Allopurinol Is an Independent Determinant of Improved Arterial Stiffness in Chronic Kidney Disease: A Cross-Sectional Study

Background Arterial stiffness is increased in patients with CKD and is a powerful predictor of cardiovascular morbidity and mortality. Use of the xanthine oxidase inhibitor allopurinol has been shown to improve endothelial function, reduce left ventricular hypertrophy and possibly improve cardiovascular outcome. We explored the relationship between use of allopurinol and arterial stiffness in patients with chronic kidney disease (CKD). Methods Cross-sectional observational study of 422 patients with CKD with evidence of, or at high risk of, renal disease progression. Arterial stiffness was determined by carotid-femoral pulse wave velocity (PWV). Results The mean age was 63±16 years, median estimated glomerular filtration rate was 25 (interquartile range: 19–31) ml/min/1.73 m2 and mean PWV was 10.2±2.4 m/s. Seventy-seven patients (18%) were receiving regular allopurinol, 61% at a dose of 100 mg/day (range: 50–400 mg/day). Patients receiving allopurinol had significantly lower peripheral pulse pressure, central pulse pressure, central systolic blood pressure, serum uric acid level tissue advanced glycation end product levels but comparable high-sensitivity C-reactive protein levels. Use of allopurinol was associated with lower PWV. After adjusting for age, gender, ethnicity, tissue advanced glycation end product level, peripheral pulse pressure, smoking pack years, presence of diabetes mellitus and use of angiotensin converting enzyme inhibitor or angiotensin II receptor blocker, the use of allopurinol remained a significant independent determinant of PWV (mean difference: −0.63 m/s; 95% CI, −0.09 to −1.17 m/s, p = 0.02). Conclusion In patients with CKD, use of allopurinol is independently associated with lower arterial stiffness. This study provides further justification for a large definitive randomised controlled trial examining the therapeutic potential of allopurinol to reduce cardiovascular risk in people with CKD.

Arterial stiffness is increased in patients with CKD and is a powerful predictor of mortality in this patient group [17][18][19]. Arterial stiffness is associated with endothelial dysfunction and increased left ventricular mass and is thought to be a key initiating factor contributing to the elevated CV risk observed in patients with CKD [17]. Carotid-femoral pulse wave velocity (PWV) is considered to be the current 'gold-standard' measurement of arterial stiffness [20]. Although allopurinol has been shown to improve endothelial function, lower central aortic pressure and regress left ventricular hypertrophy in patients with CKD, its effects on PWV remain unclear [14]. We therefore examined the relationship between allopurinol use and carotid-femoral PWV in patients with CKD recruited into the Renal Impairment In Secondary Care (RIISC) cohort study.

Ethics Statement
The study was approved by the South Birmingham Local Research Ethics committee (reference: 10/H1207/6) and all participants gave informed and written consent. The study was conducted in accordance with the Declaration of Helsinki.

Study Design and Participants
The RIISC study is a prospective, observational cohort study of patients with CKD with evidence of, or at high risk of, renal disease progression. The inclusion and exclusion criteria have previously been reported in detail [21]. In brief, patients were included if they had stage 3 CKD with a declining estimated glomerular filtration rate (eGFR) of $5 ml/min/year or $10 ml/ min/5years or a urine albumin creatinine ratio (uACR) $70 mg/ mmol on three consecutive occasions, or stage 4/5 CKD. GFR was estimated (eGFR) using the four-variable Modification of Diet in Renal Disease (MDRD) equation with serum creatinine recalibrated to be traceable to an isotope derived mass spectroscopy method [22]. Patients with established renal failure receiving dialysis treatment and patients receiving immunosuppressive medication were excluded from the study. From October 2010 to November 2012, 437 out-patients under regular follow-up were recruited from renal clinics at two large teaching hospitals in the United Kingdom.

Baseline measurements
Baseline clinical information on participants' demographics, renal diagnosis, diagnosis of diabetes mellitus (DM), CV history, past medication history, family history, concomitant medication, smoking and alcohol consumption history were recorded. Presence of CVD was defined by history or other evidence of angina, previous myocardial infarction, previous stroke or transient ischaemic attack, peripheral vascular disease, a previous revascularisation procedure or heart failure. Presence of DM was defined as receiving treatment for DM or a confirmed clinical diagnosis of diet-controlled DM. Smoking history and pack years were determined by participant self-reporting. An allopurinol user was defined as a participant who was receiving any dosage of allopurinol on recruitment. We contacted the primary care physician of all allopurinol users to obtain further details on the reason for prescription, presence of side effect related to allopurinol in the initial 3 months of treatment and start date of allopurinol to determine the duration of exposure.
Peripheral blood pressure (BP) was measured in the dominant arm using a British Hypertension Society approved automated oscillometric sphygmomanometer (BPM-100, BpTRU TM ), which obtained a series of six BP readings at 1-minute intervals after 5 minutes of rest [23]. Mean peripheral BP was derived from the average of the 2 nd to 6 th BP readings. Carotid-femoral PWV was measured non-invasively using the Vicorder system (Skidmore Medical, Bristol, UK) as previously described [24]. This is an operator independent and highly reproducible technique with low within-subject variation [24]. After 5 minutes of lying supine, PWV measurements were obtained in duplicate; the mean of the measurements was used in data analyses. Central pressure waveforms were derived and analysed using pulse wave analysis as previously described [25]. The central pressure waveform was analysed to determine the augmentation index (AIx) and central aortic pressures. AIx represents the difference between the second and first peaks of the central pressure waveform in systole, expressed as a percentage of the pulse pressure. Given the known effects of heart rate, AIx was corrected to a heart rate of 75 beats per minute (AIx 75 ) [26].
Routine laboratory testing included blood haematological (Beckman Coulter Haematology Analyzer) and biochemical profiles and urinary albumin creatinine ratios (ACR) (Roche Hitachi 702 Analyser). Additional samples were centrifuged and serum was aliquoted and stored at 280uC and subsequently batch analysed for high sensitivity C-reactive protein (hsCRP) using a commercially available assay (SpaPlus assay, Binding Site). Tissue advanced glycation end product (AGE) level was determined by skin autofluorescence (SAF) using a validated AGE Reader TM (DiagnOptics BV, Groningen, The Netherlands).

Statistical Analysis
Statistical analysis was performed using SPSS version 19.0 (SPSS Inc, Chicago, IL). Numerical values are expressed as mean 6 standard deviation for parametric data or median (interquartile range) for non-parametric data. Normality of the distribution of data was assessed by visual inspection of histogram and normal probability plot [27]. Non-parametric variables were log transformed prior to analysis to achieve normal distribution. If normal distribution was not achieved after transformation, non-parametric tests were used. Parametric continuous data were compared using student t-tests and non-parametric using Mann-Whitney tests. Pearson or Spearman's bivariate correlation analysis was used to examine the relationship between parametric and nonparametric numerical variables, respectively. Correlation coefficient factors were expressed as 'r' for Pearson correlation analyses and 'rho' for Spearman's analysis. Categorical data were compared by x 2 tests. As age is strongly correlated with arterial stiffness, we divided the studied population into 4 age quartiles. Two-way analysis of variance (ANOVA) was performed to examine the interaction between age and the use of allopurinol as well as their individual effect on PWV. In addition, multiple linear regression was performed to explore the relationship between PWV and independent variables. Missing data was excluded by cases pairwise during analyses. Statistical significance is defined as a two-tailed p value ,0.05.

Baseline characteristics
Four hundred and thirty-seven patients were recruited; of whom 14 did not have PWV measured for technical reasons and were therefore excluded from the study. One patient who was receiving febuxostat was excluded. Therefore 422 patients were included in the analyses. The numbers of individuals at each stage of study are detailed in Figure 1. The baseline demographic and biochemical characteristics of the study population are shown in Table 1. The mean age was 63616 years with 60% of male gender and 71% of white ethnicity. Use of antihypertensive agents was common and 67% were receiving either an angiotensin converting enzyme inhibitor (ACEI) or angiotensin II receptor blocker (ARB). A small number (5%) were on both an ACEI and an ARB. There was a high prevalence of hyperuricaemia; 84% had a serum uric acid concentration greater than 360 mmol/L. The frequencies of different stages of CKD were: stage 1, 0.2%; stage 2, 1%; stage 3a, 5.3%; stage 3b, 23.1%; stage 4, 61.7%; stage 5, 8.7%. Seventyseven patients (18%) were receiving regular allopurinol, 61% as a dose of 100 mg/day (range: 50-400 mg/day). Haemodynamic parameters are presented in Table 2.

Use of allopurinol
The demographic, clinical and biochemical characteristics of the cohort and a comparison between allopurinol users and nonallopurinol users is shown in Table 1. There was a significantly higher proportion of patients of male gender and white ethnicity and a significantly lower proportion of patients who were current smokers among allopurinol users. Allopurinol users had a higher body mass index (BMI) than non-allopurinol users. There was no significant difference in age, prevalence of DM, prevalence of CVD, percentage of ex-smokers, smoking pack years, total number of antihypertensive agents used and use of ACEI/ARB between the groups. Allopurinol users had significantly lower serum uric acid concentrations and lower SAF level compared to non-allopurinol users. Other biochemical variables, including kidney function, albuminuria, lipid and bone profiles and hsCRP levels were not different between the groups. Allopurinol users had significantly lower peripheral and central pulse pressures (PP), central systolic BP (SBP) and PWV (Table 2). There were no differences in heart rate, peripheral SBP, AIx and AIx 75 between allopurinol and non-allopurinol users.
Among the 77 allopurinol users, details regarding allopurinol prescription were available from their primary care physician on 59 patients. Ninety four per cent were commenced on allopurinol for gout and 6% for asymptomatic hyperuricaemia. Side effects were reported in 12% during the first 3 months of allopurinol treatment: 3 had acute gout, 2 had a skin rash, 1 had diarrhoea and 1 complained of increased thirst. The mean duration of allopurinol use at recruitment was 74 months (SD: 54 months).

Use of allopurinol and pulse wave velocity
Univariate correlations with PWV are shown in Table 3. Although BMI positively correlated with serum uric acid level (r = 0.178, p,0.001), there was no significant correlation between BMI and PWV. Uric acid levels, kidney function and hsCRP also did not correlate with PWV. In participants who were not receiving ACEI/ARB (n = 141), there was no correlation between uric acid and PWV (p = 0.66). Six per cent of participants were receiving a thiazide, use of this drug did not correlate with levels of uric acid (p = 0.58) or PWV (p = 0.66).
Pulse wave velocity positively correlated with increasing age, white ethnicity, SAF, peripheral and central SBP and PP. Exsmokers and smoking pack years had a significant positive correlation with PWV whilst current smoking did not. Use of allopurinol (mean difference: 20.8 m/s; 95% CI, 20.2 to 21.4 m/s, p = 0.006), use of ACEI/ARB and Afro-Caribbean ethnicity were also associated with lower PWV. Neither the dose of allopurinol or duration of use of allopurinol had a significant correlation with PWV. Fifty-one per cent (n = 39) of the allopurinol users had a uric acid level below 416 mmol/L. Among allopurinol users, there was no difference in PWV between those with a uric acid below or above this threshold (p = 0.92).
Two-way ANOVA was used to explore the impact of age and use of allopurinol on PWV. Participants were divided into quartiles of age (19-50, 51-65, 66-76 and 77-92 years). Pulse wave velocity increased with age and was significantly lower in  Data are presented as frequency (percentage), mean 6 standard deviation or *median (interquartile range). Parametric data was analysed using unpaired two-tailed t-test or Pearson's x 2 unless otherwise specified. 1 Log-transformed prior to analyses.

11
Analysed using Mann-Whiteney U test. Abbreviations: ACR = albumin creatinine ratio; bpm = beats per minutes; CVD = cardiovascular disease; ACEI = angiotensin converting enzyme inhibitor; ARB = Angiotensin II receptor blocker; eGFR = estimate glomerular filtration rate, hsCRP = high sensitivity C-reactive protein; SAF = skin autofluorescence. doi:10.1371/journal.pone.0091961.t001 non-allopurinol users ( Figure 2). There was no interaction between age and use of allopurinol (p = 0.27). There were significant main effects for both age and use of allopurinol, with age having a large effect size (partial eta squared = 0.201, p,0.001) and use of allopurinol having a small albeit significant effect size (partial eta squared = 0.011, p = 0.03). A linear regression model was created with PWV as the dependent variable. Variables which correlated with PWV at a p value ,0.1 were included in a standard regression model. As there was strong co-linearity among the BP measures, peripheral PP was selected from these parameters for incorporation in the regression model as it had the strongest correlation (r = 0.45, p,0.001) with PWV. Similarly, smoking pack years was selected to adjust for the relationship between smoking history and PWV in the regression model. Preliminary analyses were performed to ensure no violation of the assumptions of normality, linearity and multicollinearity. Factors entered into the model were age, gender, ethnicity, smoking pack years, diagnosis of DM, SAF level, peripheral PP, use of ACEI/ARB and use of allopurinol. Age, peripheral PP and use of allopurinol were significant independent determinants of PWV (Table 4). In the regression model, the use of allopurinol was associated with a mean reduction of PWV of 0.63 m/s (95% CI, 20.09 to 21.17 m/s, p = 0.02). The model explained 35% of the variance in PWV. Substituting peripheral SBP, central SBP or central PP, for peripheral PP and substituting smoking pack years for current or previous smoking status made no appreciable difference to the model.

Discussion
This observational study in a prospectively recruited CKD cohort suggests that the use of allopurinol was associated with reduced arterial stiffness as measured by carotid-femoral PWV, the current gold-standard measurement of arterial stiffness [20]; this association was independent of age and BP. Arterial stiffness, which is a hallmark of CKD, is a well-recognised, powerful prognostic marker of CV morbidity and mortality in both the general and CKD population [17][18][19]. Increased arterial stiffness results in higher systolic pressures, greater pressure fluctuations and leads to ventricular-arterial uncoupling, myocardial hypertrophy and fibrosis. Alterations in the extracellular matrix and endothelial dysfunction promoted by chronic inflammation, increase oxidative stress and accumulation of advanced glycation end products, vascular calcification, and activation of reninangiotensin-aldosterone system (RAAS) have been postulated to contribute to increased arterial stiffness [17,28].
Allopurinol, a xanthine oxidase inhibitor, is commonly prescribed for patients with gout as a uric acid lowering agent. During the catalytic reaction that produces uric acid, xanthine oxidase generates reactive oxidative species, which may contribute to the development of endothelial dysfunction, hypertension and vascular damage [29]. Accumulating evidence from interventional studies indicates that allopurinol improves endothelial dysfunction [14,30], lowers left ventricular mass [14], and may slow progression of CKD and lower CV risk [15]. The results of this current study suggest that some of the beneficial effect of allopurinol may occur through reducing arterial stiffness.
Even among allopurinol users there was a high prevalence of hyperuricaemia. Although several large observational studies have reported a strong association between hyperuricaemia and CV morbidity or mortality, the evidence for hyperuricaemia as a risk factor or risk marker of CVD is conflicting [31][32][33][34][35][36][37][38][39]. Despite a significant association between allopurinol use and PWV, we found no significant direct association between serum uric acid levels and PWV. In a small group of patients with chronic heart failure, George et al demonstrated that the mechanism of improvement in endothelial function with allopurinol was Abbreviations: ACEI = angiotensin converting enzyme inhibitor; ACR = albumin creatinine ratio; AIx: augmentation index; AIx 75 = augmentation index adjusted to heart rate of 75 bpm; ARB = angiotensin II receptor blocker; SBP = systolic blood pressure; DBP = diastolic blood pressure; PP = pulse pressure; SAF = skin autofluorescence. Parametric data was analysed using Pearson correlation unless otherwise specified. *Natural Log transformed prior to analyses. **Non-parametric data was analysed using Spearman's bivariate correlation analysis. doi:10.1371/journal.pone.0091961.t003 attributable to reduced oxidative stress and not to uric acid reduction [30]. CKD is known to be associated with increased oxidative stress and acute-phase inflammation, which may both contribute to increased CV risk [40]. We found no significant difference in the levels of hsCRP between allopurinol users and non-allopurinol users, suggesting that inflammation was not a prominent mechanism in this association. Advanced glycation end product (AGE) has a bi-directional relationship with oxidative stress, including studies showing that increased oxidative stress is associated with formation and accumulation of AGE [41][42][43][44][45].
Level of tissue AGE as measured by SAF, which positively correlated with arterial stiffness, was found to be significantly lower in the allopurinol users when compared to non-allopurinol users, indicating that this biological pathway may be relevant to the link between allopurinol use and PWV described here.  As arterial stiffness and BP are closely related, it is possible that the effect of allopurinol on arterial stiffness may be mediated through improved BP. A recent meta-analysis showed that allopurinol is associated with a small but significant reduction in BP [46]. Several hypotheses have been postulated to explain this apparent association. The antioxidant effect of allopurinol was considered to play a major role in improving endothelial function and blood pressure regulation [46]. Nonetheless, there was no clear consensus to-date with regards to the effects of oral antioxidant on arterial blood pressure. While some demonstrated blood pressure lowering effect of antioxidant vitamins [47,48], others did not [49][50][51][52] and one study showed paradoxical blunting of exercise training-induced improvement in endothelial function with antioxidant administration [53]. In addition to its antioxidant property, there is emerging evidence that allopurinol can block the deleterious cardiovascular effect of angiotensin II [54][55][56]. In this current study we found a significant difference in both peripheral and central BP between allopurinol users and non-allopurinol users, despite a comparable prevalence of ACEI/ARB use and total numbers of anti-hypertensive agents between groups. However, after adjustment in a multivariate analysis, the use of allopurinol remained significantly associated with arterial stiffness. This suggests a distinct direct effect of allopurinol on vascular stiffness that is independent of BP and the use of ACEI/ARB. This observation is supported by an RCT of 66 patients with mild to moderate hypertension which reported a favourable effect of allopurinol on aortic compliance, independent of ACEI or thiazide-based antihypertensive therapy [57].

Limitations
There were a number of limitations in this study. Due to the observational, cross-sectional nature of the study and nonrandomised use of allopurinol, the association between allopurinol use and lower arterial stiffness reported here does not prove causality.
Although there were unequal distributions of gender, race, BMI, current smoking status and differences in serum uric acid between the groups, these are unlikely to have resulted in bias. Male gender was associated with higher PWV, however despite a higher proportion of males amongst allopurinol users, use of allopurinol remained associated with a lower PWV. In addition, people of Afro-Caribbean ethnicity had a lower PWV; most of the Afro-Caribbean participants were non-allopurinol users. The unequal distribution of gender and race between the groups was therefore unlikely to have resulted in bias against non-allopurinol use. As the number of non-white participants was small, we were unable to confidently examine the influence of allopurinol in different ethnic groups; this should be an area for future study.
Although there were differences in BMI and serum uric acid level between the groups, these parameters did not have a significant bivariate association with PWV; hence, they were unlikely to confound the findings. Smoking history is known to have significant influence on arterial stiffness and there was a higher proportion of current smokers in the allopurinol non-user group. In addition, the comparatively lower peripheral and central pressures in the allopurinol user group might have contributed to lower PWV as BP is a strong determinant of arterial stiffness. However, after adjusted for haemodynamic parameters and smoking history in the regression model, use of allopurinol remained associated with lower PWV.
Although all available confounding variables were included in this study, there may be other potential unknown confounders as the biology of vascular disease in CKD is complex. The measurement of hsCRP was performed only at single time-point rather than the two time-points two weeks apart recommended by the American Heart Association [58]. We did not have measurements of endothelial dysfunction, which is closely linked to arterial stiffness and CKD [17]. Finally, encouraging results have been reported on an effect of allopurinol in improving renal function in patients with asymptomatic hyperuricaemia [59] or delaying disease progression in patients with CKD [15,60]. However, due to the cross-sectional nature of the data we were unable to examine this relationship.

Conclusion
In summary, the data shown here suggests that allopurinol is independently associated with lower arterial stiffness in patients with progressive CKD. This adds to accumulating evidence of the favourable effect of allopurinol on cardiovascular outcomes in a well-defined CKD cohort and indicates one mechanism by which this may occur. This study provides further justification for a large definitive RCT examining the therapeutic potential of allopurinol to reduce cardiovascular risk in people with CKD.