Uric acid and prognosis in angiography-proven coronary artery disease

Authors


Correspondence to: Gjin Ndrepepa, MD, Deutsches Herzzentrum, Lazarettstrasse 36, 80636 München, Germany. Tel.: +49 89 12181535; fax: +49 89 12184053; e-mail: ndrepepa@dhm.mhn.de

Abstract

Background

The optimal uric acid (UA) level associated with the lowest mortality and the strength of association between UA and mortality in various subgroups of patients with coronary artery disease (CAD) are unknown.

Materials and methods

This study included 13 273 patients with angiographic confirmation of CAD and UA measurements available. The primary outcome analysis was 1-year mortality.

Results

Based on the receiver operating characteristic curve analysis, the best cut-off of UA for mortality prediction was 7·11 mg/dL. Using this cut-off, patients were divided into two groups: the group with UA ≤ 7·11 mg/dL (n = 9075) and the group with UA > 7·11 mg/dL (n = 4198). Cardiac mortality was 6·3% (256 deaths) in patients with UA > 7·11 mg/dL and 2·3% (201 deaths) in patients with UA ≤ 7·11 mg/dL [hazard ratio (HR) = 2·82, 95% confidence interval (CI) 2·36–3·36; P < 0·001]. After adjustment for cardiovascular risk factors, UA remained an independent correlate of cardiac mortality (HR = 1·20, 95% CI 1·08–1·34; P = 0·001, for each standard deviation increase in the logarithmic scale of UA). The relationship between cardiac or all-cause mortality and UA showed a J-shaped pattern with lowest mortality in patients with UA between 5·17 and 6·76 mg/dL. UA predicted mortality across all subgroups of patients, with strongest association in women and patients without arterial hypertension.

Conclusions

UA predicted an increased risk of cardiac mortality across all subgroups of patients with CAD. The association between UA and cardiac or all-cause mortality had a ‘J-shaped’ pattern with lowest risk of death in patients with UA levels between 5·17 and < 6·76 mg/dL.

Introduction

Uric acid (UA) – an end product of purine metabolism – has been implicated in the genesis of cardiovascular disease since 1879 [1], and more than 60 years ago, it was included in the protocols of Framingham study to be investigated as a potential risk factor for cardiovascular disease [2]. However, the nature of the relationship between UA and cardiovascular disease remains a subject of debate. Difficulties in determining whether UA should be considered a cardiovascular risk factor may be explained by its frequent association with other cardiovascular risk factors [3] for which UA is considered as a risk marker or epiphenomenon or even an adaptive change to protect from atherosclerosis due to its antioxidant properties [4] and the controversial and conflicting findings from epidemiological studies [3, 5, 6].

Apart from disputable causality, other aspects of the association between UA and cardiovascular disease remain poorly understood. Although a threshold effect in the association between UA and mortality has been described [7, 8], the best cut-offs of UA for prediction of mortality remain unknown. A ‘J-shaped’ association pattern between UA level and mortality (with an increase in mortality for lower and higher UA levels) has been described in population-based studies [9] and in various disease states [10-12]. However, the pattern of association between UA and mortality and the most optimal UA levels associated with lowest mortality in patients with coronary artery disease (CAD) remain unknown, even though a stronger association between UA and mortality in patients with a CAD history has been reported [13, 14]. Moreover, conflicting results have been reported regarding the association between UA and mortality in subgroups according to sex [5, 15], diabetes [16, 17] and kidney disease [18]. Despite data suggesting that elevated UA levels cause endothelial dysfunction [19] and may promote atherosclerosis and vascular thrombosis [20, 21], evidence regarding the association between UA and myocardial infarction remains controversial [21, 22]. We undertook this study in a large series of patients with CAD with a double objective: first, to assess the pattern of association between UA level and cardiac or all-cause mortality and find the most optimal UA levels associated with lowest mortality in patients with CAD; and second, to assess the strength of association between UA and mortality in various subgroups of patients with CAD.

Methods

Patients

Between March 2000 and December 2009, 14 713 consecutive patients with CAD underwent diagnostic angiography and percutaneous coronary intervention (PCI) in the Deutsches Herzzentrum in Munich. Those eligible for the study were patients with the clinical diagnosis of stable CAD or acute coronary syndromes (ACS) in whom the presence of significant CAD was confirmed by coronary angiography. Patients with no UA measurements (832 patients), acute infections (130 patients), renal disease (serum creatinine level ≥ 2 mg/dL; 312 patients) and known malignancies with life expectancy < 1 year (166 patients) were excluded. Thus, the present study included 13 273 patients with angiographic confirmation of significant CAD and UA measurements available. Of these, 8149 patients had stable CAD, and 5124 patients had ACS (unstable angina in 2163 patients, acute non-ST-segment elevation myocardial infarction in 1332 patients and ST-segment elevation myocardial infarction in 1629 patients). Detailed inclusion/exclusion of the patients and the diagnostic criteria of CAD are provided in prior publications from our group [8, 23]. Patients gave written informed consent before recruitment in the study. The study was carried out in accordance with the Declaration of Helsinki and was approved by the institutional ethics committee. Reporting of the study conforms to STROBE along with references to STROBE and the broader EQUATOR guidelines [24].

Angiographic evaluation and definitions

Digital angiograms were analysed offline with an automated edge-detection system (CMS; Medis Medical Imaging Systems, Nuenen, The Netherlands) in the core angiographic laboratory. Significant CAD was diagnosed in the presence of coronary stenoses ≥ 50% lumen obstruction in, at least, one of the three major coronary arteries. The complexity of lesions was defined according to the modified American College of Cardiology/American Heart Association grading system [25]. Class B2 and C lesions were considered complex. Left ventricular ejection fraction was calculated with area–length method using left ventricular angiograms [26]. Arterial hypertension was considered to be present when a patient was receiving active treatment with antihypertensive drugs or if, on two separate occasions, the systolic blood pressure was 140 mmHg or greater or the diastolic blood pressure 90 mmHg or greater. Hypercholesterolaemia was defined as a documented total cholesterol value ≥ 220 mg/dL or prior or ongoing treatment with a lipid-lowering agent. Smokers were defined as those currently smoking any tobacco. The requirements for a diagnosis of diabetes were as follows: active treatment with insulin or an oral hypoglycaemic agent on admission; documentation of an abnormal fasting blood glucose (> 125 mg/dL); blood glucose > 200 mg/dL at any time; or abnormal glucose tolerance test based on the World Health Organization criteria. Patients' weight and height were measured, and body mass index was calculated. The glomerular filtration rate was estimated using the Cockcroft–Gault formula [27].

Stent implantation and periprocedural care were performed according to standard criteria. Antiplatelet therapy consisted of clopidogrel (300 or 600 mg as a loading dose followed by 75 mg/day for at least 4 weeks) and aspirin (200 mg/day administered orally and continued indefinitely).

Outcome definition and follow-up

The primary outcome analyses were 1-year cardiac and all-cause mortality. Secondary outcomes included vascular events such as nonfatal myocardial infarction and stroke. The follow-up protocol after discharge consisted of a phone interview at 1 month, a visit at 6 months and a phone interview at 12 months. The diagnosis of myocardial infarction was based on the development of new abnormal Q waves in ≥ 2 contiguous precordial or ≥ 2 adjacent limb leads, or an elevation of creatine kinase – myocardial band (CK-MB) > 2 times (> 3 times for the 48 h after a PCI procedure) the upper limit of normal. Stroke required confirmation by computed tomography or magnetic resonance imaging of the head. Information about death was obtained from hospital records, death certificates or phone contact with relatives of the patient or referring physician(s). Cardiac death was defined according to Academic Research Consortium criteria and included any death due to proximate cardiac cause (e.g., myocardial infarction, low-output failure, faetal arrhythmia), unwitnessed death and death of unknown cause, and all procedure-related deaths, including those related to concomitant treatment [28]. Patients reporting cardiac complaints during follow-up underwent a complete clinical, electrocardiographic and laboratory evaluation. Collection of baseline characteristics of the patients, follow-up information and adjudication of adverse events was performed by medical staff unaware of UA level.

Biochemical measurements

Detailed description of laboratory measurements has been previously shown [8, 23]. Blood samples were obtained before angiography in all patients. Blood was collected into tubes containing lithium heparin as anticoagulant (S-Monovette 4·9 mL; Sarstedt, Nümbrecht, Germany). The blood samples were immediately transported to the laboratory, centrifuged and analysed (median turn-around time 30–45 min, 98% of results within 90 min). The UA concentration in plasma was determined with an enzymatic colorimetric test on a Cobas Integra 800 analyzer (Roche Diagnostics, Mannheim, Germany). The measuring range in plasma is 0·20–25 mg/dL (11·9–1500 μM). Lower detection limit of the test is 0·20 mg/dL (11·9 μM). The reference range for men is 3·4–7·0 mg/dL (202·3–416·5 μM) and for women 2·4–5·7 mg/dL (142·8–339·2 μM). Plasma concentrations of high-sensitivity C-reactive protein (CRP) were measured using a fully automated latex-enhanced immunoturbidimetric assay on a Cobas Integra (Roche Diagnostics). The CRP assay has an analytical sensitivity of 0·085 mg/L and a measuring range up to 160 mg/L. The upper limit of the reference range in healthy adults is 5 mg/L. Creatinine was measured using a kinetic colorimetric assay based on the compensated Jaffe method. Laboratory personnel were unaware of clinical, angiographic or follow-up data of the patients.

Statistical analysis

Data are presented as median (interquartile range), counts or proportions (%). One-sample Kolmogorov–Smirnov test was used to assess the data distribution. Continuous data with skewed distribution (shown as medians with interquartile range) were compared with the Kruskal–Wallis rank sum test. Categorical data were compared with chi-square test. Receiver operating characteristic (ROC) curve was constructed to assess the best UA cut-off value regarding the prediction of cardiac mortality whilst maximizing sensitivity and specificity through minimizing the square root of (1 − sensitivity)2 + (1 − specificity)2. Survival analysis was performed by applying the Kaplan–Meier method and log-rank test. Multivariable Cox proportional hazards model was used to assess the association between UA and cardiac or all-cause mortality whilst adjusting for cardiovascular risk factors and other relevant clinical variables. The following variables were entered into the model: age, sex, body mass index, diabetes, arterial hypertension, hypercholesterolaemia, smoking status, previous myocardial infarction, previous coronary artery bypass surgery, clinical presentation (stable CAD vs. ACS), glomerular filtration rate, CRP, left ventricular ejection fraction, multivessel disease and UA. UA was entered into the model as a continuous variable after logarithmic transformation. Area under the ROC curve showing the performance of UA to predict cardiac mortality was calculated after covariate adjustment. The same variables as for the Cox model were entered into the model. The discriminatory power of the model regarding mortality was assessed by calculating the integrated discrimination improvement (IDI) according to Pencina et al. [29] Differences in the association between UA and mortality across various subgroups of patients were investigated by performing interaction testing. All analyses were performed using S-plus statistical package (S-PLUS; Insightful Corp, Seattle, WA, USA). A two-tailed P < 0·05 was considered to indicate statistical significance.

Results

Characteristics of patients

The ROC curve analysis of UA predicted 1-year cardiac mortality with an unadjusted area under the curve of 0·650 with 95% confidence interval (CI) between 0·616 and 0·675 (P < 0·001; Fig. 1). The best cut-off value of UA regarding mortality prediction was 7·11 mg/dL. Using this cut-off, patients were divided into two groups: group with UA level ≤ 7·11 mg/dL (n = 9075 patients) and group with UA level > 7·11 mg/dL (n = 4198 patients). With the exception of the proportions of patients with arterial hypertension, current smoking, hypercholesterolaemia or complex lesions and the type of therapy, all other characteristics differed significantly amongst patients in both groups (Table 1).

Table 1. Baseline characteristics
CharacteristicUric acid (UA)
≤ 7·11 mg/dL (n = 9075)> 7·11 mg/dL (n = 4198)P value
  1. Data are number of patients (%) or median [25th; 75th percentiles].

  2. Data on body mass index, glomerular filtration rate, C-reactive protein and left ventricular ejection fraction were available in 99·7% (n = 13 237), 99·7% (n = 13 238), 99·2% (n = 13 170) and 87·9% (n = 11 664) of the patients, respectively. The remaining data are complete.

  3. a

    Refers to patients with implanted stents only.

Age (years)67·0 [59·2; 74·3]68·4 [60·4;75·6]< 0·001
Females2389 (26·3)714 (17·0)< 0·001
Diabetes2409 (26·5)1296 (30·9)< 0·001
Insulin requiring751 (8·3)436 (10·4)< 0·001
Body mass index (kg/m2)26·5 [24·3; 29·1]27·8 [25·6; 30·6]< 0·001
Arterial hypertension6269 (69·1)2844 (67·7)0·123
Current smoking1450 (16·0)642 (15·3)0·314
Hypercholesterolaemia6387 (70·4)2904 (69·2)0·159
Previous myocardial infarction2751 (30·3)1458 (34·7)< 0·001
Previous coronary artery bypass surgery1297 (14·3)698 (16·6)< 0·001
Clinical presentation
Stable coronary artery disease5541 (61·1)2608 (62·1)0·011
Acute coronary syndrome3534 (38·9)1590 (37·9)
High-sensitivity C-reactive protein (mg/L)2·1 [0·9; 6·0]3·2 [1·3; 8·6]< 0·001
Creatinine (mg/dL)0·9 [0·8; 1·1]1·1 [0·9; 1·3]< 0·001
Glomerular filtration rate (mL/min)82·8 [64·1; 103·9]72·1 [52·4; 95·8]< 0·001
UA (mg/dL)5·8 [5·0; 6·4]8·1 [7·6; 9·0]< 0·001
No. of coronary arteries narrowed
11690 (18·6)625 (14·9)< 0·001
22489 (27·4)1031 (24·6)
34896 (54·0)2542 (60·5)
Multivessel disease7385 (81·4)3573 (85·1)< 0·001
Complex lesions6869 (75·7)3191 (76·0)0·688
Left ventricular ejection fraction (%)58·0 [49·0; 63·0]54·0 [43·0; 61·0]< 0·001
Vessel treated
Left main coronary artery390 (4·3)197 (4·7)< 0·001
Left descendent coronary artery3640 (40·1)1563 (37·2)
Left circumflex coronary artery2189 (24·1)1010 (24·1)
Right coronary artery2518 (27·8)1197 (28·5)
Bypass graft338 (3·7)231 (5·5)
Type of intervention  0·461
Coronary stenting8221 (90·6)3786 (90·2) 
Balloon angioplasty854 (9·4)412 (9·8) 
Drug-eluting stents6239 (75·9)a2822 (74·5)a0·109
Figure 1.

Receiver operating characteristic (ROC) curve showing the accuracy of uric acid to predict cardiac mortality. Adjusted area under the ROC curve was 0·666, 95% confidence interval 0·634–0·698; P < 0·001.

Clinical outcome

Overall, there were 646 deaths (4·9%) within the first year following PCI. Of these, 457 deaths (3·4% of patients or 70·7% of deaths) were of cardiac origin. Deaths of cardiac origin occurred in 256 patients with UA level > 7·11 mg/dL and in 201 patients with UA level ≤ 7·11 mg/dL [Kaplan–Meier estimates of mortality 6·3% and 2·3%, respectively; hazard ratio (HR) = 2·82, 95% CI 2·36–3·36; P < 0·001; Fig. 2a]. All-cause deaths occurred in 345 patients with UA level > 7·11 mg/dL and in 301 patients with UA level ≤ 7·11 mg/dL (Kaplan–Meier estimates of mortality 8·4% and 3·6%, respectively; HR = 2·54, 95% CI 2·19–2·95; P < 0·001; Fig. 2b). Nonfatal myocardial infarction occurred in 139 patients with UA level > 7·11 mg/dL and in 281 patients with UA level ≤ 7·11 mg/dL (Kaplan–Meier estimates, 3·4% and 3·1%, respectively; HR = 1·08, 95% CI 0·88–1·32; P = 0·449). Stroke occurred in 35 patients with UA level > 7·11 mg/dL and in 59 patients with UA level ≤ 7·11 mg/dL (Kaplan–Meier estimates, 0·9% and 0·7%, respectively; HR = 1·30, 95% CI 0·86–1·98; P = 0·806).

Figure 2.

Kaplan–Meier curves of cardiac (a) and all-cause (b) mortality according to UA level > or ≤ 7·11 mg/dL. UA, uric acid.

Distribution pattern of mortality according to UA level

Cardiac and all-cause mortality was assessed according to deciles of UA concentration (Table 2 and Fig. 3). For UA levels between 5·17 and < 6·76 mg/dL (3rd to 6th deciles), cardiac and all-cause mortality was lowest [1·9% (100 deaths) and 3·0% (160 deaths), respectively]. For patients with UA levels ≤ 5·17 mg/dL (the first 2 deciles), the risk of cardiac (76 deaths, 2·9%) or all-cause (108 deaths, 4·1%) mortality was significantly higher than for patients in the 3rd to 6th deciles [odds ratio (OR) = 1·53, 95% CI 1·13–2·08; P = 0·005 for cardiac mortality and OR = 1·36 (1·06–1·75); P = 0·013 for all-cause mortality]. For patients with UA levels > 6·76 mg/dL (7th to 10th deciles), the risk of cardiac (281 deaths, 5·3%) or all-cause (378 deaths, 7·1%) was significantly higher than for patients in the 3rd to 6th deciles [OR = 2·90 (2·31–3·66); P < 0·001 for cardiac mortality and OR = 1·80 (1·45–2·24); P < 0·001 for all-cause mortality]. Thus, the relationship between mortality and UA level in patients with CAD showed a J-shaped pattern (Fig. 3).

Table 2. Cardiac and all-cause mortality according to uric acid (UA) deciles
Deciles of UAUA level (mg/dL)Number of patientsMortality (number of patients; %)
CardiacAll-cause
  1. Data are range(s) or number of patients (%).

1st1·30 to < 4·52132430 (2·27)48 (3·63)
2nd4·52 to < 5·17132846 (4·46)60 (4·52)
3rd5·17 to < 5·60126415 (1·19)31 (2·45)
4th5·60 to < 6·00137527 (1·96)40 (2·91)
5th6·00 to < 6·37132135 (2·65)51 (3·86)
6th6·37 to < 6·76134823 (1·71)38 (2·82)
7th6·76 to < 7·20124429 (2·33)38 (3·05)
8th7·20 to < 7·80137550 (3·64)70 (5·09)
9th7·80 to < 8·76136470 (5·13)88 (6·45)
10th8·76–21·901330132 (9·92)182 (13·68)
Figure 3.

Cardiac and all-cause mortality according to deciles of uric acid.

UA and adjusted risk of mortality

The Cox proportional hazards model was used to test the association between UA level and mortality whilst adjusting for potential confounders (see 'Methods' for variables entered into the model). Results of multivariable analysis are shown in Table 3. After adjustment for traditional cardiovascular risk factors and relevant clinical variables, including renal function (glomerular filtration rate) and inflammation status (CRP), UA remained an independent correlate of cardiac mortality (adjusted HR = 1·20, 95% CI 1·08–1·34; P = 0·001) and all-cause mortality (adjusted HR = 1·19, 95% CI 1·08–1·30; P < 0·001) for each standard deviation increase in the logarithmic scale of UA. Adjusted risk of mortality was also calculated for each 1 mg/dL increase in the UA level. For each 1 mg/dL increase in the UA level, the adjusted risk of cardiac and all-cause mortality was increased by 10% (adjusted HR = 1·10, 95% CI 1·04–1·16; P < 0·001) and 11% (adjusted HR = 1·11, 95% CI 1·06–1·16; P < 0·001), respectively. The inclusion of UA in the multivariable model was associated with a trend for an improvement of the discriminatory power of the model regarding prediction of cardiac mortality (absolute IDI = 0·004, relative IDI = 2·5%; P = 0·074) and a significant improvement of discriminatory power of the model regarding all-cause mortality (absolute IDI = 0·005, relative IDI = 3·0%; P = 0·022).

Table 3. Results of multivariable Cox proportional hazards model regarding cardiac and all-cause mortality
CharacteristicCardiac mortalityAll-cause mortality
HR [95% CI]P valueHR [95% CI]P value
  1. ACS, acute coronary syndrome; CAD, coronary artery disease; CI, confidence interval; SD, standard deviation; HR, hazard ratio.

Uric acid (for each SD increase in the natural logarithm)1·20 [1·08–1·34]0·0011·19 [1·08–1·30]< 0·001
Age (for 10-year increase)1·12 [0·96–1·29]0·1471·20 [1·05–1·31]0·006
Female sex1·14 [0·89–1·47]0·2981·14 [0·93–1·42]0·212
Body mass index (for 5 kg/m2 increase)1·11 [0·96–1·29]0·1681·06 [0·93–1·20]0·412
Diabetes1·60 [1·28–2·01]< 0·0011·61 [1·33–1·94]< 0·001
Arterial hypertension0·55 [0·44–0·70]< 0·0010·63 [0·52–0·77]< 0·001
Hypercholesterolaemia0·79 [0·63–1·00]0·0500·77 [0·63–0·94]0·009
Current smoking0·93 [0·66–1·31]0·6811·05 [0·79–1·40]0·724
Previous myocardial infarction0·83 [0·65–1·06]0·1440·89 [0·73–1·10]0·274
Previous coronary artery bypass surgery0·75 [0·53–1·05]0·1000·89 [0·68–1·16]0·379
Clinical presentation (ACS vs. stable CAD)2·25 [1·75–2·90]< 0·0012·00 [1·63–2·46]< 0·001
Glomerular filtration rate (for 30 mL/min decrease)2·04 [1·66–2·50]< 0·0012·00 [1·66–2·38]< 0·001
C-reactive protein (for 5 mg/L increase)1·02 [1·01–1·03]< 0·0011·02 [1·01–1·03]< 0·001
Left ventricular ejection fraction (for 10% decrease)1·61 [1·48–1·75]< 0·0011·52 [1·41–1·62]< 0·001
Multivessel disease (vs. single-vessel disease)1·87 [1·27–2·77]0·0021·71 [1·24–2·27]0·001

UA and mortality in various subgroups of patients

The association between UA and cardiac or all-cause mortality was assessed in various subgroups of patients. For this analysis, patient subgroups were obtained by dividing patients according to age (cut-off, 65 years), sex, diabetes, arterial hypertension, glomerular filtration rate (cut-off, 60 mL/min), body mass index (cut-off, 30 kg/m2), left ventricular ejection fraction (cut-off, 50%) and CRP (cut-off, 5 mg/L). This analysis showed that UA predicted an increased risk of cardiac and all-cause mortality across all subgroups. An interaction was observed between UA and sex – demonstrating a stronger association with mortality in women than in men (P for interaction = 0·059 for cardiac mortality and P for interaction = 0·019 for all-cause mortality) – and between UA and arterial hypertension – demonstrating a stronger association between UA and mortality in patients without arterial hypertension than in patients with arterial hypertension (P for interaction = 0·050 for cardiac mortality and P for interaction = 0·064 for all-cause mortality). The results of subgroup analyses are shown in Figs 4 and 5.

Figure 4.

Cardiac mortality in various subgroups of patients with CAD. ACS, acute coronary syndrome; BMI, body mass index; CAD, coronary artery disease; CRP, C-reactive protein; GFR, glomerular filtration rate; LVEF, left ventricular ejection fraction; No., number; Pint, P for interaction.

Figure 5.

All-cause mortality in various subgroups of patients with CAD. ACS, acute coronary syndrome; BMI, body mass index; CAD, coronary artery disease; CRP, C-reactive protein; GFR, glomerular filtration rate; LVEF, left ventricular ejection fraction; No., number; Pint, P for interaction.

Discussion

In this study, we investigated the pattern of association between UA and mortality (cardiac and all-cause) in a large series of patients with CAD and the strength of this association in various subgroups of patients. The main findings of the study can be summarized as follows: (i) UA predicted an increased risk of cardiac and all-cause mortality independent of traditional cardiovascular risk factors, left ventricular function, CRP and renal function across the whole spectrum of patients with CAD. (ii) The association between UA and cardiac or all-cause mortality followed a ‘J-shaped’ pattern with a significant increase in the risk of death in patients with the lowest and highest UA levels. UA levels between 5·17 and < 6·76 mg/dL were associated with the lowest risk of mortality (both cardiac and all-cause). There was a significant increase in the risk of death for UA levels either < 5·17 mg/dL or higher than 6·76 mg/dL. (iii) UA predicted an increased risk of cardiac and all-cause mortality across all subsets of patients defined by age (younger or older than 65 years), sex, diabetes, arterial hypertension, obesity, CAD presentation, renal function, left ventricular systolic function or CRP level. However, the strongest association between UA and mortality was observed in women and patients without arterial hypertension.

Prior studies in patients with known CAD have shown that UA is an independent correlate of mortality [30, 31]. Of note, it has also been suggested that the presence of CAD strengthens the association between UA and mortality [13, 14]. The present study is the largest one to provide support for an independent association between UA and both cardiac and all-cause mortality in patients with known CAD. We also showed that the association between UA and mortality has a J-shaped pattern with higher mortality rates for patients with high and low UA levels. Based on our analysis, 60% of patients with CAD had suboptimal UA levels (the first two deciles too low and the upper four deciles too high) in terms of association with mortality. The association between elevated levels of UA and mortality from cardiac disease has been the subject of intense research, and a host of putative mechanisms to explain it has been proposed and recently reviewed [32]. In brief, an association with a more adverse cardiovascular risk profile and the possibility that UA per se may promote atherosclerosis or vascular events leading to cardiovascular deaths have been proposed to explain increased risk of death associated with elevated UA levels. Evidence available suggests that UA may have atherogenic actions. Thus, it has been suggested that UA causes endothelial dysfunction [19, 33], proliferation of vascular smooth muscle cells [34], and it promotes inflammation in vascular tissue [33], acts as pro-oxidant in atherosclerotic environment [35], and enhances lipid peroxidation [36]. UA and xanthine oxidase have been identified in atherosclerotic plaques and involved in the aetiology of atherosclerosis [37]. However, the Hill's criteria of causality for UA as a risk factor for cardiovascular disease [38] are only partially fulfilled at best [39]. Reasons for an association between low UA levels and mortality are unknown. In a recent study of patients on hemodialysis by Lee et al. [40], an UA level < 5·2 mg/dL (very close to our cut-off point of 5·17 mg/dL) was associated with a > 2-fold increased risk of death in the first year of dialysis. Low UA levels were also associated with more comorbid conditions and markers of protein–energy wasting, suggesting a higher oxidative stress due to reduced antioxidant combat mechanisms conferred by low UA level. These factors were proposed as putative mechanisms for increased risk of death in these patients [40]. Other studies have implicated malnutrition as a causative factor for low UA level, which also affects mortality [41]. Reduced UA levels have been linked to several neurological diseases such as multiple sclerosis, Parkinson's disease, Alzheimer's disease and optic neuritis [42]. Although the risk conferred by these diseases in our patients remains unaccounted for, any substantial contribution in mortality by these diseases would seem unlikely. Moreover, because patients with impaired renal function (serum creatinine level ≥ 2 mg/dL) were excluded from the study, mechanisms that may be operative in a patient population with renal failure and dialysis may not be applicable in explaining the increased risk of death with low UA levels in our patients with CAD.

Subgroup analyses performed in the setting of present study showed that UA level predicts an increased risk of cardiac and all-cause mortality across all subsets of patients with CAD. We believe that these findings are novel and important in the light of existing controversy regarding the association between UA and mortality in various diseases [16-18]. Although, UA level predicted mortality across all subgroups, differences in the strength of association were observed. Specifically, a stronger association of UA with mortality was observed in women and patients without arterial hypertension. Whilst the association between UA and mortality was observed in both sexes, the finding of a stronger association in women in our study is in line with prior studies showing a stronger association of UA with mortality in women than in men [15]. A recent meta-analysis by Kim et al. [43] showed that association between hyperuricemia and CAD incidence and mortality was significant in women but not in men. Moreover, in recent studies by Strasak et al. [44, 45], a higher mortality due to CAD was observed only in women. To our knowledge, finding of a stronger association between UA and cardiac mortality in patients without arterial hypertension is novel. In the Rotterdam study, the association between UA and stroke was weakened by the presence of arterial hypertension [21]. The magnitude of differences in the risk estimates (HRs) in subgroup analyses (Figs 4 and 5) provides a signal for a stronger association between UA and mortality in the absence of cardiovascular risk factors. Specifically, trends for a stronger association of UA with cardiac mortality were observed in patients without arterial hypertension, diabetes, obesity and in those with low CRP. Although, this finding may be a signal for causality in the relationship between UA and cardiovascular disease, which may be more evident in the absence of masking effects of risk factors, further studies are needed to confirm this hypothesis.

The present study may have implications. In particular, this study may strengthen the need for interventional studies with UA-lowering therapies to maintain UA levels in the range associated with the lowest mortality. Although current guidelines recommend an UA level ≤ 6 mg/dL [46], these target concentrations are reached in < 50% of the patients [47]. Moreover, allopurinol – a xanthine oxidase inhibitor – has been shown recently to significantly improve endothelial function and abolish vascular oxidative stress [48], has a clinically relevant anti-ischaemic effect and has been well tolerated in patients with angina [49]. In analogy with homocysteine-lowering therapy, interventional studies with UA-lowering agents, apart from exploring clinical benefits of these agents, may provide valuable additional information regarding causality in the UA–cardiovascular disease relationship.

We recognize several limitations of the present study. First, a follow-up longer than 1 year would be desirable. However, the curves of mortality were diverging, and the difference in mortality rates expanding throughout the first year suggests that even greater differences in mortality would be expected in the years to come. Second, when analysing the results of subgroup analyses, the impact of errors introduced by multiple testing should be considered. However, all subgroup comparisons of UA–mortality association had P < 0·001, which almost excludes the possibility of chance findings. Although, in general, subgroup analyses suffer from limited numbers of subjects/patients due to division into subgroups, large number of patients in the present study allowed us to perform subgroup analyses with large numbers of patients per subgroup. Third, the study has an observational design and is based on a single UA measurement. Thus, eventual changes in UA level during the follow-up remain unaccounted for. Finally, patients with advanced renal disease and those with malignancies with life expectancy < 1 year are not included in this analysis.

In conclusion, UA level predicted an increased risk of cardiac and all-cause mortality across all subsets of patients with CAD with the strongest association observed in women and patients without arterial hypertension. The association between UA level and cardiac and all-cause mortality followed a ‘J-shaped’ pattern with a significant increase in mortality for UA levels < 5·17 and > 6·76 mg/dL. Specifically designed interventional studies are needed to confirm these findings.

Author contribution

Drs Ndrepepa and Kastrati involved in the conception and design of the study. Drs Ndrepepa, Braun, King, Fusaro, Tada, Cassese, Hadamitzky and Haase involved in the acquisition of the data, analysed and interpreted the data. Drs Ndrepepa, Kastrati and Schömig involved in the writing and critiquing of drafts of the manuscript. Drs Ndrepepa, Braun, King, Fusaro, Tada, Cassese, Hadamitzky, Haase, Schömig and Kastrati approved the final manuscript for submission.

Conflicts of interest

None.

Sources of funding

None.

Address

Deutsches Herzzentrum, Technische Universität, Lazarettstrasse 36, 80636 München, Germany (G. Ndrepepa, S. Braun, L. King, M. Fusaro, T. Tada, S. Cassese, M. Hadamitzky, A. Schömig, A. Kastrati); 1. Medizinische Klinik rechts der Isar, Technische Universität, Munich, Germany (H.-U. Haase, A. Schömig).

Ancillary