Introduction

Cardiovascular disease (CVD) is one of the most severe complications of hypertension,1, 2, 3, 4, 5 and many risk factors are reported to be associated with cardiovascular events in hypertensive patients. Uric acid (UA) is one of the risk factors for CVD in hypertensive patients.6, 7 Previous meta-analyses revealed that hyperuricemia may marginally increase the risk of coronary heart disease events, independent of traditional coronary heart disease risk factors,8 and that hyperuricemia may modestly increase the risks of both stroke incidence and mortality.9 In addition, the results of several studies indicate that a higher incidence of CVD occurs not only in patients with ‘hyperuricemia’ but also in patients with normal-to-high serum UA levels.10, 11, 12

In contrast, some multivariate analyses have not shown that the serum UA level is a significant risk factor for CVD, independent of traditional risk factors.13, 14, 15 One of the reasons proposed for this finding is that subjects with high CVD risk often also have conditions that affect the serum UA level, such as renal dysfunction, hyperinsulinemia, oxidative stress, tissue ischemia and taking diuretics for hypertension.16, 17 Moreover, UA can act as an antioxidant; thus, some studies have indicated that UA might have a beneficial effect on CVD.18, 19 These factors could affect the correlation between serum UA levels and CVD, which is therefore not fully understood and still controversial.

In this study, to investigate whether the possible confounding effects described above have an effect on the association between UA and CVD, we performed a longitudinal and prospective analysis of a cohort of essential hypertension patients and evaluated various factors, including biochemical and physiological measurements, usage of antihypertensive agents, complicating disorders and lifestyle. We then assessed the hypothesis that serum UA levels are an independent predictor of CVD and mortality in hypertensive patients.

Methods

Study subjects and study design

The present hospital-based cohort study was designed as part of the Non-invasive Atherosclerotic evaluation in Hypertension study.20 In this study, a total of 774 serial outpatients who had been diagnosed with essential hypertension were sequentially recruited between January 1998 and June 2004 at Osaka University Medical Hospital. Out of the 774 patients, 34 were excluded because of arterial fibrillation, 6 were excluded because of malignant disease and 23 were excluded because of incomplete evaluation for UA. A clinical survey was conducted for each patient, and 42 were excluded because of lack of follow-up. Thus, a total of 669 patients were included in the analysis (Figure 1). The study protocol was approved by the hospital ethics committee, and written informed consent was obtained from all participants.

Figure 1
figure 1

Flow chart of the selection process of study participants.

Full size image

Follow-up evaluation

Clinical follow-up was conducted by clinical visits, mailed questionnaires and telephone contacts every September from 2003 to 2010. The questionnaire included events related to hypertensive complications and cause of death. We also confirmed the responses of each patient in detail by checking them against patient medical sheets. The primary end point of this study was new onset of stroke (paralysis and diagnosis with computed tomography and/or magnetic resonance imaging), new onset of CVD, including angina pectoris (typical chest pain with ST segment changes on electrocardiograms), myocardial infarction (ST segmental elevation with more than two-fold elevation of creatinine kinase), heart failure (diagnosed using American Heart Association criteria) and rupture of an aortic aneurysm (diagnosed employing ultrasound echography or computed tomography). The follow-up duration was considered to encompass the interval from the initial evaluation to the time of event onset or September 2010. The average follow-up period was 7.1±0.1 years and the follow-up ratio was 93.5%.

Blood pressure (BP) measurements

Conventional BP measurements were performed by trained observers with an electronic sphygmomanometer (HEM-705IT or HEM-711; OMRON, Tokyo, Japan). Following the guidelines for the management of hypertension in Japanese Society of Hypertension 2009, clinic BP was measured at least two times in a sitting position after 5 min of rest; we used the average of the two readings as the official BP if the difference in the measured values was <5 mmHg. When the difference in the measured values was >5 mmHg, additional measurements were conducted to obtain stable BP readings, and we adopted the average of the two stable readings as the official BP.

Renal function

Estimated glomerular filtration rate was calculated using the following equation:

Pulse wave velocity (PWV) measurement

We evaluated carotid–femoral PWV (cfPWV) as a representative PWV to evaluate arterial stiffness. For the cfPWV measurements, participants visited the hospital in the morning; they were instructed not to take any antihypertensive drugs, nitrate or aspirin for 8 h preceding their visit. Measurements were performed in the morning with each patient in the supine position after 30 min of rest. cfPWV measurement was performed in a controlled environment at 22±2 °C. Three measurements taken 2 min apart were averaged. The intra-observer coefficient of variation was calculated using three measurements from each of seven healthy men. The intra-observer coefficient of variation for cfPWV was 2.8±1.2%.

We determined cfPWV using a model FCP-4731 (Fukuda Denshi, Tokyo, Japan), which allows on-line pulse wave recording and automatic calculation, using a previously reported method.21

Statistical analysis

Data were analyzed using JMP ver. 9.0.1 (SAS, Cary, NC, USA) and are presented as the mean±s.e.m. An event-free curve was estimated employing the Kaplan–Meier method. The log-rank test was used to compare the differences in event-free rates among subjects who were placed into four groups according to their UA level. Cox proportional hazard models were used to detect the relative risk of UA with regard to the prognoses. Analysis of variance and Student’s t-test were used to test for significant differences between patients with higher and lower variability in systolic BP. A value of P<0.05 was regarded as statistically significant.

Results

During a mean follow-up period of 7.1±0.1 years, 71 strokes, 58 cases of CVD, 117 cases of stroke+CVD, 64 deaths and 156 cases of stroke+CVD+death were recorded. Baseline clinical characteristics are shown in Table 1. We classified the subjects into four groups according to serum UA levels: >8.0 mg dl−1, high (n=27); 7.0–8.0 mg dl−1, middle-high (n=60); 6.0–7.0 mg dl−1, middle-low (n=142); and <6.0 mg dl−1, low (n=440). We then compared various parameters among these four groups. Subjects with higher UA levels showed significantly higher male-to-female ratio; to have a higher body mass index (BMI) and triglyceride and serum creatinine levels; and to have lower high-density lipoprotein cholesterol levels and diastolic BP. Subjects with lower UA levels had significantly higher concentrations of hemoglobin A1c and a higher estimated glomerular filtration rate (Table 1). There was no correlation between UA levels and systolic BP or cfPWV representing arterial stiffness. Chi-square analysis also revealed that patients in the high UA group were significantly more likely to be taking diuretics (high UA group: n=5, 19.2%; middle-high UA group: n=6, 10.0%; middle-low UA group: n=16, 11.4%; and low UA group: n=25, 5.69%; P=0.0162). There was no significant correlation between the serum UA level and other classes of antihypertensive agents.

Table 1 Baseline clinical characteristics of the study participants
Full size table

Kaplan–Meier analysis revealed that subjects in the high UA group had a higher frequency of stroke and CVD (P=0.0120; Figure 2c), total mortality (P=0.0021; Figure 2d) and mortality, stroke and CVD (P=0.0028; Figure 2e). On the other hand, there was no significant difference in CVD (P=0.1487; Figure 2a) and stroke (P=0.1499; Figure 2b) among the subjects in the four UA groups.

Figure 2
figure 2

Kaplan–Meier analysis of occurrence of CVD (a); stroke (b); stroke and CVD (c); mortality (d); and mortality, stroke and CVD (e) among the four groups of patients sorted by UA level (low, middle-low, middle-high and high).

Full size image

To further investigate, we used a Cox proportional hazard model to reveal whether the serum UA level could predict the onset of cardiovascular events or mortality, independent of other risk factors. First, we adjusted for gender, BMI, diastolic BP, serum creatinine, high-density lipoprotein cholesterol, tryglyceride and hemoglobin A1c, because these parameters were significantly correlated with UA in this cohort. Table 2a shows the results of the Cox proportional hazard model in which the UA level was treated as a group variable; the serum UA level was significantly and independently predictive of CVD and stroke (P=0.0420) and mortality, CVD and stroke (P=0.0096). Table 2b shows the results of the Cox proportional hazard model in which UA was treated as a continuous variable; the serum UA level was significantly and independently predictive of CVD (relative risk=1.36; P=0.0149), CVD and stroke (relative risk=1.23; P=0.0332), and mortality, CVD and stroke (relative risk=1.20; P=0.0315). Additionally, in a Cox proportional hazard model that adjusted for the usage of diuretics instead of diastolic BP, the serum UA level was significantly and independently predictive of cardiovascular events. A gender-segregated Cox proportional hazard model in which UA was treated as a continuous value revealed that in female subjects, the serum UA level was significantly and independently predictive of CVD (relative risk=1.74; P=0.0043), CVD and stroke (relative risk=1.52; P=0.0055), and mortality, CVD and stroke (relative risk=1.46; P=0.0077). However, in male subjects, the serum UA level was not predictive of CVD (P=0.5507), CVD and stroke (P=0.5385) or mortality, CVD and stroke (P=0.9807).

Table 2 Cox proportional hazard models adjusted for gender, BMI, DBP, creatinine, HDL cholesterol, triglyceride and HbA1c
Full size table

Finally, we conducted a Cox proportional hazard model in which we adjusted for gender, age, presence or absence of diabetes mellitus, presence or absence of dyslipidemia and systolic BP, because these parameters are widely recognized as traditional risk factors for cardiovascular events and mortality. Table 3a shows the results of the Cox proportional hazard model in which UA was treated as a group variable; the serum UA level was significantly and independently predictive of mortality (P=0.0285) and mortality, CVD and stroke (P=0.0316). Table 3b shows the results of the Cox proportional hazard model in which UA was treated as a continuous variable; the serum UA level was significantly and independently predictive of CVD (relative risk=1.30; P=0.0073); CVD and stroke (relative risk=1.19; P=0.0141), mortality (relative risk=1.23; P=0.0353), and mortality, CVD and stroke (relative risk=1.18; P=0.0083), but not stroke (P=0.4268), after adjustment for traditional risk factors. A gender-segregated Cox proportional hazard model in which UA was treated as a continuous value revealed that in female subjects, serum UA levels were significantly and independently predictive of CVD (relative risk=1.65; P=0.0002), CVD and stroke (relative risk=1.40; P=0.0023), total mortality (relative risk=1.71; P=0.0028) and mortality, CVD and stroke (relative risk=1.44; P=0.0002). However, in male subjects, the serum UA level was not predictive of CVD (P=0.9950), CVD and stroke (P=0.5773), total mortality (P=0.6480) or mortality, CVD and stroke (P=0.7342).

Table 3 Cox proportional hazard models adjusted for traditional risk factors
Full size table

Discussion

Many studies have investigated the correlation between serum UA levels and CVDs such as coronary heart disease,8 stroke9 and chronic kidney disease.22, 23, 24 Serum UA levels have also been reported to be correlated with metabolic syndrome25 and diabetes mellitus.26 However, various clinical conditions in subjects at high risk of CVD, such as renal dysfunction, hyperinsulinemia, oxidative stress and taking diuretics for hypertension, can have an effect on serum UA levels. Moreover, many studies have reported that elevated serum UA levels could cause hypertension.27, 28, 29 The complex interactions between CVD risk and serum UA levels have made it difficult to interpret the results of previous studies.

As in previous studies, in this cohort study, subjects with higher serum UA levels had significantly higher BMIs, higher levels of triglycerides and serum creatinine and a higher rate of taking diuretics. However, there was no significant correlation between the serum UA level and systolic BP, because 59.3% of the subjects were already under medical treatment for hypertension. A Cox proportional hazard model showed that, after adjustment for confounding factors (that is, serum creatinine, BMI, hemoglobin A1c, dyslipidemia, diastolic BP and using diuretics) that were reported to be affected by serum UA levels in previous reports and were significantly correlated with serum UA levels in this study, serum UA levels were also a significant and independent risk factor for cardiovascular events.

Some studies have reported J-shaped relationships between the serum UA level and CVD or total mortality.30, 31, 32 This was also the case in the present study; when UA was treated as a group variable, the Cox proportional hazard models showed that patients with middle-high UA levels had a lower relative risk than those with middle-low UA levels for cardiovascular events; however, other risk factors, such as BMI and triglyceride and serum creatinine levels, were correlated with the serum UA level in a linear fashion. Previous studies reported that increased UA levels might have a beneficial effect on CVD by serving as an antioxidant;18, 19 our results may support a beneficial effect of UA on CVD. To determine the optimal UA levels for preventing CVD, a larger cohort is required, as well as a longer period of follow-up.

Our findings revealed that the serum UA level was a significant and independent risk factor for cardiovascular events after adjustment for other traditional risk factors or confounding factors in female subjects, but not in male subjects. Females are known to show lower serum UA levels than males because of estrogenic compounds that enhance the renal urate excretion and increase urate clearance.33 Our results support the findings of previous reports, suggesting that the correlation between hyperuricemia and CVD differs between the sexes.34, 35

The present study has several limitations. First, the sample size in this study was relatively small; to avoid study bias and confirm the observations of this study, a larger cohort and multi-center trials are needed. Second, the subjects enrolled in this study were receiving medical treatment, such as angiotensin receptor type II blockers, angiotensin-converting enzyme inhibitors and statins; these drugs may have contributed to better outcomes.

In conclusion, we have presented evidence that in patients with essential hypertension, especially women, the serum UA level is a significant and independent risk factor for cardiovascular events after adjusting for other traditional risk factors and confounding factors.