blood flow into the heart from the systemic circulation (22). In our study individuals with a larger ΔE_AI (i.e., those with a higher ΔTACI, and a smaller ΔPVRI) have a negative ΔEDVI and a smaller positive ΔSVI. This suggests that these individuals have a blunted reliance on the Frank-Starling mechanism to augment SVI during exercise. Further, the pattern of change in ΔEDVI mimicked the pattern of change in the TACI to PVRI relationship. Similarly, individuals with the largest EDVI at rest, and also at peak exercise, expressed the highest TACI/PVRI ratio. Individuals in E_AI group 3 were able to initially augment EDVI from rest to 25% of MWL, which may be attributed to the mobilization of blood in the extremities to the central circulation thereby increasing venous return.. However, at 50% of MWL the EDVI began to decline, corresponding to the higher absolute E_AI values, which may be attributed to an elevated PVRI, inadequate venous return (from the higher hydraulic load and higher peripheral resistance), increased LV diastolic stiffness (thereby impairing LV filling) (1), abnormal venous tone (thereby impairing endothelial function), or a combination of these factors. In line with our results, Borlaug at al. (7) found that blunted ΔPVRI was associated with attenuated increase of EDVI with exercise in heart failure patients with a preserved EF (HFpEF). Previous work (21, 50) had reported that the utilization of the Frank-Starling mechanism during exercise to simply ensure that SVI during exercise did not decline with age. In contrast, in our study, individuals with a smaller ΔE_AI who exhibited greater utilization of the Frank-Starling mechanism (E_AI group 1) during exercise, also achieved a higher SVI (and a higher CI) independent of age, sex or MWL. Thus, our findings provide novel insights in that they establish, for the first time, a link between ΔE_AI and the utilization of the Frank-Starling mechanism to increase SVI.

The deficit in cardiac performance (blunted preload and CI response) in E_AI group 3 did not translate into a deficit in peak aerobic capacity. Since VO_2peak is determined by the product of CI and arterial-venous oxygen extraction, the similar VO_2peak because arterial-venous oxygen extraction was increased in E_AI group 3 vs. 1, which was not due to differences in hemoglobin concentrations between groups. Thus, the arterial and cardiac alterations during exercise also appear to be linked to adaptations in peripheral oxygen extraction. The signaling pathways that underlie this adaptive response require further study.

Interestingly, Borlaug et al. (8) recently identified evidence of global impairment in cardiovascular reserve function in HFpEF patients including a blunted contractile, endothelial, and vascular (E_AI and PVRI) reserve. Further, it has been suggested that the depressed reserve responses correlated with a reduced exercise capacity in HFpEF patients (8). Unlike our study where the blunted increase in E_AI was accompanied by a greater reduction in PVRI, the HFpEF patients exhibited a blunted response in both E_AI and PVRI, which may suggest that they are no longer able to compensate for the limited E_AI response by reducing PVRI to maintain adequate LV filling. Indeed ΔEDVI was smaller in HFpEF patients compared with healthy controls (8).

Finally, the patterns that were observed during exercise are also linked to distinct (albeit more modest) arterial and ventricular patterns in the resting state. At rest, individuals E_AI group 1 exhibited higher E_AI, and a lower TACI, EDVI and SVI. However, these are the same individuals who, at peak exercise, have a lower E_AI, and a higher TACI, EDVI and SVI. This would suggest a more compliant system during exercise but not at rest in E_AI group 1 and vice versa in E_AI group 3. Acute shifts in compliance from rest to exercise are related to the arterial structure and vasoreactivity (e.g., sympathetic activity, prostacyclins or nitric oxide bioavailability), and mechanical forces (e.g., sheer circumferential stress, and HR), and how these change during exercise and interact with arterial structure determines the compliance of the artery during exercise. Unfortunately we are unable to determine the underlying mechanisms that alter the compliance of the artery from rest to peak exercise and contribute to the differences noted between groups.

The effects of hypertension and a hypertensive response to exercise.

We examined the effects of hypertension on the relationship between ΔE_AI and the CV performance by categorizing individuals into normotensives and hypertensives groups (14), and by categorizing individuals based on their BP response during the exercise stress (26). Individuals who normotensive and hypertensive at rest had similar arterial and cardiac responses to exercise. The only significant E_AI group by BP group interaction was found for ΔE_AI whereby hypertensive individuals in E_AI group 1 had a smaller ΔE_AI response, whereas hypertensive individuals in E_AI group 3 had a greater ΔE_AI response. These data illustrate that the findings of the study are not due the inclusion of individuals with hypertension.

Several studies (19, 31, 44) have shown that a BP response to exercise can predict the development of hypertension and target organ damage. Individuals with a hypertensive response to exercise had a greater ΔE_AI. With the exception of the change in E_LVI during exercise, no other hemodynamic differences were noted during exercise between BP response groups. However, differences were noted at rest and peak exercise; individuals with a hypertensive response to exercise had a higher E_AI, SVRI, and BP, and a lower TACI both at rest and at peak exercise compared with those with a normotensive BP response. This illustrates that examining the change in SBP alone during exercise does not fully reveal alterations in CV hemodynamic during exercise that are clearly evident when the ΔE_AI response is examined. Future research is required to determine whether the change in E_AI during exercise could predict future CV events.

Study limitations.

Some limitations of the study are worth noting. Because of the noninvasive nature of this study, some of the arterial and LV parameters were calculated using brachial rather than central pressures. Because of BP amplification across the arterial tree, brachial BP is known to overestimate central pressures (35); the magnitude of this overestimation is inversely related to age, and is markedly augmented during exercise (45). Although the estimation of ESP as 0.9 × brachial SBP has been shown to closely estimate central ESP (25), this was only assessed in the resting state. Because Vo (volume-axis intercept of the end-systolic pressure volume relationship) was not measured and assumed to be zero, E_LVI should not be considered an absolute measure of LV contractility but only an approximation of that. Further, arterial compliance was not directly assessed; instead the ratio of SVI/PP was used as a crude estimate of total system arterial compliance (17). This index, which has been shown to be a predictor of cardiovascular events (17) has well known limitations (17, 29). However, the ratio of SVI/PP is a practical and feasible measure to acquire during exercise, particularly since this study was largely conducted prior to the availability of modern devices that facilitate the assessment of arterial stiffness during exercise (45). Finally, because the calculation of PVRI and TACI were measured indirectly through changes in pressure and volumes, caution should be applied to the interpretation of any cross talk between large and small arteries. More direct and independent assessment is required to identify whether there is any cross talk between small and large arteries involved in the change in E_AI during exercise.

Conclusions

This is the first study to demonstrate the effects of graded exercise on the changes in E_AI and the corresponding changes in cardiovascular dynamics and to show the heterogeneity in E_AI among individuals across a range of exercise intensities. The changes in E_AI from rest to peak exercise reflect a complex interaction among changes in arterial and cardiac function that appear to be linked to adaptive features in the peripheral cardiovascular system. Because some of the components of ΔE_AI, at least at rest, have been shown to be predictors of outcomes, in future studies it would be worthwhile to examine whether the specific patterns of change in arterial load and the coupled patterns of change in LV response confer any clinical or prognostic information.

GRANTS

This work was supported by contract N01AG82109 (Gary Gerstenblith), and, in part, by the Intramural Research Program of the NIH, National Institute on Aging. VM is currently supported from the State Department of Health, Czech Republic (grant NS 10497-3/2009).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: P.D.C. analyzed data; P.D.C., V.M., S.P.S., G.G., L.C.B., L.F., J.L.F., E.G.L., and S.S.N. interpreted results of experiments; P.D.C. prepared figures; P.D.C., V.M., J.L.F., E.G.L., and S.S.N. drafted manuscript; P.D.C., V.M., S.P.S., G.G., L.C.B., L.F., J.L.F., E.G.L., and S.S.N. edited and revised manuscript; P.D.C., V.M., S.P.S., G.G., L.C.B., L.F., J.L.F., E.G.L., and S.S.N. approved final version of manuscript; J.L.F., E.G.L., and S.S.N. conception and design of research; J.L.F. performed experiments.

*Present address of J. L. Fleg: Division of Cardiovascular Diseases, National Heart, Lung, and Blood Institute, Bethesda, Maryland. ″ Present address of S. S. Najjar: Division of Cardiology, Washington Hospital Center.