The current study was performed in accordance with the Helsinki declaration as revised in 2013. The University of Witwatersrand Human (Medical) Research Committee approved the protocol (protocol number: M15-08-43). Each patient provided informed written consent prior to participation. The study design was reported previously (28–30). One hundred and fifteen patients participated. These comprised 67 non-dialysis and 48 dialysis patients. Patients with active infection or/and cancer, previously diagnosed heart failure and a Chronic Kidney Disease Epidemiology Collaboration estimated glomerular filtration rate (eGFR) of ≥60 ml/min/1.73 m² were excluded.

Methods that were employed in this study have been reported previously (28–30). Recorded baseline characteristics comprised demographic features, anthropometric measures, major traditional cardiovascular risk factors, non-traditional or renal cardiovascular risk factors and the use of cardiovascular drugs and erythropoietin stimulating agents. Established cardiovascular disease comprised ischemic heart disease (acute myocardial infarction, percutaneous transluminal coronary angioplasty and coronary artery bypass graft), cerebrovascular disease (stroke and transient ischemic attack) and peripheral vascular disease, the presence of which was confirmed by a cardiologist, neurologist and vascular surgeon, respectively. All investigations were performed on a single day. In dialysis patients, this was done on a day prior to a haemodialysis session. Dialysis was performed thrice weekly using 4-hour sessions. Most dialysis patients (72.9%) had an arteriovenous fistula (68.7%) or arteriovenous graft (4.2%). All patients were in sinus rhythm at the time of investigation.

Mean arterial pressure for the peripheral waveform was determined electronically by the SphygmoCor device (see below) and using the formulaMP=∑i=T0TFPinwhere T₀= start of the waveform; T_F= end of the waveform; P_i= pressure points and n = number of pressure points.

Central systolic blood pressure and forward wave pressure were determined using a high-fidelity SPC-301 micromanometer (Miller instrument, Inc., Houston, Texas), interfaced with a computer using SpygmoCor software, version 9.0 (AtCor Medical Pty. Ltd., West Ryde, New South Wales, Australia), as previously reported (31, 32). After resting for 15 min in the supine position, arterial waveforms at the radial (dominant arm), carotid and femoral artery were recorded for a time period of ten consecutive waveforms (heart beats). Calibration of the pulse wave was done by manual measurement (auscultation) of the brachial blood pressure taken immediately prior to recordings. A validated generalized transfer function incorporated in the SphygmoCor software was used to convert the peripheral pressure waveform into a central aortic waveform. The results were discarded when systolic and diastolic variability of consecutive waveforms exceeded 5% or the amplitude of the pulse wave signal was less than 80 mV. All measurements were made by single experienced observer (CR) who was unaware of the cardiovascular risk factor profiles of the patients. Technically sound measurements of the central pressure wave were obtained in 109 patients. Brachial or peripheral blood pressure were recorded in all patients employing the oscillometric SunTech device (SunTech Medical, USA) (33). Recorded peripheral blood pressure in this study represents the average of ≥3 measurements taken at least 30 s apart after sitting quietly for at least five minutes. The optimal systolic blood pressure target in CKD is currently uncertain (34). In the present study, a systolic blood pressure of ≥130 mmHg was considered elevated.

Echocardiography was performed in accordance with the American Society of Echocardiography convention (35) and employing a Philips CX50 POC Compact CompactXtreme Ultrasound System [Philips Medical Systems (Pty) Ltd, USA] equipped with a 1.8–4.2 MHz probe that allowed for M-mode, 2-D and tissue Doppler measurements, as previously described (30). Patients were examined in the partial left decubitus position. We assessed left ventricular geometry and systolic (lateral s' and midwall fractional shortening as measures of longitudinal and circumferential myocardial contractility, respectively, and ejection fraction as an index of chamber or pump function) function.

Left ventricular dimensions were determined by measuring the left ventricular internal end diastolic and end systolic diameters and wall thickness (left ventricular septal and posterior wall thickness) in the parasternal long axis view by two-dimensional directed M-mode echocardiography. The Teichholz method was used to assess left ventricular end diastolic volume. Stroke volume was determined from the difference between left ventricular end diastolic and systolic volumes as evaluated upon employing the Z-derived method. Cardiac output was determined as stroke volume × heart rate. Left ventricular midwall fractional shortening was assessed using the previously reported formula as [(LVID_ed + 0.5 H_ed) − LVID_es + 0.5 H_es)]/LVID_ed + 0.5 H_ed) where LVID is left ventricular internal diameter, H is wall thickness (mean of septal + posterior wall thickness), ed is end diastole and es is end systole (36, 37). Left ventricular ejection fraction was calculated as [(left ventricular end diastolic volume − left ventricular end systolic volume)/left ventricular end diastolic volume] × 100. Systemic vascular resistance was calculated as (mean arterial pressure − right atrial pressure)/cardiac output, assuming the right arterial pressure is 0 mmHg. Heart rate was determined from the length of an averaged peripheral waveform captured over a 10 s period, using the formula: 1,000/the length of an averaged peripheral waveform captured over a 10 s period ×60. Total arterial compliance (TAC) was calculated as stroke volume/central or aortic pulse pressure.

Echocardiographic measurements were made by the same observer that performed the arterial function evaluation. Intra-observer echocardiographic measurement variability is low in our setting with Pearson's correlation coefficients and variances (mean % difference (SD)) for left ventricular end-diastolic diameter, septal wall thickness, posterior wall thickness, E and e' of 0.92, 0.72, 0.76, 0.88 and 0.93 (p < 0.0001 for all), and −0.41 (4.16), 0.45 (7.74), 1.74 (6.08), 0.16 (9.95) and −1.46 (8.58), respectively.

Results are expressed as mean (SD), median (interquartile range) or proportions as appropriate. Non-normally distributed variables were logarithmically transformed prior to entering them in linear multivariable regression models.

Data among subgroups including non-dialysis vs. dialysis patients and those with controlled vs. uncontrolled systolic blood pressure were compared in age, sex and black population origin adjusted models. Potential confounders for subsequent multivariate analysis were also identified in age, sex and black population origin adjusted models. Black population origin was included in regression models because the mean (SD) mean and central as well as peripheral systolic blood pressure were or tended to be consistently larger in black patients compared to those from other population origins (105 (14) mmHg vs. 100 (11) mmHg (p-Value = 0.06), 136 (22) mmHg vs. 127 (17) mm Hg (p-Value = 0.006) and 148 (24) mmHg vs. 137 (18) mmHg (p-Value = 0.004)), respectively.

Associations of stroke volume and vascular mechanisms (systemic vascular resistance and total arterial compliance) with mean and central and peripheral systolic blood pressure were assessed in confounder adjusted models. These relationships were also analysed in stratified analysis by dialysis status. They were also re-evaluated in a sensitivity analysis among patients without established cardiovascular disease. The associations of stroke volume and vascular mechanisms with elevated systolic blood pressure were determined in receiver operator characteristic (ROC) curve analysis.

As only 27.1% (n = 13) of dialysis patients did not have an arteriovenous fistula or graft, we did not compare the associations of stroke volume and vascular mechanisms with blood pressure measures in appropriately adjusted multivariate analysis between patients with and without the respective vascular access mode.

The data were analysed on IBM SPSS statistical program (version 27.0, IBM, USA) and Statistica 8.0 application package (version 14.0, TIBCO, USA). The power of this study was calculated using STATA (version 18, Stata Corp, USA).

The baseline characteristics in non-dialysis and dialysis patients are shown in Table 1. Black patients were more frequently on dialysis, whereas white patients did more often not require dialysis. Dialysis patients exercised more frequently than their non-dialysis counterparts. Phosphate and parathyroid hormone concentrations were larger, whereas haemoglobin levels were lower in dialysis compared to non-dialysis patients. Dialysis patients used calcium channel blockers, diuretics and erythrocyte stimulating agents more often than those that were not on dialysis. Sodium-glucose cotransporter-2 inhibitors were not available in South Africa at the time of the study (2016) and hence not used in the present cohort.

The hemodynamic characteristics in non-dialysis and dialysis patients are given in Table 2. Central systolic blood pressure was larger in dialysis compared to non-dialysis patients. The other recorded hemodynamic characteristics did not differ among the two groups.

Data analysis aimed at identifying potential confounders for subsequent multivariate analysis was performed in all and non-dialysis as well as dialysis patients. Table 3 shows that age, sex, black population origin, exercising status, diabetes, haemoglobin levels and erythrocyte stimulating agents were each associated at p value ≤ 0.1 with one or more blood pressure measures in all, non-dialysis or/and non-dialysis patients.

Table 4 gives the relative contributions of cardiac output, stroke volume and systemic vascular resistance to the variation in mean arterial blood pressure in all included patients. None of the baseline characteristics were independently associated with mean arterial pressure (first model). Baseline characteristics explained only 10.5% (model R²= 0.105) of the variation in mean arterial pressure. When cardiac output (second model) or stroke volume (fourth model) were added to the model, both these characteristics were independently associated with mean arterial pressure and the model R² increased to 0.190 and 0.179, respectively. When stroke volume was replaced by stroke volume indexed to body surface area in the fourth model, the results were unaltered (model R²= 0.190, standardised β = 0.280, p value = 0.004). By contrast, neither heart rate (third model) nor systemic vascular resistance (fifth model) was related to mean arterial pressure and, accordingly, the model R² remained unchanged upon entering these characteristics. When systemic vascular resistance was re-calculated on the assumption that right atrial pressure was 10 mmHg (30) in the fifth model, the results were unaltered (model R²= 110, standardised β = 0.020, p value = 0.8). When stroke volume and systemic vascular resistance were entered into the same model (sixth model), both characteristics were strongly and to a similar extent associated with mean arterial pressure. The respective model R² increased to 0.324. In this regard, stroke volume and systemic vascular resistance were strongly interrelated (Spearman's rho = −0.797, p value < 0.001). Upon entering stroke volume x systemic vascular resistance without (seventh model) and with (eighth model) the individual interaction terms, the model R² remained unaltered. In the latter model, only the interaction between stroke volume and systemic vascular resistance tended to be associated (p value = 0.05) with mean arterial pressure. The partial correlations (95% confidence intervals) for the relationships in Table 4 are shown in Figure 1A.

Based on the sample size of 115 patients, α = 0.05 and 10 covariates in multiple regression models, the calculated power of the study was 0.999. This was established using the associations with mean arterial pressure.

Table 5 gives the relative contributions of stroke volume and the inverse of TAC (inv TAC) to the variation in central and peripheral systolic blood pressure in all included patients. Among the baseline characteristics, only black population origin tended to be (p value = 0.05) associated and was related to (p value = 0.04) central and systolic blood pressure, respectively. In the respective models, baseline characteristics explained ∼16% of the variation in systolic blood pressures. When stroke volume or inv TAC were added to the model, both these characteristics were independently associated with systolic blood pressures and the model R² increased to 0.246 to 0.266. When stroke volume and inv TAC were entered into the same model, both characteristics were strongly and to a similar extent associated with systolic blood pressures. The respective model R² increased to 0.708 for central systolic blood pressure and 0.685 for peripheral systolic blood pressure. In this regard, stroke volume and inv TAC were strongly interrelated (Spearman correlation coefficient = −0.581, p value < 0.001). Upon entering stroke volume x inv TAC without and with the individual interaction terms, the model R² increased further from 0.752 to 0.771. In the latter models, only the interaction between stroke volume and inv TAC were associated with systolic blood pressures (p value < 0.001 for both). In parallel to these findings, Pf was strongly associated with systolic blood pressures with a model R² of 0.910 and 0.929 for central and peripheral systolic blood pressure, respectively. The partial correlations (95% confidence intervals) for the relationships in Table 5 are shown in Figures 1B,C.

Based on the sample size of 115 patients, α = 0.05 and 10 covariates in multiple regression models, the calculated power was 0.999. This was established using the associations with both central and peripheral blood pressure.

Table 6 gives the mutually independent impact of stroke volume and vascular mechanisms on the variation in blood pressure measures in confounder adjusted stratified analysis by dialysis status. The relative contribution of stroke volume and vascular mechanisms (systemic vascular resistance and inv TAC) to the variation in mean arterial blood pressure and central as well as peripheral systolic blood pressure were assessed in models that included both characteristics. The relative potential impact of stroke volume and vascular mechanisms on blood pressure measures was similar in each of these models. Also, the potential impact of stroke volume and vascular mechanisms on blood pressure measures was consistently similar in non-dialysis and dialysis patients. The partial correlations (95% confidence intervals) for the relationships in Table 6 are shown in Figure 2.

Table 7 gives the mutually independent potential impact stroke volume and vascular mechanisms on blood pressure measures in confounder adjusted models among patients without established cardiovascular disease [n = 83 (72.2%)]. The relative contribution of stroke volume and vascular mechanisms (systemic vascular resistance and inv TAC) to the variation in mean arterial blood pressure and central as well as peripheral systolic blood pressure were assessed in models that included both characteristics. The respective relationships were similar to those in all patients (Tables 4, 5).

Stroke volume is a valuable marker of volume effects on systemic flow and hence volume status (38, 39). However, stroke volume is reportedly determined by not only cardiac preload (left ventricular end diastolic volume), but also left ventricular contractility and cardiac afterload (38, 39). Cardiac afterload (systemic vascular resistance and the inverse of TAC) was included in our multivariate regression models in Tables 4–7. Table 8 shows the contribution of left ventricular end diastolic volume, contractility (left ventricular midwall fractional shortening and lateral s') and pump function (ejection fraction) to the variation in stroke volume in confounder adjusted regression models. Based on the variable R²s in the models, left ventricular end diastolic volume, midwall fractional shortening and ejection fraction explained 39.4%, 16.1% and 12.5% of the variation in stroke volume. Hence, left ventricular end diastolic volume contributed 2.4 fold more than left ventricular midwall fractional shortening and 3.1 fold more than ejection fraction to the variation in stroke volume. Left ventricular lateral s' was not associated with stroke volume.

In view of these findings and in order to confirm that the potential impact of stroke volume on blood pressure measures is not attributable to left ventricular contractility and pump function, the confounder adjusted and mutually independent relationships of stroke volume and vascular mechanisms with blood pressure parameters as given in Tables 4, 5, were re-evaluated in regression models that included left ventricular midwall fractional shortening or ejection fraction as additional potential confounders. Table 9 shows that the associations of stroke volume (as well as systemic vascular resistance and the inverse of TAC) with blood pressure parameters were unaltered and independent of left ventricular function measures.

Table 10 gives the baseline characteristics in patients with normal and elevated systolic blood pressure. Eighty-eight (76.4%) of the patients had elevated systolic blood pressure. Systolic blood pressure was below 120 mmHg in only 13 (11.3%) of study participants. Age and CKD duration were associated with elevated systolic blood pressure. Black patients experienced elevated systolic blood pressure more often, whereas the reverse applied to Asian patients. Patients who exercised had more frequently elevated systolic blood pressure. Patients with elevated systolic blood pressure used calcium channel blockers and statins more frequently and less often had established cardiovascular disease.

Table 11 shows the hemodynamic characteristics in normal and elevated systolic blood pressure. As expected, all blood pressure measures were larger in patients with elevated systolic blood pressure. Stroke volume, cardiac output and left ventricular end diastolic volume were larger, whereas TAC was smaller in patients with uncontrolled elevated systolic blood pressure. By contrast, systemic vascular resistance was similar in both groups. Pf was larger in patients with uncontrolled elevated systolic blood pressure.

The hemodynamic correlates of elevated systolic blood pressure were assessed in ROC curve analysis. As given in Figure 3, stroke volume (AUC (95% CI) = 0.722 (0.590–0.853)) and inv TAC (AUC (95% CI) = 0.653 (0.590–0.853)) were associated with uncontrolled elevated systolic blood pressure whereas systemic vascular resistance was not (AUC (95% CI) = 0.476 (0.317–0.635)). Compared to stroke volume x inv TAC (AUC (95% CI) = 0.901 (0.843–0.950)), stroke volume x systemic vascular resistance was more weakly related to elevated systolic blood pressure (AUC (95% CI) = 0.668 (0.539–0.796)). The stroke volume x inv TAC association with elevated systolic blood pressure was paralleled by an equally strong relationship between Pf and elevated systolic blood pressure. Table 12 shows the associations of hemodynamic characteristics with elevated systolic blood pressure in stratified analysis by dialysis status. The respective relationships were consistently similar in non-dialysis and dialysis patients.