Heart Failure: Insights From the Arterial Waves

The heart and the arterial system are the main components of the total cardiovascular system. They interact with each other (ventricular–arterial coupling), and the function from 1 of those 2 should only be analyzed in the context of the function from the other one. Thus, it is not surprising that severe systolic dysfunction of the heart, because it is present in patients suffering from heart failure with reduced ejection fraction (HFrEF), markedly influences the arterial pulse wave shape. The influence of HFrEF on measures of arterial wave shape and wave reflection, as well as the prognostic value of such parameters, is the focus of the study from Steinberg et al published in this issue of the Journal of the American Heart Association (JAHA).¹ We congratulate the authors for performing this valuable study and for the clear discussion of the results. First, it is shown that the augmentation index corrected for heart rate (AIx75) is different between patients and controls, whereas there was no significant difference in carotid–femoral pulse wave velocity (cfPWV). This confirms that AIx75, and thus the aortic pressure wave shape, is not solely influenced by arterial stiffness. Second, parameters obtained from pulse wave separation analysis showed a prognostic value within the cohort of patients with HFrEF after 2 years of follow‐up, whereas cfPWV was not significantly associated with the clinical outcomes. This indicates that waveform information, especially when obtained from both pressure and flow, has additional prognostic value beyond classical risk scores and beyond pulse wave velocity in patients with heart failure.

In addition to the comprehensive discussion of the results by the authors, we would like to add some thoughts in the following paragraphs.

Subjects with HFrEF are in a completely different cardiovascular equilibrium state. Because of ventricular–arterial coupling, this disease does not have an isolated effect on cardiac function, as commonly investigated by imaging such as echocardiography, but all arterial parameters are somehow affected. This makes matching of control groups difficult, because it is basically impossible to match for all parameters. If the groups are matched for pressure, as in the study by Steinberg et al, timing parameters such as heart rate or left ventricular ejection time will notably differ. A proper adjustment for effects would be needed for the correct interpretation of study results, but it is difficult. AIx75 is assumed to be corrected for heart rate; however, the formula applied to perform the correction is obtained from a small and different cohort of patients with pacemakers, which included only 2 patients with heart failure, and the relationship between augmentation index and heart rate may be different in patients with HFrEF and controls. When parameters from pulse wave analysis are corrected for heart rate and left ventricular ejection time, patients with HFrEF and controls may show similar values again for certain parameters, whereas only the correction could reveal the real difference between groups for other parameters.² It was also found by Steinberg et al that cfPWV is similar between groups. Although cfPWV is influenced by blood pressure, ventricular dysfunction probably does not have such a direct effect on cfPWV. Consequently, lower values of indices of wave reflection do not necessarily indicate reduced arterial stiffness in patients with HFrEF.

The timing of wave reflections in the aorta seems to be crucial for the effectiveness of ventricular–arterial coupling. This is comprehensively discussed by Steinberg et al, and it is also noted that current wave separation techniques may provide misleading parameters, because re‐reflections at the aortic valve are not explicitly included in wave separation methods.^{3, 4} Currently, it is also not fully understood how the progression of heart failure or valvular disorders is related to wave reflection at the valve and how the generated waveform by these re‐reflections depends on ventricular diseases.^{5, 6} Adaptation of wave separation techniques, including re‐reflections, could be worthwhile to increase the precision and interpretability of pulse wave parameters.⁷ Technically, the correct time information related to wave reflection assessment depends critically on the methods used, and measurements with high‐quality standards and analysis methods are needed to prevent misleading conclusions. This involves the appropriate sampling frequency as well as a proper alignment of pressure and flow traces in combination with a suitable technique to assess wave reflections.

As stated by Steinberg et al, the combined analysis of pressure and flow has a clear advantage to pressure‐only approaches, especially in patients with HFrEF. However, models to simulate measurement‐specific flow waveforms exist.^{8, 9} Their validity in subjects with impaired systolic function still needs to be determined, but promising results have been already achieved.^{10, 11} In real life (ie, in clinical studies but even more in clinical practice), there is always a tradeoff between accuracy and applicability. The measurement of flow in the aortic outflow tract and its alignment with aortic pressure traces is cumbersome. Probably for this reason it was not done by Steinberg et al for the control subjects in their study.

The wasted pressure effort and the wasted pressure effort ratio, which were used by Steinberg et al in their analysis, are parameters that have a similarity to parameters of reservoir theory, which are well known to the field and have been strongly discussed over the past decade.¹² It is noteworthy that the reservoir theory can be applied both including flow measures but also in a pressure‐only approach. Again, it still needs to be fully understood under which conditions the pressure‐only approach is a valid methodological simplification.

Unfortunately, subjects with HFrEF have high event rates; thus, outcome studies are possible with a relatively small number of subjects and a short follow‐up period. Steinberg et al decided to include only hard end points (death, heart transplantation, implantation of left ventricular assist device), which can be clearly traced. This decreases the noise introduced by weaker or intermediate end points often used in other studies, and in consequence this increases the statistical power. Nevertheless, weaker end points, such as a change in natriuretic peptides or exercise capacity or, typical for studies in heart failure, heart failure hospitalizations, also reflect important aspects of clinical practice. Fortunately, there is already evidence that parameters of wave reflection are helpful for therapy guidance and quantification of disease progression.¹³

When analyzed together with subjects with preserved ejection fraction, nonlinear risk models are needed because of the U‐shaped behavior of risk prediction for many typical biomarkers.^{14, 15} Even within an HFrEF cohort, nonlinear effects can occur. In previous works, we found such nonlinear behavior when we stratified patients with HFrEF into quartiles by arterial pressure indices.^{16, 17} These effects are probably attributable to different heart failure causes leading to a reduced ejection fraction. To investigate this further, we analyzed the data previously presented in Parragh et al.¹⁶ The amplitude of the reservoir pressure is used to stratify the cohort into quartiles. In the Figure, we show the percentage of patients in 3 groups (first quartile, second+third quartile, fourth quartile), broken down by cause. It can be seen that in the first quartile, mainly subjects with dilated cardiomyopathy are represented, whereas these patients are underrepresented in the fourth quartile.

Beyond traditional pulse wave parameters, which are based on signal analysis methods or physical/physiological models of the cardiovascular system, new methods from the field of machine learning and artificial intelligence could also be helpful to obtain information from pulse wave signals in the future. Machine learning can help to reduce variability introduced by measurement modalities and to select the most appropriate beats.¹⁸ Furthermore, it offers a wide range of possibilities for developing methods for the detection, prognosis, and therapy support of patients with cardiovascular disease.^{19, 20}

In conclusion, when analyzed properly, the heart and the arterial system are coupled, and systolic dysfunction is visible in arterial waveforms. Consequently, arterial pulse wave analysis might serve as a method for screening, risk prediction, as well as therapy guidance for patients with heart failure with reduced ejection fraction.