Background. Patients with kidney disease suffer from high cardiovascular risk due to classic and disease-specific risk factors. Arterial stiffness is a novel cardiovascular risk factor whose role is yet to be established. High-resolution echo-tracking is a developing method for the assessment of local arterial stiffness.
Objectives. To assess carotid stiffness in patients on long-term hemodialysis (HD) using high-resolution echo-tracking and to analyze the impact of arterial stiffness on mortality in the mid-term follow-up.
Materials and methods. Fifty-eight HD patients (28 female (F), 30 male (M)) underwent clinical examination, laboratory tests and carotid stiffness assessment. Local arterial stiffness parameters such as beta stiffness index (β), Young’s modulus (Ep), arterial compliance (AC), and one-point pulse wave velocity (PWVβ) were measured both before and after HD, allowing to calculate their change (∆). The survival of patients was analyzed up to 48 months. The multivariate analysis of survival with the use of Cox proportional hazard stepwise regression was performed to determine the factors significantly correlated with the survival.
Results. After 48 months, 33 patients were alive (16 F, 17 M) and 25 patients (12 F, 13 M) died. The deceased group was significantly older (66.5 ±12.3 years compared to 56.6 ±17.8 years), had more pronounced coronary artery disease (percutaneous coronary intervention (PCI) 36% compared to 9%, p < 0.05, respectively). Deceased patients had significantly higher ∆AC than survivors. The results showed that age, history of PCI, left ventricular ejection fraction (LVEF), ∆AC, fasting glucose, serum total protein, sodium level after HD, and potassium level before HD were significantly associated with mortality.
Conclusions. Echo-tracking-based arterial stiffness assessment in patients with chronic kidney disease (CKD) yields the clinical information regarding mid-term mortality risk. A paradoxical increase in AC is among independent risk factors for mid-term mortality in patients undergoing maintenance HD. The proper estimation of the correlations among vascular, hemodynamic and sympathetic-dependent changes in a given patient with kidney failure is complex.
Key words: hemodialysis, chronic kidney disease, arterial stiffness, mortality, echo-tracking
Cardiovascular diseases are the leading cause of death in patients with chronic kidney disease (CKD).1 These patients suffer from high cardiovascular risk due to traditional and disease-specific risk factors. Arterial stiffness has been proven to be independently associated with a higher global risk of death in patients suffering from CKD.2 The measurement of local arterial stiffness in CKD patients may be of particular value due to the limitations of classic diagnostic methods in this group of patients,3, 4 such as electrocardiographic exercise test or stress echocardiography.5 These patients are less likely to have the classic symptoms of myocardial ischemia, hence the need to look for new tools to improve clinical evaluation. Damage to the kidneys also limits the possibility of using tests with contrast media – coronary angiography and computed tomography.3 In this group of patients, the diagnostic value of laboratory exponents of cardiovascular function, such as N-terminal pro B-type natriuretic peptide (NT-proBNP) or cardiac troponin, is also lower.6, 7 Chronic kidney disease is affecting 10–16% of the world population.8, 9, 10 As glomerular filtration rate (GFR) decreases, the likelihood of hypertension increases.11 People with CKD are also characterized by an increased tendency to develop advanced atherosclerotic lesions.12 It should be emphasized that the presence of classic cardiovascular risk factors such as diabetes, hypertension, nicotinism, and dyslipidemia does not sufficiently explain the clear increase in the risk of cardiovascular death in patients with CKD.4, 10
Stiffness is one of the properties of arteries resulting from the vessel wall structure, in particular from the ratio of collagen to elastic fibers.13, 14 Left ventricular contraction causes an ejection of a certain volume of blood into the systemic circulation. The appearance of additional blood volume in the ascending aorta causes its distension, which is possible due to its high elasticity. This distension is then transferred to the distal parts of the arterial system, forming a pulse wave. The pulse wave, reaching the resistance vessels, is reflected due to the increase in the stiffness of arteries. The reflected wave returns to the ascending aorta in diastole, supporting coronary perfusion. The final shape of the pulse wave is therefore a result of a progressive wave and reflected wave. The increase in the stiffness of arteries, particularly in the aorta, leads to a shorter distance between the heart and the place where the reflected wave is formed, as well as to an increase in pulse wave speed. This causes many adverse effects on the cardiovascular system. The resulting reduction in diastolic blood pressure impairs coronary perfusion. At the same time, high central blood pressure leads to an increase in left ventricular afterload in the pressure overload mechanism. Therefore, an increase in arterial stiffness is a part of the development of hypertension, ischemic heart disease and heart failure.15, 16 Both aging and diseases such as essential hypertension, diabetes and CKD contribute to the increase in arterial stiffness.17, 18, 19
High-resolution echo-tracking is a direct and noninvasive method of the evaluation of local arterial stiffness. It is obtained by ultrasound examination at one point of the arterial system, usually at the site of the common carotid artery. According to experts from the European Society of Cardiology (ESC), it is recognized that methods for measuring local arterial stiffness, such as high-resolution echo-tracking, are useful in pathophysiology and therapy studies.13 In recent years, the interest in using high-resolution echo-tracking in patients with CKD has increased.20, 21, 22 There is growing evidence that carotid stiffness assessment may contribute to a better risk stratification of CKD patients. Studies have shown that carotid stiffness parameters obtained with high-resolution echo-tracking may be used in prediction of all-cause mortality in patients with kidney failure treated with hemodialysis (HD).21
The aim of the study was to assess local arterial stiffness parameters in patients undergoing maintenance HD and relate these parameters to mortality in the mid-term follow-up.
Materials and methods
Study design and participants
This was a prospective cohort study. The analysis was performed in 58 patients (28 women and 30 men) with kidney failure treated with HD at Dialysis Center in Department of Nephrology and Transplantation Medicine (Wroclaw Medical University, Wrocław, Poland). Hemodialysis sessions were conducted 3 times a week for 4 h per session. The study group was formed in 2015 and patients were observed until 2019. The patients included in the study were free of active infection. We excluded patients with persistent or permanent atrial fibrillation, history of malignancy and diseases requiring immunosuppressive treatment.
All patients were treated with erythropoietin and intravenous iron supplementation according to standards. Beta-blockers, angiotensin converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARBs), calcium channel blockers, and alpha blockers were used in the treatment of hypertension. All patients were dialyzed using a native arteriovenous fistula.
The study was approved by Institutional Ethics Committee of the Wroclaw Medical University, Poland.
Variables and data sources
In all subjects, data on cardiovascular morbidity and causes of kidney failure were collected. Clinical data of patients were extracted from the hospital records.
At the study onset, the following factors were analyzed: baseline characteristics, duration of dialysis, adequacy of dialysis (Kt/V), body mass index (BMI), and laboratory parameters. Blood samples were collected prior to the initiation of HD session. Routine laboratory tests were measured in the Central Hospital Laboratory, University Hospital in Wrocław, Poland, as a part of the standard care.
Heart rate, blood pressure measurements and carotid elasticity with echo-tracking technique were investigated at the study onset, before and after a single mid-week HD session. All patients underwent transthoracic echocardiogram.
Assessment of local arterial stiffness
Images were obtained with an Aloka Alpha 6 ultrasonograph (Aloka Co., Ltd., Tokyo, Japan) equipped with an integrated and automated Doppler and high-resolution echo-tracking system using a linear probe. Patients were examined in supine position 15 min before the start of HD and 15 min after the end of HD. A clear ultrasonographic visualization of both anterior and posterior wall of the common carotid artery opposite to the arteriovenous fistula was taken 1–2 cm below bifurcation in the longitudinal axis. Consequently, after establishing the intima-media complex, the echo-tracking samples were positioned at the end of intima, with an 1 kHz sampling rate for continuous detection of movement of the arterial wall. Thus, a graphical representation of change in the diameter of the artery was recorded as a waveform. Three to five evolutions were recorded to obtain a representative waveform.
To calculate local arterial stiffness parameters, it is necessary to register high-resolution echo-tracking, heart cycle and blood pressure. Heart cycle was recorded using standard electrocardiographic limb leads I, II and III. Blood pressure was measured over the brachial artery opposite the arteriovenous fistula in supine position, directly before the ultrasonographic examination.
As a result, the following parameters were calculated:
– β – beta stiffness index – ratio of the natural algorithm of systolic/diastolic blood pressure to the relative change in diameter:
β = ln (Ps/Pd)/[(Ds − Dd)/Dd]
where: ln – the natural logarithm, Ps – systolic blood pressure, Pd – diastolic blood pressure, Ds – diameter of the artery in systole, Dd – diameter of the artery in diastole;
– Ep – epsilon, Young’s modulus:
Ep = (Ps − Pd)/[(Ds − Dd)/Dd]
– AC – arterial compliance, calculated from the arterial cross-section and blood pressure:
AC = π(Ds×Ds − Dd×Dd)/[4×(Ps − Pd)]
– PWVβ – one-point pulse wave velocity, derived from the time delay between the 2 consecutive waveforms representing the diastole of the artery, with the use of beta stiffness index:
PWVβ = √(β×Pd/2×r)
where: β – beta stiffness index, Pd – diastolic blood pressure, r – blood density (1.050 kg/m3).
The change (∆) in carotid stiffness parameters was calculated as follows: ∆β = β after HD – β before HD; ∆Ep = Ep after HD – Ep before HD; ∆AC = AC after HD – AC before HD; ∆PWVβ = PWVβ after HD – PWVβ before HD.
Statistical analyses were performed using STATISTICA v. 12 (TIBCO Software, Palo Alto, USA). Continuous variables were checked for the distribution using the Shapiro–Wilk W test and if the value of p was < 0.05, the assumption for normality was discarded. The Shapiro–Wilk W test showed normal distribution for the following variables: age, adequacy of HD, ultrafiltration during HD session, BMI, all echocardiographic parameters, echo tracking parameters except for Ep after HD, all blood pressure and heart rate parameters, and laboratory parameters except for fasting glucose, C-reactive protein (CRP) and high-density lipoprotein (HDL) cholesterol. Continuous variables with normal distribution were presented as means and standard deviations (SDs), and compared according to the survival groups (deceased compared to alive) with Student’s t-test. Continuous variables without normal distribution were presented as medians and interquartile ranges (IQRs) and compared using Mann–Whitney U test. Discrete variables were presented as counts and percentages, and compared using χ2 test. The comparison of the investigated parameters between the deceased patients depending on the causes of death (non-cardiovascular compared to cardiovascular mortality) was performed separately.
The multivariate survival analysis with the use of Cox proportional hazard stepwise regression was performed based on a model in which the dependent variable was the survival time since the inclusion in the study, while independent variables incorporated in the analysis were characteristics which differed in the one-way analysis with a significance of p < 0.15 between the 2 evaluated groups (survivors and non-survivors), or were relevant from the clinical point of view. In case of highly correlated variables such as heart rate before HD and heart rate after HD, serum total protein and serum total albumin, history of coronary artery disease and history of percutaneous coronary intervention (PCI), AC after HD and change in the arterial compliance (∆AC), the more statistically significant variable was chosen. The following variables were considered for multivariate analysis: age, hypertension, history of PCI, left ventricular ejection fraction (LVEF), ∆AC, heart rate after HD, hemoglobin, serum total protein, fasting glucose, CRP, low-density lipoprotein (LDL) cholesterol, urea after HD, creatinine before HD, sodium level after HD, and potassium level before HD.
The survival analysis using Cox proportional hazard regression analysis was performed with the Cox proportional hazard analysis module. In the assumptions of the model, the reliability of Efron was selected, and models were created taking into account all effects and using the forward method. Then, Cox regression assumptions were assessed by analyzing Martingale-based residuals for survival models and Schoenfeld residuals over time, visually assessing a plot of the residuals compared to time. The correlation of residuals and time was examined using the Spearman’s correlation. In case of finding a variable that violates the assumptions of the Cox proportional hazard regression, the variable was excluded from the model.
Receive operating characteristic (ROC) analysis was performed to find the cutoff point of echo-tracking parameters which differ between survivors and non-survivors. The Kaplan–Meier curve was constructed and the log-rank test was performed to present the survival in the groups of patients chosen on the basis of the founded cutoff points, as described above.
The multiple stepwise regression analysis was performed to find the association between echo-tracking parameters and age, gender, diabetes, and the presence of cardiovascular complications assessed as at least one of the following: previous myocardial infarction, stroke, history of coronary artery bypass graft surgery, or PCI. The variance inflation factor was assessed as the diagonal elements of the inverse correlation matrix. The Durbin–Watson test was performed to examine the autocorrelation of the residuals.
The results were considered statistically significant when the p-value was < 0.05.
Clinical characteristics of the investigated patients at study onset are presented in Table 1.
Thirty-three patients from the initial cohort of 58 (56.89%) survived the whole 48-month observation period. Of the non-survivors, 11 patients (44%) died of cardiovascular diseases and 14 patients (56%) died of non-cardiovascular causes. The comparison of the investigated parameters between deceased and surviving patients is depicted in Table 2, Table 3, Table 4. The comparison of the investigated parameters between deceased patients depending on the causes (non-cardiovascular compared to cardiovascular mortality) is depicted in Table 5, Table 6, Table 7. The predictors of the overall mortality among the studied patients using Cox proportional hazards model are depicted in Table 8.
Based on the data obtained in the study, we created a multivariate analysis of survival with the use of Cox proportional hazard stepwise regression. The dependent variable was the survival of the whole 48-month observation period. Independent variables were: age, hypertension, history of PCI, LVEF, ∆AC, heart rate after HD, hemoglobin, fasting glucose, serum total protein, CRP, LDL cholesterol, urea after HD, creatinine before HD, sodium level after HD, and potassium level before HD. The stepwise forward model showed that factors which significantly correlated with survival were: age, history of PCI, LVEF, ∆AC, fasting glucose, serum total protein, sodium level after HD, and potassium level before HD.
In Figure 1, a plot of Martingale-based residuals compared to survival time was depicted. In Table 9, the Spearman’s correlation between Schoenfeld residuals of the model and survival time was shown.
The ROC curve analysis was performed to find the cutoff point between patients who survived the 48-month observation period and the deceased. In the Figure 2, the ROC curve analysis for the ∆AC in survivors and non-survivors was presented. The cutoff point aimed at differentiating the groups was −0.06 mm2/kPa. The area under curve was 0.71; 95% confidence interval (95% CI): [0.56; 0.85]; p = 0.005.
In Figure 3, the crude Kaplan–Meier survival curves for the difference in survival regarding ∆AC for patients with ∆AC > −0.06 and ∆AC < −0.06 are presented. Patients with ∆AC > −0.06 have worse outcome than patients with ∆AC < −0.06. We assessed the significant cutoff point of survival as −0.06 mm2/kPa.
The stepwise multivariate regression analysis revealed that age was only associated with β after HD, with Ep after HD and with PWVβ after HD. However, gender, diabetes and the presence of cardiovascular complications were not related to any parameters of arterial stiffness. The results of the stepwise multiple regression analyses were presented in Table 10, Table 11, Table 12.
In the present study, we observed that the change in AC correlates with mid-term survival in patients undergoing maintenance HD. The AC is a parameter proportional to change in cross-sectional area of the artery, and inversely proportional to the change in blood pressure. In general population, the higher the AC, the more elastic the artery is. Previous studies have shown that the decrease in pulse pressure – which is a basic surrogate of arterial stiffness – after HD is connected with a longer survival of patients undergoing maintenance HD.23 In contrary, the results of our study show that subjects with ∆AC equal or higher than −0.06 mm2/kPa had shorter survival time in the mid-term follow-up than patients with ∆AC below the cutoff point. The Kaplan–Meier survival curves drifted apart after 600 days. Furthermore, the lone value of AC after HD was also correlated with worse outcome – survivors were characterized by a significantly lower AC after HD than deceased patients. We have not observed such correlation in AC before HD or for any other of the calculated carotid stiffness parameters. We hypothesize that in the group of patients with kidney failure, the pathophysiological process and degeneration of the arterial wall are so advanced that in most severe cases, the arteries lose ability to reduce inner diameter after dehydration during HD. This particular phenomenon may predict worse outcome.
There is some evidence stating that among patients with kidney failure treated with HD, a U-curve relationship between change in arterial stiffness and survival may occur. In one study, Lertdumrongluk et al. found that modest decline in pulse pressure, rather than either large reduction or rise in pulse pressure, is associated with the greatest survival.24 This result may correspond with our findings, leading to a concept in which a potential significant decrease in arterial stiffness would correlate with a worse outcome.
In our study, we used high-resolution echo-tracking as a source of stiffness data. This method has already been used in studies on arterial stiffness in patients undergoing maintenance HD; however, only several of them focused on the role of carotid stiffness as a predictor of mortality in this group. Sato et al. found a correlation between high β and a higher risk of all-cause mortality in a 4-year follow up.21 Another study on β has shown a predictive value of carotid stiffness on cardiovascular mortality independent of arterial thickness.22 Blacher et al. investigated a potential role of the common carotid artery incremental modulus of elasticity (Einc), showing that Einc is an independent predictor of both cardiovascular and all-cause mortality in patients with kidney failure treated with HD.25 In contrary to these studies, we explored a wider spectrum of carotid stiffness parameters directly before and after HD. To the best of our knowledge, there are no studies exploring the potential predictive value of change in these particular arterial stiffness parameters during a single HD session.
Apart from the findings considering arterial stiffness, our study has shown the association between the cardiological burden and survival of patients with kidney failure. Deceased patients had lower LVEF and more often had a history of PCI, which are the results of heart failure and coronary artery disease (CAD), respectively. Also, the deceased patients were characterized by a lower serum potassium level, which may be related to a higher risk of tachyarrhythmias.
The assessment of hemodynamic changes in patients with kidney failure is not a simple task. It should be taken into account, that changes in stiffness parameters during HD may indicate both pathological changes in the vessels themselves and unfavorable processes in the autonomic system. Some blood pressure changes are provoked by higher fluid or sodium removal. On the other hand, cardiovascular changes can be disturbed in the state of raised sympathetic tone present in CKD patients.26 Sympathetic hyperactivity is taken into account as one of the reasons of intradialytic blood pressure changes.27 The hydration status may also be considered a factor impacting arterial stiffness. In a study on a group of HD patients, hypervolemia (assessed as left ventricle end-diastolic volume) was significantly associated with pulse pressure.28 Moreover, it cannot be ruled out that the increase in AC after HD is a result of decrease in systolic blood pressure following dehydration. Hemodialysis may impact local arterial stiffness in a complex mechanism, in which the decrease of blood pressure, dehydration and individual properties of the cardiovascular system overlap and lead to the change in AC. This reaction of cardiovascular system is correlated with the survival.
The correlation with mortality of other echo-tracking parameters such as β, Ep and PWVβ remained statistically insignificant both before and after the HD, which stays in opposition to the previous findings.21, 22 Because of the existing shortage in the studies on the abovementioned parameters in kidney failure patients, with regard to all-cause mortality, and in particular cardiovascular mortality, the discussion of those relationships would be complex or probabilistic one.
Our study shows that high-resolution echo-tracking may provide a unique insight into pathophysiology of cardiovascular disease in patients with kidney failure, and particularly, in patients treated with maintenance HD. It also gives hope for a better risk stratification among those patients. Further pathophysiological studies will be needed in order to recognize the underlying cause of paradoxical change in AC and a potential utility of this phenomenon. Moreover, further studies are required to establish how to control arterial stiffness in order to improve the outcomes.
Our study has several limitations. It was observational in nature and thus, the causality could not be directly derived from the results. It had also a relatively small number of participants. The mid-term follow-up of 48 months seems to be also a limitation. Nevertheless, the subject did not undergo extensive investigations, so our study adds an important piece of data regarding the studied issue.
Echo-tracking-based arterial stiffness assessment in patients with CKD yields the clinical information regarding mid-term mortality risk. The independent risk factors for mid-term mortality in patients with kidney failure treated with HD are age, PCI, LVEF, ∆AC, fasting glucose, serum total protein, sodium level after HD, and potassium level before HD. The proper estimation of the correlations among vascular, hemodynamic and sympathetic-dependent changes in a given patient with kidney failure is complex.