Novel Measures of Arterial Hemodynamics and Wave Reflections Associated With Clinical Outcomes in Patients With Heart Failure

Data are available upon reasonable request from the authors.

(1) Control cohort: Healthy, nonsmoking volunteers aged 30 to 70 years (n=129) from the Emory Predictive Health Initiative were recruited as control participants from 2006 to 2008 after careful screening for the absence of all of the following conditions: hypertension (systolic blood pressure <130 mm Hg or diastolic blood pressure <90 mm Hg on 3 occasions), hyperlipidemia (total cholesterol <200 mg/dL, low‐density lipoprotein <120 mg/dL), impaired fasting glucose or diabetes (fasting glucose <100 mg/dL), overweight or obesity (BMI ≤25 kg/m²), and history of any cardiovascular or valvular heart disease or use of prescription medications, as previously described.^{10, 11} (2) Patients with HFrEF: Self‐identified Black and White subjects aged ≥18 years (n=205) were screened for eligibility from the outpatient HF clinics at the Emory University Hospitals from 2015 to 2019. Patients were recruited using the following inclusion criteria: (1) ejection fraction (EF) ≤40% by echocardiogram attributable to ischemic or nonischemic etiology; and (2) New York Heart Association class II to IV HF symptoms for 3 months despite guideline‐directed medical therapy. Patients were excluded for the following reasons: (1) HF etiology including hypertrophic or restrictive cardiomyopathy, constrictive pericarditis, or complex congenital heart disease; (2) prior heart transplant or left ventricular assist device (LVAD); (3) any conditions other than HF likely to alter the patient's status over 6 months; (4) end‐stage HF requiring continuous inotrope infusion; and (5) serum creatinine >3 mg/dL or estimated glomerular filtration rate <20 mL/min per 1.73 m². After informed consent and enrollment, interviews and medical records were used to collect demographics, medical and procedure history, current medications, and laboratory measures. All study visits were performed in the Emory General Clinical Research Center. Both studies were approved by the Emory University Institutional Review Board.

Severity of illness was assessed using New York Heart Association functional class¹² and B‐type natriuretic peptide (BNP, Quest Diagnostics).^{13, 14}

Patients with HFrEF were actively followed for the occurrence of clinical events every 6 months after study enrollment. The primary clinical outcomes of interest were: (1) all‐cause death at 2 years or (2) a composite of death, heart transplant, or LV assist device implantation at 2 years. Patients without available follow‐up data were contacted by phone to ascertain events. For patients completely lost to follow‐up, death data were ascertained by Social Security Death Index query. Because controls were not expected to have clinical events, they were not followed longitudinally.

Large artery stiffness and wave reflections were estimated using the SphygmoCor device (Atcor Medical, Australia), according to previously published methods.¹¹ Patients with HFrEF were asked to hold vasoactive medications on the day of their study visit. All measurements were performed before venipuncture for biomarkers in a quiet, temperature‐controlled environment set at 22 °C after an overnight fast and were made with participants in the supine position after a 10‐minute rest period. Using applanation tonometry with a high‐fidelity micromanometer, CF‐PWV, the standard noninvasive reference metric of large artery stiffness, was measured by acquiring pressure waveforms at the carotid and femoral arteries, and velocity (distance/time in m/s) was calculated using the “foot‐to‐foot” method (Figure 1A). Pulse wave analysis (PWA) of the pressure waveforms at the radial artery estimated central (proximal) aortic pressures and the degree of systolic pressure augmentation. This permitted derivation of an augmentation index (AIx=augmented pressure/total central pulse pressure). Although AIx is affected by wave reflections, AIx is also sensitive to heart rate. Therefore, we used unadjusted AIx and AIx adjusted for heart rate in multivariable analysis. Patients with HFrEF with poor quality measurements attributable to poor signal (n=68) as assessed in fashion blinded to outcomes were excluded from this analysis.^{11, 15}

Pressure‐flow analyses were performed offline in participants with HFrEF using custom‐designed software written in MATLAB (The MathWorks, Natick, MA) as previously described.¹⁶ We performed 10‐s acquisitions of radial arterial pressure waveforms using high‐fidelity applanation tonometry. The radial waveform was calibrated using brachial systolic and diastolic pressure and used to estimate the aortic pressure waveform using the generalized transfer function of the Sphygmocor device (Naperville, IL). Central systolic blood pressure and pulse pressure were assessed. Pulse wave characteristics and morphology (first and second shoulders of systolic pressure and dicrotic notch) were used for determining the aortic augmentation index (P2−P1×100/pulse pressure). Flow measurements were performed using pulsed wave Doppler at the left ventricular outflow tract (Figure 1B).¹⁷ Pressure and flow waves were then aligned to maximize a linear rapid early systolic upstroke of pressure and flow and concordance of (1) first shoulder of systolic pressure with peak systolic flow and (2) aortic pressure dicrotic notch with the cessation of flow (Figure 1C).¹⁶ Notably, pressure and flow acquisitions were not simultaneous, which might have introduced noise in the measurements. Figure 2 illustrates the pressure‐flow relations and wave separation analysis. All pressure‐flow pairs underwent a quality control check by J.A.C. to ensure acceptable primary signal quality and good alignment. Aortic root characteristic impedance (Zc) was computed in the frequency domain as the mean value of higher harmonics of input impedance (Figure 2), as previously described.¹⁸ Zc was then used to calculate the forward (Pf) and backward (Pb) waves as Pf=(P+Q×Zc)/2 and Pb=(P−Q×Zc)/2, respectively, where P is pressure and Q is flow.¹⁸ Reflection magnitude (RM) was calculated as [Pb/Pf×100].⁸ Reflected wave transit time (RWTT) was calculated as the time delay between the zero crossings of Pf and Pb as previously described (Figure 2A).¹⁹ The time integral of the proximal aortic characteristic impedance (QZc) product (instantaneous flow multiplied by characteristic impedance, depicted in dark maroon in Figure 2B) was computed, which represents the pressure–time integral required to push the given stroke volume through the aortic root in the absence of wave reflections. Wasted pressure effort (WPE, depicted in red in Figure 2B) was computed as the difference between QZc and the observed systolic pressure–time integral (ie, the pressure–time integral before the incisura).⁴ It represents the additional effort required to overcome the load by wave reflections while generating the same flow/stroke volume. The wasted pressure effort ratio, a dimensionless ratio representing the ratio of pulsatile systolic pressure generated as a result of wave reflection versus load by the aortic root, was computed as WPE/QZc.

Data are presented as mean±SD, median (interquartile range [IQR]), or n (%) of patients. Baseline characteristics were compared between patients according to race and sex using Student t‐test for normally distributed continuous variables, nonparametric Mann–Whitney U test for non‐normally distributed continuous variables, and the Chi‐square or Fisher exact test for categorical variables. Given the differing physiology of patients with HFrEF compared with controls, ANCOVA was used to adjust CF‐PWV for age, heart rate, and mean arterial pressure, and AIx for age, heart rate, and mean arterial pressure. We similarly compared arterial stiffness and pulsatile hemodynamic measures between patients who experienced the primary end point and those who did not achieve the primary end point.

The association between measures of large artery stiffness and pulsatile load and the primary clinical end points was assessed using Cox proportional hazards models in patients with HFrEF only. All hemodynamic measures were transformed into z‐scores to obtain standardized hazard ratios, which are more intuitively comparable across different hemodynamic indices. All sample values had the mean value subtracted and were divided by 1 SD. Models were adjusted for race, BNP level, and the Meta‐Analysis Global Group in Chronic Heart Failure risk score. The Meta‐Analysis Global Group in Chronic Heart Failure risk score includes 13 clinical variables used to predict mortality in patients with HF, including age, sex, EF, systolic blood pressure, creatinine, New York Heart Association class, current smoking, diabetes, chronic obstructive pulmonary disease, duration of HF diagnosis, and guideline‐directed medical therapy use.²⁰ Power calculations were based on the expected association of traditional measures of arterial stiffness with the frequency of clinical HF‐related events (death, hospitalization) in patients with HF.²¹ Assuming a 2‐sided type 1 error of 0.05 and a test statistic based on the 2‐sample t‐test, we anticipated a sample size of n=200 would provide 80% power to detect a difference of 1.38 m/sec in femoral PWV (SD=3.1) and 6.37% in AIx (SD=14.3) between subjects with and without an HF‐related event.

Data were analyzed using SAS statistical software version 9.4 (SAS Institute Inc.) and R Studio version 1.3.1073 (The Comprehensive R Archive Network: https://cran.r‐project.org). A 2‐sided P value <0.05 was considered statistically significant.

Baseline characteristics of the 137 participants with HFrEF and 124 age‐matched controls are displayed in Table 1. Three‐quarters of participants with HFrEF had a nonischemic HF etiology, and almost 90% exhibited New York Heart Association functional class II or III. Approximately half of participants had a history of hypertension or diabetes. Measured blood pressures were consistent with clinical guideline targets, and use of guideline‐directed medical therapy was high.

Results of large artery stiffness and hemodynamic measures are shown in Table 2. Reproducibility studies in our laboratory on 10 subjects on consecutive days demonstrate a coefficient of variation of 8.1% and 6.4% for CF‐PWV and AIx, respectively. Central systolic, central diastolic, and peripheral diastolic blood pressures were higher in controls than patients with HFrEF. ANCOVA‐adjusted CF‐PWV and AIx were similar in controls compared with patients with HFrEF. Pressure‐flow analyses further characterized pulsatile arterial hemodynamics in patients with HFrEF. Median reflection magnitude was 47% (IQR, 39%–54%). In patients with HFrEF, median RWTT was 115 (IQR, 84–158) ms, whereas median ejection duration was 281 (IQR, 256–307) ms, implying substantial overlap between the reflected wave and systolic ejection. Accordingly, the wasted effort ratio (median, 0.81 [IQR, 0.34–1.29]) implied there was significant additional effort required to overcome the additional load from wave reflections added to the pulsatile load imposed by QZc (median, 0.09 [IQR, 0.06–0.14]).

Within a 2‐year follow‐up time in the HFrEF cohort, 27 (20%) patients died, 9 (7%) patients underwent LV assist device implantation, and none underwent heart transplant. Patients who experienced the primary end point had lower central systolic and diastolic pressure, lower central pulse pressure, lower Pf and Pb amplitudes (though comparable RMs), and shorter RWTT compared with patients who did not experience the primary end point (Table S1). On univariate analysis, shorter RWTT was associated with an increased risk of death at 2 years (hazard ratio [HR], 1.79 [95% CI, 1.12–2.78]), and this association remained significant in a multivariable model adjusted for race, Meta‐Analysis Global Group in Chronic Heart Failure score, and BNP (HR, 1.67 [95% CI, 1.03–2.70]) (Table 3, Figure 3A). Greater wasted effort ratio was significantly associated with increased risk of death at 2 years (HR, 1.44 [95% CI, 1.05–1.97]) on univariate analysis, though this association was attenuated in the multivariable model (HR, 1.37 [95% CI, 0.94–1.98]). Shorter RWTT and greater wasted effort ratio were both associated with an increased risk of the primary composite end point on univariate analysis (RWTT: HR, 1.54 [95% CI, 1.05–2.27]; wasted effort ratio: HR, 1.42 [95% CI, 1.07–1.88]) and remained significant in the multivariable model (RWTT: HR, 1.61 [95% CI, 1.06–2.44]; wasted effort ratio: HR, 1.40 [95% CI, 1.02–1.93]) (Table 3, Figure 3B).