Impaired renal function impacts negatively on vascular stiffness in patients with coronary artery disease

We included 160 patients with severe triple vessel CAD who underwent elective coronary artery bypass graft surgery at the Western Infirmary Glasgow between 2004 and 2008 [7, 12‐14]. Thirty nine patients with CKD stages 2 to 4 who attend renal clinics at the Western Infirmary Glasgow [15] and 130 healthy control subjects who were recruited from a local sports centre, from a surgical ward in Gartnavel General Hospital Glasgow or were Glasgow University employees, also participated in this study. Hence, a total of 329 subjects, 160 with CAD and 169 without CAD, were studied (Table 1). As a result of rigorous quality checks not all of the parameters below are available in all study participants.

Demographic data were collected and a blood sample was taken to assess lipid profile, CRP, serum creatinine, calcium and phosphate using standard biochemical methods [7]. A lithium-heparinate sample was kept on ice for a maximum of 1 hour prior to centrifugation. Plasma was stored for biomarker analysis at −80°C. eGFR was determined using the 4-variable MDRD formula, calibrated to isotope dilution mass spectrometry reference. Impaired renal function was defined as eGFR <60 mL/min.

The study adheres to the principles of the Declaration of Helsinki and was approved by the West of Scotland Research Ethics Committee (Reference Number 06/S0703/110). All participants gave written informed consent.

Carotid-femoral PWV was assessed using the SphygmoCor® device (AtCor Medical Ltd., Sydney, Australia) as previously described [16]. A measuring tape was used to assess the distance between the carotid and femoral artery recording sites. PWV was calculated automatically by dividing this distance by the time interval between the rapid upstroke in the pulse wave at the carotid and femoral arteries using the peak of the R-wave on electrocardiography as a reference point.

Oxidised (GSSG) and reduced glutathione levels (GSH) were determined in whole blood using a photometric method (OXIS International Inc., Foster City, California, USA) as previously described [17]. Oxidised low-density lipoprotein (LDL) cholesterol (oxLDL) was determined using a competitive enzyme-linked immunosorbent assay (Mercodia AB, Uppsala, Sweden).

Mononuclear cells were extracted from whole blood as previously described and diluted to 5×10⁶ cells/mL [18]. Superoxide production was measured in triplicate by electron paramagnetic resonance (EPR) spectroscopy (Bruker BioSpin e-scan R, Bruker Corporation, Rheinstetten, Germany) using the spin probe 1-Hydroxy-3-carboxy- 2,2,5,5-tetramethylpyrrolidine (CPH; Noxygen, Elzach, Germany) [19]. Maximum superoxide production was assessed after stimulation with phorbol 12-myristate 13-acetate (PMA, 3.2 μM; Sigma-Aldrich, Dorset, UK) and fold changes compared to basal production have been calculated. For assessment of whole blood superoxide generation blood was collected in lithium heparinate containing tubes, kept on ice and processed within half an hour. EPR measurements were performed after adding the spin probe CPH to a final concentration of 500 μmol/L [12]. Instrument settings were: microwave power, 22 mW; centre field, 3375 G; modulation amplitude, 2.27 G; sweep time, 5.24 s; sweep width, 60 G; 10 scans. Superoxide levels were recorded once a minute for 10 min and the rate of superoxide anion production was calculated as counts per minute. For mononuclear cells this has been standardised to a rate per 10⁶ cells per minute whereas for whole blood arbitrary units (AU) are presented.

Statistical analysis was conducted using SPSS statistical package. Data are presented as mean [95% confidence interval] or median [interquartile range] as appropriate. Two sample Student’s t-test or Mann–Whitney U-test were used to compare data between groups of patients. Correlation was assessed by calculating Pearson’s (r) and Spearman’s (ρ) correlation coefficients as appropriate. Linear regression analysis was performed to show determinants of PWV using both a model where all variables were forced in and then a stepwise model (P_in < 0.05 and P_out > 0.10). For interaction analysis between eGFR and CAD status we centered eGFR around the mean of eGFR and calculated an interaction term by multiplying centered eGFR with CAD status (coded as 1 = “no CAD”, 2 = “CAD”); these terms were then used in linear regression analysis. A P-value of 0.05 (two sided) was considered significant.

Clinical and demographic characteristics of study participants are presented in Table 1. As expected, patients with CAD and/or eGFR < 60 mL/min were older than control subjects, and due to lipid-lowering therapy, total and LDL cholesterol levels were lowest in patients with CAD. Blood pressure was similar across the study groups.

PWV was greater in patients with eGFR < 60 mL/min compared to those with eGFR ≥60 mL/min, both in patients without CAD (10.2 [9.1;11.2] vs 7.3 [6.9;7.7] m/s; P < 0.001) and with CAD (10.1 [8.9;11.2] vs 8.1 [7.5;8.7] m/s; P = 0.001). In those with an eGFR <60 mL/min, there was no difference in PWV between patients with and without CAD (P = 0.935). In contrast, PWV was greater in the CAD group when only patients with normal renal function were selected (P = 0.029) (Figure 1A).

Assessing eGFR as a continuous variable, a close correlation was found between PWV and eGFR (r = −0.476; P < 0.001) in the whole study cohort (Figure 1B). Linear regression analysis demonstrated that eGFR (β = −0.293; P < 0.001) contributed to PWV independently of age (β = 0.453; P < 0.001) and systolic blood pressure (β = 0.153; P = 0.009); whereas presence or absence of CAD (β = −0.015; P = 0.798), sex (β = 0.040; P = 0.523) and body mass index (β = 0.064; P = 0.294) did not contribute significantly. A model containing only age, systolic blood pressure and eGFR explained 42% of the variability of PWV.

Linear regression analysis to study interaction between CAD and eGFR confirmed the key role of eGFR as determinant of PWV. Centered eGFR (β = − 0.603, P = 0.002), but not CAD status (β-0.081; P = 0.191) and the interaction term (β = 0.137, P = 0.479) significantly contributed to the model explaining PWV.

We explored if markers of oxidative stress differed between patients with CAD and those without CAD. In line with our previous observations [16, 17] we found significantly increased levels of oxidative stress markers in patients with CAD compared with those without (Table 2). In contrast, no significant differences were found in patients with CAD when stratified into groups with normal or impaired renal function (Table 2). Oxidative stress therefore does not seem to explain the increased vascular stiffness associated with impaired renal function in patients with CAD.

We then studied circulating biomarkers of inflammation and cell adhesion in control subjects and in patients with CAD of which both were stratified into groups with normal or impaired renal function. As expected, a number of markers were different between those with and without impaired renal function in the control group (Table 3). In a next step we analysed correlations between these markers and PWV and eGFR in the whole cohort (Table 4). We found that levels of E-selectin, VCAM-1, adiponectin, osteopontin and leptin were consistently different between groups with and without impaired renal function and were correlated with both eGFR and PWV. Scatterplots for these markers illustrating potential correlations that could provide a link between renal function and vascular stiffness are shown in Figure 2.

When added to the multivariate model to study the association of the markers with PWV independently of eGFR, age, systolic blood pressure, sex, presence or absence of CAD and body mass index only E-selectin (β = 0.307, P < 0.001; also included: age [β = 0.617, P < 0.001]), osteopontin (β = 0.336, P < 0.001; also included: age [β = 0.538, P < 0.001], body mass index [β = 0.188, P = 0.021]) and adiponectin (β = 0.203, P = 0.017; also included: age [β = 0.502, P < 0.001], eGFR [β = − 0.186, P = 0.036] and systolic blood pressure [β = 0.215, P = 0.014]) were part of the final models. Of note, the contribution of osteopontin to PWV remained statistically significant even after additional adjustment for phosphate (osteopontin, β = 0.206, P = 0.049; also included: age [β = 0.638, P < 0.001] and systolic blood pressure [β = 0.205; P = 0.046]) or calcium (osteopontin, β = 0.287, P = 0.005; also included: age [β = 0.631, P < 0.001]). Phosphate, CRP, VCAM-1 and leptin were no significant determinants of PWV when adjusted for other factors.

Increased vascular stiffness is an independent predictor of cardiovascular outcome [2, 20, 21]. Impaired renal function is associated with vascular stiffening, but the underlying mechanisms are incompletely understood [1, 10, 11, 22, 23]. From a recent study by Ilyas et al. [8] it appears that there is an additive contribution of impaired renal function to vascular stiffness even in patients with moderately severe CAD. In our present study we have gone one step further and studied patients with severe triple vessel CAD who require surgical revascularisation, compared them with patients without CAD and assessed a wide range of markers of oxidative stress, inflammation and cell adhesion.

We have previously reported reduced aortic compliance in patients with ESRD [7]. In the present study we not only found a linear relationship between renal excretory function and vascular stiffness in mild-to-moderate renal impairment, but also demonstrated that this association is independent of other cardiovascular risk factors including systolic blood pressure and age.

Inflammatory processes play an important role in the pathogenesis of vascular disease [9], and renal impairment has been identified as a chronic low grade inflammatory state [11]. The present study demonstrates that levels of E selectin and VCAM-1 are increased in patients with CAD and renal impairment relative to patients with CAD alone. In multivariate analysis only E-selectin was an independent predictor of PWV, and the final model only contained age and E-selectin but not eGFR or blood pressure. It therefore appears plausible, although not directly proven in this cross-sectional study, that effects of impaired renal function on vascular stiffness are at least in part mediated by increased levels of E-selectin. We have previously demonstrated a role of E-selectin in the early stages of pre-eclampsia, a condition that is also associated with renal damage and endothelial dysfunction [24]. Other studies [25, 26] reported that adhesion molecules are independent predictors of cardiovascular events in patients with ESRD, although not all studies confirm these data [27]. Adhesion of inflammatory cells to the endothelium is an early step in the development of vascular disease and further induces local vascular inflammation [28].

We have also found independent associations of osteopontin and adiponectin with vascular stiffness. Osteopontin is an extracellular matrix protein that contributes to the development of atherosclerosis as it acts as a pro-inflammatory cytokine to induce macrophage adhesion and migration, it promotes vascular smooth muscle cell proliferation and may mediate vascular calcification [29]. Osteopontin mRNA is expressed in aortic atherosclerotic plaques in patients with CAD, but not in the arteries of healthy controls, and expression is proportional to severity of disease [30]. In a prospective study of patients with established CAD, osteopontin was an independent predictor of fatal and nonfatal cardiovascular events even after adjustment for traditional risk factors [31]. Lorenzen et al.[29] demonstrated that in patents with CKD, osteopontin levels increase linearly with progressive decline in GFR. Our study suggests that osteopontin levels are further increased in patients with CAD and concomitant renal impairment, and thus may contribute to the development of vascular disease in this group. It is important to note that the association between osteopontin and vascular stiffness in our study was independent of calcium or phosphate levels. The findings on adiponectin are more difficult to interpret. Adiponectin is an adipose tissue-derived protein that is generally thought to have vasoprotective properties. Lower levels of adiponectin have been found in a number of cardiovascular conditions including diabetes [32] and hypertension [33]. On the other hand, renal failure is associated with reduced clearance of adiponectin, and a recently proposed interaction between cystatin C and adiponectin could explain the paradoxical finding of high adiponectin levels in renal failure and the absence of a vasoprotective effect under these conditions [34].

Oxidative stress has been proposed as a unifying concept to explain the pathophysiology of cardiovascular disease in uraemic patients [11]. We have previously shown that vascular superoxide production is a significant determinant of vascular stiffness in patients without renal disease [16]. Levels of reactive oxygen species are increased in patients with CAD [12, 17], but in our present study there was no further rise in oxidative stress to explain the increased vascular stiffness in those with concomitant renal impairment. However, increased expression of adhesion molecules may lead to increased numbers of inflammatory cells in the vessel wall and thereby promote functional and structural changes.

A limitation of our study is that renal impairment was assessed by estimating GFR from serum creatinine using the MDRD equation. eGFR is a measure of renal function which also integrates other important determinants of vascular stiffness, namely age and sex. Without exact measurement of GFR (e.g. by inulin clearance) we cannot precisely separate the effect of renal impairment from that of age on vascular stiffness. However, this issue has been raised in other studies as well and it is not necessarily a limitation [35]. In fact, eGFR captures risk related to a number of characteristics and is an excellent surrogate marker in clinical practice [35, 36]. We acknowledge, however, that assessment of renal function at a single time point does not take variability of renal function in response to diet, fluid intake and medication into account.

Another limitation of this study is the cross-sectional design, which precludes inferences about the contribution of renal impairment on vascular stiffness and the effect on future cardiovascular risk, especially as vascular stiffening itself is a process progressing over a long period of time, where improvements of vascular stiffness are associated with improved outcome in patients with CAD [37]; rather we must limit ourselves to showing differences between patient groups. However, vascular stiffness is an established independent predictor of cardiovascular outcome in the general population [38, 39], in patients with CAD [21] and in individuals on dialysis [2, 40]. Recently, Ilyas et al.[8] suggested that assessment of arterial stiffness in patients with established CAD and mild renal impairment has some prospective value, as lower GFR and higher PWV predicted a shorter time to cardiovascular hospitalization and all cause mortality.

We also acknowledge that we did not control for drug therapy in this study. Therefore a number of potentially important factors, in particular lipid levels that are affected by statin therapy which is especially prevalent in patients with CAD, may not be adequately represented in our analyses.

Finally, residual confounding is a potential limitation of any study using multivariate analysis. We apreciate that other factors than renal function may have additional effects on vascular stuffness.