On-admission serum uric acid predicts outcomes after acute myocardial infarction: systematic review and meta-analysis of prognostic studies

Aim To evaluate the prognostic value of serum uric acid (SUA) in acute myocardial infarction (AMI) patients. Methods Systematic review and random-effects meta-analysis of prognostic studies assessing AMI outcomes (death, major adverse cardiac events, MACE) in relation to on-admission SUA. Results Nine studies (7655 patients) were identified, 6 in the ST-segment elevation AMI patients treated with invasive revascularization and three in mixed AMI type cohorts with variable reperfusion strategies. “High” SUA (vs “low,” different cut-offs) was univariately associated with higher short-term mortality (8 studies/6805 patients; odds ratio [OR], 3.24; 95% confidence interval [CI], 2.47-4.27) and incidence of MACE (7/6467; OR, 2.46; 95% CI, 1.84-3.27, moderate heterogeneity, mild bias), and with higher medium-term mortality (5/5194; OR, 2.69; 95% CI, 2.00-3.62, moderate heterogeneity, mild bias) and MACE (4/4299; OR, 1.93; 95% CI, 1.36-2.74, high heterogeneity, mild bias). It was independently associated with a higher short-term (4/3625; OR, 2.26, 95% CI, 1.85-2.77) and medium/long-term (3/2683; hazard ratio [HR], 1.30; 95% CI 1.01-1.68, moderate heterogeneity, mild bias) occurrence of poor outcomes (death/MACE). As a continuous variable (by 50 μmol/L), higher SUA was also independently associated with poorer medium/long-term outcomes (4/3533; HR, 1.19; 95% CI, 1.03-1.37, high heterogeneity, mild bias). All individual study effects (unadjusted or adjusted) were in the same direction, but differed in size. Heterogeneity was mainly due to the included AMI type and/or definition of MACE. All bias-corrected pooled effects remained significant. Conclusion Based on the available data, high(er) on-admission SUA independently predicts worse short-term and medium/long-term outcomes after AMI. However, the number of data are modest and additional prospective studies are warranted.

In humans, uric acid is the end product of purine metabolism due to a genetically determined lack of uricase activity (1). It is generated by oxidation of xanthine, primarily in the liver, gut, kidneys, and apparently in the heart, but xanthine oxidase (XO) is a ubiquitous enzyme (2). Serum levels of uric acid (serum uric acid, SUA) are governed by its production and elimination rates (via the kidney). Concentrations >420 μmol/L in men and >360 μmol/L in women are conventionally considered to represent hyperuricemia, values <310 μmol/L and <250 μmol/L, respectively, are considered low-normal, whereas concentrations in-between these limits are considered high-normal (3). High purine intake (eg, animal foods, herring, anchovies, alcohol, fructose, sweetbreads) and a number of morbidity (reduced renal function, conditions with a high cellular turnover), pharmacological (eg, diuretics) and genetically determined factors (eg, urate transporter or organic anion transporter mutations) may contribute to development of high SUA (4).
Uric acid has several effects of potential interest in cardiovascular diseases (CVD). It is a potent antioxidant but can also promote oxidative stress, particularly at high concentrations and/or in surroundings with a low pH and/or low levels of other antioxidants (3)(4)(5). Furthermore, in vitro, it has several effects on the vascular smooth muscle and mononuclear cells that are considered important in pathophysiology of CVD (5). Consequently, high(er) SUA has been extensively evaluated as a prognostic factor for different CVDs (3)(4)(5). However, increased SUA is linked to various conditions that per se are CVD risk factors (eg, hypertension, dyslipidemia, diabetes, metabolic syndrome, renal failure) and it has not been always possible to distinguish whether it is a cause or a consequence of such conditions (3)(4)(5). Next, in a failing heart or a hypoxic heart, activation of XO occurs (2,6). This inevitably results in increased SUA, but XO per se promotes oxidative stress and endothelial dysfunction (2). Inhibition of XO re-establishes endothelial function, whereas lowering of SUA by uricosuric agents does not seem to achieve this effect (2). Hence, the role of SUA in CVD has been accompanied by a controversy: should it be viewed as a "true" risk factor (ie, a "direct pathogen") or as a mere marker of conditions that actually are the risk factors (6).
Considering coronary artery disease, a recent meta-analysis of 26 large prospective cohort studies indicated an independent association between hyperuricemia and occurrence of the disease and related mortality (7). Less is known about SUA as a potential predictor of outcomes in patients affected by the acute myocardial infarction (AMI). By 2009, two studies indicated independent association between high(er) on-admission SUA and worse outcomes (8,9). The aim of the current study was to perform a systematic review and, if feasible, meta-analysis of observational studies in order to evaluate the prognostic value of SUA in this setting.

MaTerialS and MeThodS literature search and study eligibility
We searched PubMed Medline, Ovid Medline, and Embase (till February 2012, no language restrictions; Web extra material 1) and bibliographies of studies selected for full-text retrieval. Considered were only prognostic studies (prospective or retrospective) that assessed major AMI outcomes (death or major adverse cardiac events, MACE), irrespectively of the post-index event observational period, AMI type (with, STEMI, or without, NSTEMI, ST-segment elevation), and treatment strategy, specifically in relation to on-admission SUA.

Study selection and abstracting
The two authors independently reviewed the titles, abstracts, and full-text articles to decide on study inclusion and also abstracted the selected studies. Disagreements and uncertainties were resolved by a consensus.

outcomes
The outcomes of interest were mortality and incidence of MACE. Expectedly, definitions of MACE differed across studies, but always included all or most of the events such as re-infarction, angina, need for revascularization, cardiac arrest, death. Therefore, data for this composite outcome were used as reported. Extracted were time-to-event data, as well as "single time-point" data. Data were extracted for SUA as a continuous variable or as a categorical variable, eg, "high" vs "low" SUA irrespectively of the cut-off values used to define the categories.

data adjustments
Raw data on a previously described cohort (9) were available and were re-calculated to fit the format suitable for data pooling (Web extra material 2). Published data only were used for all other studies. Effect measures for SUA as a continuous variable expressed by dif-ferent "steps" in SUA increase (eg, by 1, 10 or 100 μmol/L or by 1 mg/dL) were converted to SI units (μmol/L, 1 mg/ dL = 59.48 μmol/L) and expressed per 50 μmol/L. For data adjustments see Web extra material 2.

data pooling and statistical analysis
Outcomes reported at individual time points were grouped into two time intervals after AMI. The short-term interval referred to the in-hospital period or, if data not available, to the first 30 days after AMI (in-hospital data were preferred assuming greater reliability). Medium-term interval referred to the outcomes reported at 1 year after AMI, or alternatively at 6 months to two years after AMI. Pooled estimates were generated for unadjusted and independent (from multivariate analysis) SUA effects for each outcome/ time interval available from ≥3 studies. Estimation of unadjusted effects of "high" vs "low" SUA was based on reported frequencies. Since some studies were declared as retrospective and some as prospective, both odds ratios (OR) and risk ratios (RR) were calculated. Estimation of the effects of SUA as a continuous variable and estimation of all independent effects was based on ln(OR). For the timeto-event data, pooled estimates were obtained only for independent effects based on ln(hazard ratio, HR). Considering the differences in evaluated settings (STEMI/NSTEMI patients combined, only STEMI patients, different reperfusion strategies, different covariate adjustments), randomeffects (DerSimonian-Laird) meta-analysis was performed. Since the number of studies was small, for assessment of bias we combined visual inspection of the funnel plots, position of the 95% confidence intervals around the intercept in Egger's regression and Duval and Tweedie's trim and fill method. Heterogeneity (based on I 2 and Q test) was explored only in relation to categorical moderators. Mixedeffect approach was used to analyze the subsets and to generate the across-subset estimates. We used CMA software version 2.2 (Biostat Inc., Englewood, NJ, USA).

Characteristics of the selected studies
Of the 391 identified individual study reports, 14 studies were retrieved in full text and 9 (7655 patients) met the inclusion criteria ( Figure 1, Table 1, descriptions in the Web extra material 1). Three studies included mixed STE-MI/NSTEMI cohorts with variable reperfusion strategies (8,9,14), whereas six studies included exclusively STE-MI patients who underwent primary percutaneous coronary intervention (PCI) (10-13,15,16) ( Table 1). The reported outcomes were short-term (in-hospital, 30-day) or medium/long-term outcomes with variably long followup periods (Table 1). In all studies and for all reported outcomes, high(er) on-admission SUA was univariately associated with worse results ( Table 2). All studies reported also on independent associations between SUA and at least one of the assessed outcomes ( Table 2) (Web extra material 1 for multivariate models). High(er) SUA was independently associated with worse outcomes in all studies (Table 2). In one study (12), it was independently associated with higher in-hospital incidence of MACE but not with inhospital mortality ( Table 2).
Of the five excluded studies (3411 patients, Web extra material 1), one (17) reported on a subset of patients from one of the included studies, another one (18) was poor by methodology/reporting (both indicated univariate association between higher on-admission SUA and worse in-hospital AMI outcomes), one was a small randomized controlled trial (19) indicating beneficial effects of allopurinol (an XO inhibitor) on short-term biochemical and electrophysiological indicators in STEMI/PCI patients, and two (20,21) were prognostic studies in patients undergoing revascularization procedures, but only a minority were AMI patients. One indicated independent association of "high" SUA and higher long-term all-cause mortality (20), and the other one reported independent association of "high" SUA and occurrence of PCI-induced renal impairment (21).
†Two reports by the same group stated to be completely separate cohorts. ‡Partial overlap of patients in the two reports is possible: the admission periods overlapped and the authors of the earlier smaller report (15) are listed among the authors of the later larger report (16). §no description given about the methods of identification of specific cardiovascular deaths. *abbreviations: Ci -confidence interval, hr -hazard ratio, or -odds ratio, rr -relative risk (risk ratio). †MaCe (major adverse cardiac events) includes death. definitions by study are depicted in Table 1. ‡Primary analysis was not in respect to "high" or "low" Sua. however, incidence of MaCe at 30 d could be derived for 749 patients classified based on this cut-off, and incidence of MaCe for "high" vs "low" at 6 mo could be derived for all 850 patients.

diSCuSSion
The role of SUA as a potential risk predictor of outcomes in AMI has been more closely investigated only recently: 8/9 studies identified in the present thorough literature search were published within the last three years. Although the Figure 2. random-effects meta-analysis of unadjusted (univariate) effects of on-admission serum uric acid (Sua) on different outcomes after acute myocardial infarction (aMi). The effects are expressed as a contrast between "high" (above a defined cut-off value, in μmol/l) and "low" (below the cut-off ) Sua. asterisk represents original cohort data recalculated (see text and Web extra material 1). dagger indicates when death is included. M -men, Fwomen, STeMi -aMi with ST-segment elevation, nSTeMi -aMi without ST-segment elevation, PCi -percutaneous coronary intervention.
overall number of patients was reasonably large, the present results should be viewed in the light of the limitations of individual studies and the fact that estimates based on a small number of studies could be compromised by heterogeneity and/or bias, particularly in non-randomized settings.

Main findings
Based on the agreement of univariate and independent associations indicated by the pooled estimates (and individual studies), the present data suggest that high onadmission SUA predicts higher short-term and medium/ long-term mortality and incidence of MACE in AMI patients. "High" SUA appears to be in the range of hyperuricemia or high-normal values.

limitations of the individual studies
Most of the studies (6/9) were retrospective, and hence inherently susceptible to bias. Still, in the case of AMI, this limitation might not be particularly relevant as the condition is handled after pre-defined protocols with patient monitoring and accurate (real-time) data recording as a part of the good clinical practice standard, particularly con-sidering short(er) time-periods (eg, during in-hospital stay). It is reasonable to expect susceptibility to bias/inaccuracy to increase with longer observational periods. All studies that reported on outcomes over periods longer than 30-day (8,9,11,13,14,16) were retrospective. However, two were based on databases generated prospectively after a pre-defined strategy (8,14), whereas methods implemented in the others (9,11,13,16) convincingly suggested a reasonably accurate identification of events and their timing. Hence, it seems plausible to consider individual study data as acceptably reliable.
No study matched the "high" and "low" SUA patients in respect to the major AMI outcome risk factors (eg, age, Killip class, or renal function) suggested by the established risk stratification systems (eg 22,23, ). Hence, although most of the multivariate models were quite complex, important effects or effect modifiers might have been occasionally omitted. While it seems unlikely that this fact could have resulted in misinterpretation of "no effect" or a "favorable effect" of high SUA as an "adverse effect" (particularly considering the qualitative agreement between univariate and multivariate estimates), it might have troubled the attempts to accurately quantify the true effect of high SUA.

Strengths and limitations of the pooled analysis
The present literature search (no language or publication year restrictions) most likely omitted no relevant publication. The decision to exclude four of the identified prognostic studies was justified as they clearly referred to patients embraced in other included reports (17), were so poor in respect to reporting/methodology (18) that it was unclear whether they indeed dealt with the topic of interest, or included only sporadic AMI patients (20,21). The variability of the study particulars (AMI type/treatment, definition of MACE, consideration of SUA, follow-up duration, timing of the outcome assessment) further restricted the number of studies eligible for pooled analysis per outcome/time-period, particularly in the case of independent associations. We therefore addressed the issue of bias and, within the limitations of the available methodology, demonstrated that the observed SUA effects were not artifacts. Considering the methodological differences between studies (including differences in multivariate models), we a priori considered that no common effect size should be assumed and applied the random-effects meta-analysis. Still, for the short-term and simply defined outcomes, heterogeneity was non-existing or low and it increased when the composite outcome with variable definitions (MACE) or medium/long-term periods were considered. It should be noted, however, that heterogeneity was not due to essentially different study results (all individual unadjusted or adjusted estimates were in the same direction), but rather to the differences in size of the high(er) SUA effect. The undertaken exploration of heterogeneity should be viewed with caution, given the limited number of studies, but it indicates the disease/treatment characteristics (STEMI/PCI vs mixed STEMI/NSTEMI patients with variable reperfusion strategies) as its main source. Overall, the concordant primary random-effect and across-subset mixed-effect estimates provide evidence of an independent association of high(er) on-admission SUA and worse AMI outcomes that may vary in strength across the disease/treatment characteristics and the observed post-AMI period.
Some of the individual studies indicated that the effect of high SUA on medium/long-term outcomes was conditional on Killip class (more pronounced in patients with Killip class III/IV) (8), age (more pronounced in patients <63 years of age) (9), or on the levels of N-terminal pro-Btype natriuretic peptide (14) (more pronounced at higher concentrations). Since not based on the individual patient data, the present analysis could not address these issues.

Practical relevance of the present observations
Considering that the individual studies were conducted in different settings reflecting daily situations and in different parts of the world that could differ in genetic/cultural factors affecting SUA (eg, nutritional habits, alcohol consumption, cigarette smoking), the present analysis supports a view about a robust association between the high(er) on-admission SUA and poor AMI outcomes. This observation has two potential practically relevant implications. First, it suggests that on-admission SUA should be considered in the risk stratification in AMI patients. In this respect, it should be noted that the independent effects of SUA (high-normal or hyperuricemic values) on short-term adverse outcomes are comparable in size to the effects of age >65 years, Killip class>II/III, or renal insufficiency, which are among the most important individual elements in the established risk stratification systems in AMI (22,23). Next, regardless of the known controversy about the mechanistic relationship between SUA and cardiovascular outcomes (a "direct pathogen" or a "mere marker" of the increased XO activity) (2-6), high on-admission SUA might depict situations that could benefit from XO inhibition. The results of a small placebo-controlled trial (19) showing beneficial short-term effect of allopurinol in STEMI/PCI patients support such a view.

Conclusions
Over the past few years, several studies have evaluated the predictive value of on-admission SUA for the outcomes in AMI patients. Although still modest in number, the available data suggest that on-admission SUA in the high-normal or hyperuricemic range independently predicts worse short-term and medium/long-term outcomes. Additionally, preferably prospective studies are needed to more precisely quantify the relationship between SUA and the outcomes in different settings (eg, AMI type, reperfusion strategies) and to characterize the seemingly complex relationships between high SUA and other predictors.
acknowledgment We are thankful to the Reviewer No. 3 for his/her benevolent and useful comments.
Funding None.
ethical approval Not required.
declaration of authorship VT conceived and carried-out the presented work, and drafted and approved the manuscript. SC participated in designing, conducting, and reporting of the presented work.
Competing interests All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organization for the submitted work; no financial relationships with any organiza-