Uric acid is a product of the activity of xanthine oxidase, an enzyme linked to oxidative stress, endothelial dysfunction, and heart failure. It is unclear whether adding uric acid levels to the assessment of cardiovascular risk might contribute to the improved ability to stratify cardiovascular risk. The purpose of this study was to evaluate the prognostic value of serum uric acid levels in a large cohort of men and women at high risk of cardiovascular disease.
Serum uric acid levels were determined in all patients seen for primary/secondary cardiovascular disease prevention at the Cleveland Clinic Section of Preventive Cardiology and Rehabilitation between 1998 and 2004, and all data were entered into the Preventive Cardiology Information System (PreCIS) database. Vital status of the patients was determined through the Social Security Death Index. Death from all causes was summarized across quartiles of uric acid values.
A total of 3,098 patients (age range 18–87 years) were identified in the database, among whom 43% had cardiovascular disease. There were 156 deaths (5%) during the 14,262 person-years of followup. For each 1-mg/dl increase in the serum uric acid level, there was a 39% increase in the risk of death (by Cox regression analysis). After adjusting for age, sex, smoking status, alcohol consumption, weight, body mass index, waist circumference, blood pressure, history of cardiovascular disease, estimated glomerular filtration rate, levels of cholesterol fractions, and plasma glucose levels, the serum uric acid level continued to predict the risk of death (hazard ratio = 1.26 [95% confidence interval 1.15–1.38], P < 0.001). This association was present regardless of diuretic use. Concordance index (C statistic) analyses showed that uric acid significantly improved the predictive accuracy of a model that included Framingham Heart Study score factors, metabolic syndrome components, and fibrinogen levels.
Serum uric acid levels are an independent predictor of death in patients at high risk of cardiovascular disease. Further studies are warranted to evaluate its prognostic implications and potential utility in the monitoring of therapy.
Uric acid (urate) is a product of the activity of xanthine oxidase, an enzyme increasingly implicated as a mechanistic participant in oxidant stress and cardiovascular disease. Xanthine oxidase activity is increased during ischemia and heart failure, and treatment with xanthine oxidase inhibitors has favorable effects on myocardial oxygen consumption and endothelium-dependent vascular function (1, 2). Hyperuricemia has also been linked to multiple proatherogenic processes, including increased oxidative stress (3, 4), vascular smooth muscle cell proliferation (5), leukocyte activation (6), and crystal formation within coronary atherosclerotic plaques (7).
Although numerous studies have noted an association of elevated serum uric acid levels with hypertension and cardiovascular disease, particularly in patients with heart failure, a clinical role of uric acid as a predictor of death is still uncertain. Among the confounders is the association of uric acid with markers of insulin resistance, as well as its variance with sex, renal function, and diuretic use.
Several studies have shown that elevated uric acid levels predict an increased risk of cardiovascular events and death from cardiovascular causes as well as all causes (7–12), while others have not found such independent association (13–16). Differences in the age, sex, and other baseline characteristics of the study population, as well as differences in the cohort size, duration of followup, and design of analyses may explain these disparate conclusions. Some of the studies that failed to show uric acid as an independent predictor of death were underpowered (15) or included too few events (13, 15). On the other hand, the studies that supported uric acid as a risk factor for death did not always adjust for renal function or markers of metabolic syndrome (8) or did not include women (9, 10, 12). Studies demonstrating a relationship between uric acid and death suggested that it was a more important risk factor for death from cardiovascular causes in women than in men (8).
Uric acid assays are clinically available and inexpensive. Clarification regarding patients who may benefit from its measurement is lacking. We evaluated the predictive role of serum uric acid levels with respect to all-cause mortality in a large cohort of men and women at high risk of death from cardiovascular disease in whom the uric acid levels were routinely obtained at the baseline evaluation and their data entered into the Preventive Cardiology Information System (PreCIS) database. The PreCIS database is comprehensive and allowed for adjustments for numerous potential confounders, including diuretic use.
PATIENTS AND METHODS
The PreCIS database includes information from patients referred to the Cleveland Clinic Section of Preventive Cardiology and Rehabilitation for evaluation of the prevention of coronary artery disease (CAD). This referral clinic is run by physicians and nurse-clinicians, and assessments and treatments are algorithm-driven. From the PreCIS database, we performed a retrospective review of patients who had their baseline visit to the clinic between April 1, 1998, and July 14, 2004 (n = 3,098). During this period, determination of uric acid levels was part of the initial visit evaluation protocol.
The vital status of the patients in the database is updated monthly using the Social Security Death Index. Mortality rates in our study were calculated based on the number of deaths that occurred in the study cohort between April 1, 1998, and November 15, 2005. For survivors, followup time was defined as the time from the uric acid determination (i.e., initial visit) until November 15, 2005. For nonsurvivors, followup time was defined as the time from the uric acid determination to the event of interest (i.e., death).
Analysis of the PreCIS database was approved by the Institutional Review Board of the Cleveland Clinic.
During the visit to the clinic, demographic information, medical history, physical examination, and laboratory data on each patient were entered into an electronic medical record and corresponding database. All laboratory measurements were performed at the same time as the uric acid determination. Hypertension was diagnosed according to the patient's history or the taking of antihypertensive medications. Diabetes mellitus was diagnosed according to the medical history, a fasting glucose level >125 mg/dl, or the taking of glucose-lowering medications. Alcohol consumption was defined as >3 alcoholic drinks per week. Blood pressure values consisted of an average of 3 blood pressure measurements obtained at the time of the initial visit. Serum uric acid levels were determined with the uricase colorimetric test (MAS ChemTRAK Liquid Unassayed Chemistry Controls; Medical Analysis Systems, Camarillo, CA). Fasting low-density lipoprotein (LDL) cholesterol values were calculated using the Friedewald equation unless the levels of triglycerides were >250 mg/dl, in which case, a direct LDL cholesterol determination was used. The glomerular filtration rate (GFR) was estimated with the Cockcroft-Gault equation (17).
All numeric parameters are presented as the mean ± SD or as the median (interquartile range [IQR]) for continuous variables, and as percentages for categorical variables. Demographic and traditional risk factors for cardiovascular disease were compared across quartiles of uric acid values using analysis of variance (or Kruskal-Wallis for non-normally distributed data) for continuous variables or chi-square test for discrete variables. Correlations between serum uric acid levels and other parameters were tested with Spearman's test.
Kaplan-Meier survival curve analysis was used to represent the proportional risk of death for the serum uric acid values, and the log-rank test was performed to assess differences across quartiles of serum uric acid values. The analyses were also performed separately for men and women.
A multivariate Cox proportional hazards model was performed to determine if serum uric acid was an independent predictor of death from all causes. Confounding anthropometric and biochemical factors considered for adjustment included age, sex, smoking status, alcohol consumption, diuretic use, weight, body mass index (BMI), waist circumference, systolic and diastolic blood pressure, history of either diabetes mellitus, hypertension, CAD, or stroke, the log-transformed GFR values, fasting levels of lipids (LDL cholesterol, high-density lipoprotein [HDL] cholesterol, triglycerides), and fasting glucose levels. Stepwise selection was used in development of the model. Serum uric acid was analyzed as a continuous variable. Separate models, which were adjusted for age, BMI, race, smoking status, and alcohol consumption, were generated for sex and diuretic subgroups. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) are reported, along with 2-tailed P values. Significance was set at P < 0.05.
To evaluate the contribution of using uric acid as a predictive marker, we calculated the concordance indexes (C statistic) (18) with and without uric acid in separate multivariate models that included age, sex, current smoker, HDL cholesterol level, systolic blood pressure, total cholesterol level, hypertension, diabetes mellitus, fasting glucose level, waist circumference, triglyceride level, and fibrinogen level. The C statistic is the probability that, for 2 randomly selected patients, the patient with the higher predicted mortality risk actually died. The improvement in predictability with the addition of uric acid to these well-known risk factors was assessed by the difference in the C statistics (19). The difference in the C statistics was bias-corrected, and bootstrapping was used to generate the 95% CIs. A 1-sample t-test was performed to determine if the difference was equal to zero.
All analyses were performed using SAS version 8.2 software (SAS Institute, Cary, NC).
Characteristics of the study patients at baseline.
Baseline characteristics of the patients categorized by quartiles of serum uric acid values are presented in Table 1. A total of 2,003 men and 1,095 women with a mean ± SD age of 55.4 ± 12.0 (range 18–87 years) had a uric acid determination between April 1, 1998, and July 14, 2004. At the time of the uric acid measurement, 1,322 patients (42.7%) had a history of CAD, 291 (9.4%) were previously diagnosed as having had a stroke, and 1,489 (48.1%) had hypertension. Primary prevention subjects (n = 1,641) had a 10-year estimated global risk of CAD of 9.1 ± 8.0% in men and 4.4 ± 5.2% in women (mean ± SD). A family history of premature CAD was documented in 1,621 patients (52%): 844 in the primary prevention group and 777 in the secondary prevention group. Treatment with diuretics at baseline was documented in 684 patients (22.0%): 427 were taking thiazides, 274 were taking loop diuretics, and 44 were taking other diuretics. Other medications known to influence uric acid concentrations were niacin in 197 patients (6.4%), allopurinol in 91 (2.9%), and probenecid in 5 (0.2%). Aspirin treatment was documented in 1,897 patients (61.2%) and statins in 1,273 (41.1%).
Table 1. Baseline clinical characteristics at the time of uric acid determination*
Alcohol consumption was defined as >3 alcoholic drinks per week. Categorical variables are presented as the number (%), and continuous variables are presented as the mean ± SD or as the median (interquartile range). CAD = coronary artery disease; BMI = body mass index; BP = blood pressure; GFR = glomerular filtration rate (calculated by the Cockcroft-Gault equation); LDL = low-density lipoprotein; HDL = high-density lipoprotein.
By Kruskal-Wallis test for non-normally distributed continuous variables. P values for normally distributed continuous variables were determined by analysis of variance; P values for discrete variables were determined by chi-square test.
Among the 1,095 women, 800 (73.1%) were older than 50 years. In women, but not in men, age increased across quartiles of uric acid values. Otherwise, there were no significant differences between the sexes with regard to the baseline parameters.
Serum uric acid values.
Mean ± SD uric acid levels were higher in men than in women (6.5 ± 1.5 mg/dl versus 5.4 ± 1.7 mg/dl; P < 0.001). The mean uric acid levels were higher in patients previously diagnosed as having CAD than in the rest of the study patients (6.3 ± 1.7 mg/dl versus 5.9 ± 1.6 mg/dl; P < 0.001) and in the patients with diabetes mellitus than in the rest of the study patients (6.3 ± 1.8 mg/dl versus 6.0 ± 1.6 mg/dl; P < 0.001). Uric acid levels were higher in the subgroup of 684 patients who were taking diuretics than in the rest of the patients (6.9 ± 1.9 mg/dl versus 5.9 ± 1.5 mg/dl; P < 0.001) and in patients taking thiazides than in the rest of the patients (6.7 ± 1.8 mg/dl versus 6.0 ± 1.6 mg/dl; P < 0.001).
Serum uric acid levels did not correlate with age in men, whereas in women, the correlation coefficient (r) with age was 0.26 (P < 0.001). Among the markers of metabolic syndrome (20), the serum uric acid level correlated with waist circumference (r = 0.35, P < 0.001), BMI (r = 0.27, P < 0.001), and triglyceride levels (r = 0.24, P < 0.001) and correlated inversely with HDL cholesterol (r = −0.26, P < 0.001). Uric acid levels correlated with serum creatinine levels (r = 0.31, P < 0.001) but did not correlate with the GFR (r = −0.03, P = 0.18). Uric acid levels also correlated with homocysteine levels (r = 0.34, P < 0.001) but did not correlate with levels of fibrinogen, C-reactive protein, or LDL cholesterol.
During followup, which consisted of 14,262 person-years, 156 subjects died (5%). Figure 1 shows the distribution of deaths per quartile of uric acid values. Death occurred after a median of 3.3 years from the time of uric acid determination. The median followup time in the subgroup of survivors (i.e., time from initial visit to November 5, 2005) was 4.6 years. Uric acid levels were significantly higher in the subgroup of subjects who died compared with those who survived during the study period (7.1 ± 2.1 mg/dl versus 6.0 ± 1.6; P < 0.001).
Mortality rates during followup were higher in the 684 patients taking diuretics at baseline as compared with those who were not (17.7% versus 6.9%; P < 0.001). Mortality rates were also higher in patients taking thiazides as compared with those who were not (12.3% versus 8.8%; P < 0.001).
Kaplan-Meier analysis (Figure 1) showed that although the risk of death increased gradually across the quartiles of uric acid values, only patients whose values were in quartile 4 had significantly higher mortality rates as compared with those whose values were in quartile 1, and the risk became evident at 3 years of followup. This was found to be true for both sexes when separate survival curves were built.
In the unadjusted Cox regression analysis, for every 1 mg/dl elevation in the serum uric acid level, the risk of death from all causes increased by 39% (95% CI 1.28–1.50, P < 0.001). After stepwise adjustment for age, sex, history of either diabetes mellitus, CAD, stroke, or hypertension, smoking status, alcohol consumption, diuretic use, weight, BMI, systolic and diastolic blood pressure, fasting lipid levels, fasting glucose levels, and log-transformed GFR, the serum uric acid level remained a predictor of death from all causes (HR = 1.26 [95% CI 1.15–1.38], P < 0.001) (Table 2). The attributable risk per each standard deviation of uric acid for this model was 1.47 (95% CI 1.26–1.70, P < 0.001). A separate Cox model that included serum creatinine values instead of the log-transformed GFR yielded similar results (HR = 1.31 [95% CI 1.20–1.43], P < 0.001).
Table 2. Multivariate model factors predicting all-cause mortality*
HR (95% CI)
HR = hazard ratio; 95% CI = 95% confidence interval; GFR = glomerular filtration rate.
For each increase of 1 mg/dl in the serum uric acid level, the risk of death from all causes increased by 26%.
Cox models were performed separately in patients with and those without a history of CAD. After adjusting for potential confounding factors, uric acid was a significant predictor of death in patients with a history of CAD (HR 1.35 [95% CI 1.20–1.52], P < 0.001), whereas it was not significant in the patients with no history of CAD (HR = 1.16 [95% CI 0.99–1.35], P = 0.06).
The C statistic obtained when the model included uric acid was 0.82, as compared with a C statistic of 0.79 when the model did not include uric acid. The bias-corrected difference was 0.03 (95% CI 0.013–0.056, P = 0.002). This indicates that inclusion of the uric acid value significantly improved the predictive accuracy of the model that included Framingham Heart Study score factors, metabolic syndrome components, and fibrinogen levels.
Multivariate regression was also performed separately in men and women according to the use of diuretics. The following HRs were obtained: in men using diuretics (n = 367) HR = 1.25 (95% CI 1.07–1.46, P = 0.01), in women using diuretics (n = 317) HR = 1.28 (95% CI 1.09–1.49, P = 0.02), in men not using diuretics (n = 1,636) HR = 1.22 (95% CI 1.03–1.45, P = 0.02), and in women not using diuretics (n = 779) HR = 1.61 (95% CI 1.20–2.16, P = 0.002).
In this cohort study of men and women at high risk of death from cardiovascular causes whose cases were followed for up to 7.6 years, the serum uric acid level proved to be an independent risk factor for all-cause mortality. The association was stronger in patients with a known history of CAD. The thorough phenotypic characterization of our cohort allowed us to analyze the independence of uric acid levels relative to a multitude of risk factors and comorbidities that may have an impact on uric acid levels, including traditional risk factors for CAD, markers of insulin resistance, the estimated GFR, and use of diuretics. Our study in patients at high risk of CAD events adds to the reported evidence that uric acid is a predictor of death (Figure 2) and is the first study to use C statistic analysis to establish the clinical significance of uric acid levels as a predictor of death. The C statistic appears to be a more robust way to determine whether a putative risk factor actually contributes to risk (18, 19, 21). It is especially valuable when the putative risk factor is associated with other established risk factors. For example, uric acid levels are associated with age, hypertension, and renal disease, all of which are associated with increased risk of CAD. In our cohort, the uric acid level added to the predictive value of a comprehensive multivariate model that included Framingham Heart Study score factors, components of metabolic syndrome, and markers of inflammation (C statistic 0.82 versus 0.79 when uric acid was not included in the model; P = 0.002).
From the perspective of previous studies of uric acid, almost 50% of our cohort were patients with hypertension, 55% had a family history of CAD, and 46% had CAD and/or stroke at the time of the uric acid determination. Patients are sent to our Prevention Clinic because of known CAD or for treatment of CAD risk factors. Thus, they have higher cardiovascular risk than the cohorts in the Framingham Heart Study (13) or the first National Health and Nutrition Examination Survey (NHANES-I) (8). Among the studies performed in Europe and Asia, the Progetto Ipertensione Umbria Monitoraggio Ambulatoriale (PIUMA) (11), Systolic Hypertension in the Elderly Program (SHEP) (14), and Systolic Hypertension in China (SYST-China) (12) studies exclusively included patients with hypertension. The findings of the PIUMA and SYST-China studies supported the link between uric acid levels and death from cardiovascular causes (11, 12), while the findings of the SHEP study supported only the relationship between uric acid levels and cardiovascular events (14). A review of uric acid studies since 1998 (22) showed that the findings of 10 of 11 studies performed in high-risk patients and 6 of 10 studies performed in low-risk subjects supported an independent association of uric acid levels with mortality risk.
Our data show a significant association between uric acid levels and mortality risk in both sexes. The NHANES-I study showed that serum uric acid levels predicted mortality in both men and women; the association was stronger in women than in men (8). Other studies that support the independent association of uric acid levels with mortality risk in both sexes are the SYST-China (12) and PIUMA (11) studies. In the Framingham Heart Study, the association of uric acid with mortality risk was absent in men in the univariate analysis and was absent in women after adjustment for diuretic use, blood pressure, and total cholesterol level (13). A recent prospective study showed that men with gout have increased overall risk of death and increased risk of death from cardiovascular causes compared with men without a history of gout (23).
Since the uric acid concentration increases with impaired renal function, which itself is a risk factor for mortality (24–27), we performed adjustments for the serum creatinine level and GFR in separate Cox models. The uric acid association with mortality risk remained significant. The serum creatinine level was available in 60% of participants of NHANES-I cohort and did not affect the association of uric acid levels with mortality risk (8). The Framingham Heart Study did not report the serum creatinine level or the GFR. Adjustment for serum creatinine levels was performed in the PIUMA (11), SHEP (14), and SYST-China (12) studies.
Uric acid is associated with several CAD risk factors, especially when clustered as metabolic syndrome (20). Uric acid remained a predictor of mortality risk in our study even after adjustments for history of hypertension, BMI, waist circumference, blood pressure, serum triglyceride levels, and fasting plasma glucose levels. The Kuopio Ischemic Heart Study showed similar findings after adjustment for many markers of metabolic syndrome (10). Our study and the Kuopio Ischemic Heart Study are the only uric acid cohort studies that controlled for all fractions of the lipid panel.
Treatment with diuretics is a significant confounder of serum uric acid levels. Our cohort included 684 patients who were taking diuretics at baseline, two-thirds of whom were also taking thiazides. We found significantly higher serum uric acid levels in patients who were taking diuretics at baseline as compared with patients who were not taking diuretics, which is consistent with the findings of previous studies (8, 14, 28). In our study, patients taking diuretics had higher mortality rates than did those not taking diuretics, which is similar to the findings of the Framingham Heart Study (13) and the SHEP study (14). The hypothesis that uric acid elevations in patients taking diuretics might decrease the cardiovascular benefit of the diuretics should be tested in future prospective studies. Our study controlled for the use of diuretics in a multivariate analysis, and this did not attenuate the association of uric acid levels with overall mortality risk. This is similar to findings of the NHANES-I (8) and PIUMA (11) studies. In contrast, the Framingham Heart Study showed that diuretic use was a major confounder of the relationship between serum uric acid levels and mortality risk (13).
There are some limitations of the present study. The retrospective design prevented us from controlling for medications during followup. However, controlling for changes in medications during followup was also not performed in the previous large prospective studies (8, 13). Because some patients in our study underwent only a single consultation, we did not report the cardiovascular events during followup in those patients. However, in studies showing an association between uric acid levels and overall mortality risk, the association between uric acid levels and cardiovascular mortality risk was reported to be stronger than the association between uric acid levels and all-cause mortality (8, 10–12). The majority of our patients were white, so our results cannot be readily extrapolated to other populations. However, in the NHANES-I study, the value of uric acid as a risk factor for death was more significant in the black population than the white population (8).
An association between uric acid levels and mortality risk does not imply that uric acid has a causal effect on mortality. Animal studies have shown that xanthine oxidase activity increases during vascular ischemia and reperfusion (29). In contrast, uric acid has proinflammatory and proliferative effects on vascular smooth muscle (5, 6) and inhibits both basal and vascular endothelial growth factor–induced production of nitric oxide by endothelial cells. The latter effect was reversed by lowering the uric acid levels with allopurinol therapy (30). The link between hyperuricemia and impaired nitric oxide generation might provide a pathogenetic basis for the link between uric acid and cardiovascular disease. The identification of the uric acid transporter URAT1 (31, 32) has created new potential molecular targets for directly manipulating serum uric acid levels and for dissecting whether hyperuricemia has direct detrimental effects on the vascular system (31, 32).
Whether a reduction of uric acid levels would have an impact on survival is still to be determined. Effective medications that lower uric acid levels have been used for many years in clinical practice to treat patients with gout. Some studies have suggested that reduction of uric acid levels has a favorable effect on surrogate markers of atherosclerosis and cardiovascular events. Allopurinol, an inhibitor of xanthine oxidase, was shown to improve flow-mediated dilatation after 3 months of therapy in patients with hyperuricemia at high risk of cardiovascular events (33). The angiotensin receptor blocker losartan has a uricosuric effect, and the contribution of uric acid reduction to the effect of losartan on the primary end point in the Losartan Intervention For Endpoint reduction in hypertension (LIFE) study was 29% (34). Among the statins, atorvastatin was shown to decrease uric acid levels via a uricosuric effect (35).
In conclusion, based on our data and data from previous studies (8–12), the serum uric acid level is an independent predictor of mortality in populations at high risk of CAD. Measurement of serum uric acid levels contributes to the risk stratification beyond the CAD scores that are currently used. Routine measurement of serum uric acid levels may therefore be useful in patients at high risk of CAD. Further analyses from large interventional studies are needed to clarify whether therapies that lower serum uric acid levels may benefit patients at high risk for cardiovascular events.
Dr. Hoogwerf had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Ioachimescu, Hazen, Hoogwerf.
Acquisition of data. Ioachimescu, Hoar, Hazen, Hoogwerf.
Analysis and interpretation of data. Ioachimescu, Brennan, Hazen, Hoogwerf.