Heart failure (HF) is a major health problem whose attributable burden of morbidity and mortality continues to increase as the population ages.1 The pathogenesis of HF remains incompletely understood. The key pathways implicated include the effect of neurohormonal activation,2 oxidative stress,3 and immune activation4 on myocytes and the interstitium. Several biomarkers have emerged to evaluate the prognosis and progression of HF. Currently, however, no marker is used in the clinical decision making for HF patients.
Serum uric acid (SUA) is a byproduct of purine catabolism, the terminal steps of which are catalyzed by xanthine oxidoreductase (XOR). In HF, SUA levels may rise due to increased purine catabolism resulting from tissue hypoxia, apoptosis, and/or enhanced or upregulated XOR activity.5 Therefore, SUA could be used as a prognostic marker in HF progression.
Recent studies have noted that an elevated level of SUA predicts the development of cardiovascular heart disease, hypertension, obesity, kidney disease, and diabetes.6 The purpose of the present study was to evaluate the evidence supporting SUA as a predictor of all-cause mortality in patients with HF and to determine whether there is an SUA cut-off for the increase in risk.
Search Strategy. We accessed MEDLINE through PubMed, the Internet access site provided by the National Library of Medicine. MEDLINE includes articles from 1966 until March 2009. Our search strategy included the following search terms: uric acid OR trioxopurine OR potassium hydrate OR urate OR ammonium acid urate OR sodium urate monohydrate OR monosodium urate monohydrate OR sodium acid urate monohydrate OR sodium urate OR monosodium urate OR sodium acid urate AND heart failure OR cardiac failure OR myocardial failure OR left-sided heart failure OR left-sided heart failure OR right-sided heart failure OR right-sided heart failure OR congestive heart failure OR heart decompensation. To ensure a comprehensive literature search, we examined reference lists from our retrieved articles and reference literature from journals cited most frequently in the literature searches.
Selection Criteria. As a first step in the review process, two members of the study team independently reviewed the abstracts identified by the search to exclude those that did not meet our eligibility criteria. When there was no abstract or when the reviewers could not determine from the abstract whether the article met the eligibility criteria, the team obtained a full copy of the article to review.
Articles were considered for inclusion if they reported on original data from included patients with HF where SUA was measured and mortality rates were reported. Two investigators independently reviewed each eligible article identified by the abstract review process (LT and AH). One investigator was responsible for completing both the quality assessment (AP) and content abstraction forms (AH), and the second confirmed the accuracy of the data abstracted (LT). Differences between the two reviewers in either quality or content abstraction were resolved by consensus. Inter-rater reliability between reviewers was 98%.
Data Extraction. The key exposure variable was the SUA measurement at baseline in mg/dL. One study reported SUA in μmol/L, and we converted those values using the following conversion equivalence: 1 mg/dL = 59.48 μmol/L. All studies measured SUA using the uricase-peroxidase enzymatic method.
Three studies reported the exposure as a dichotomized variable while 3 other studies reported SUA in quartiles. For the primary analysis we defined hyperuricemia as an SUA >7 mg/dL7; however, the selected studies did not have consistent cut-offs and a few ranges crossed 7 mg/dL. If this occured we assigned the category based on where the majority of the values in the range fell. For example, if a cut-off range was 5.8 mg/dL to 7.2 mg/dL in one study, that group would be assigned to the “no hyperuricemia” group. For our evaluation using different SUA cutoffs, we abstracted and categorized the SUA levels for each study into 5 groups: the reference group (lowest quartile of SUA reported in one of the studies <5.0 mg/dL8), 5 mg/dL to 6 mg/dL,8 6 mg/dL to 7 mg/dL,9,10 7 mg/dL to 8 mg/dL,8 and >8 mg/dL.9–11 We used the same strategy described above to assign ranges of values from the different studies to each category. The rationale for grouping studies that reported on outcomes with SUA levels >8 mg/dL was that we were interested in understanding the relationship between lower levels of SUA and outcomes.
We also abstracted the variables adjusted for in the multivariate analysis and whether the studies included patients with acute or chronic HF and the method of SUA collection. Because hemodynamics and treatment strategies are different between acute and chronic HF, we performed a secondary analysis stratifying by that variable.
The outcome variable of interest was all-cause mortality. All-cause mortality was reported in all 6 articles defined as death due to any cause during the follow-up time. At the same time, we abstracted event data in each of the reported SUA cutoffs.
Qualitative Assessment. The study team developed article review forms that were pilot tested and have been validated in prior projects.12 The quality assessment forms included items about study quality in the following categories: representativeness of study population, bias and confounding; statistical quality and interpretation, and conflict of interest. The items in these categories were derived from study quality forms used in previous projects and were modified for this project. The study team responded to each question with a score of zero (criteria not met), one (criteria partially met), or two (criteria fully met), unless only 2 criteria were specified: yes (criteria fully met) or no (criteria not met). The score for each category of study quality was presented as a total score adding the individual scores from each question in the category. The overall quality score was the average percentage of the 4 categoric scores and is described in the results.
Statistical Analysis. We reported relevant baseline characteristics as median values with the range of observations. To assess for heterogeneity across studies we used the Cochran Q chi-square statistic (significance level of P<.10). Publication bias was assessed using a visual inspection of the Begg’s funnel plot and tested using the Egger’s weighted regression method.13
For quantitative analysis we used Stata 10.0 (StataCorp LP, College Station, TX) to calculate the relative risk (RR) of all-cause mortality with the respective 95% confidence intervals (CIs) and P values. For our main analysis we categorized the data by mortality and hyperuricemia rates. We used both fixed-effects and DerSimonian and Laird random-effects models to calculate the pooled RR across levels of SUA. Both models yielded similar findings, thus we elected to present results from the fixed-effects model since no significant heterogeneity was identified among studies.
Because the pooled studies had a cohort study design, we evaluated the influence of variables used for adjustment in the individual studies (Table I) and the different levels of SUA cut-offs in a meta-regression model to test the effects of those variables on the results.
Table I. Confounders Included in the Multivariate Models Evaluated in 6 Studies Included in the Analysis
Abbreviations: BNP, brain natriuretic peptide; GFR, glomerular filtration rate; HF, heart failure; NE, norepinephrine; NYHA, New York Heart Association classification; SUA, serum uric acid.
To determine the SUA cut-off where mortality increases, we performed two independent secondary analyses. First, we calculated the pooled RR for all-cause mortality across all studies for each of our selected cut-offs. The reference used for this calculation was the lowest cut-off reported in one of the studies, which was <5.0 mg/dL. Second, to determine the β-coefficient of increase in the risk of all-cause mortality by each SUA cut-off, we performed a weighted meta-regression with no intercept term using generalized least squares for trend estimation. We used the natural logarithm of the RR of all-cause mortality as the dependent variable and the SUA cut-offs as the independent variable to estimate a dose response. We used the “pool-first” method proposed by Greenland and Longnecker.14 In this method, we used as a reference group for each study, its own lowest value of SUA cut-off reported. We did not include the studies that did not report multiple cut-offs in this analysis.
Literature Search. The search strategy retrieved 354 unique citations. Of these, 348 were excluded due to the following reasons: 100 were review articles, 46 reported outcomes that were not mortality, 21 were not cohort studies, 53 did not measure uric acid, 4 had insufficient data, and 124 did not include HF patients. Therefore, 6 articles met our inclusion criteria.8–11,15,16
Study Characteristics. Baseline characteristics of the 6 studies are shown in Table II. All studies were single-center, prospective cohorts that included 1456 HF patients and evaluated mortality in a median follow-up of 40 months (range, 12–51 months). Four studies reported results on patients with acute HF hospitalizations. The median ejection fraction (EF) in the studies was 32% (range, 26%–40%), with a median serum creatinine of 1.3 mg/dL (range, 1.0–1.3 mg/dL). The studies included patients with ischemic cardiomyopathy in a median of 58% (range, 25%–71%).
Table II. Baseline Characteristics of 6 Studies Evaluating the Association Between SUA and All-Cause Mortality
Ischemic Etiology, %
Abbreviations: NR, not reported; SUA, serum uric acid reported as categoric variable.
Publication Bias and Qualitative Analysis. There was no evidence of publication bias (P=.14). The population was well represented in the majority of the studies. The description of the method of SUA measurement did not provide reliability measures and the timing of the measurement was reported in only one of the studies that included acute HF patients. Also, the ascertainment of the outcome did not provide enough detail for replication. All studies were adjusted for known important prognostic factors such as EF and New York Heart Association (NYHA) classification at baseline but did not adjust for the use of angiotensin-converting enzyme inhibitors and β-blockers (Table I).
Risk of All-Cause Mortality by SUA Level. The median mortality rate for patients with hyperuricemia was 50% (range, 26%–100%) compared with patients with normal uric acid (22% [range, 8%–41%]). There was no significant heterogeneity between studies (P=.10), and all studies found a statistical significant association between mortality and SUA levels. The RR of all-cause mortality was 2.13 (95% CI, 1.78–2.55) for SUA >7 mg/dL compared with <7 mg/dL SUA level (P<.01) (Figure). Neither the number of confounding variables (P=.69) nor the variability of the cutoffs (P=.76) affected the results.
Risk of All-Cause Mortality by Study Population. Four studies reported mortality in patients who were admitted to the hospital for acute HF exacerbations. Two studies reported the timing of the SUA measurement. The first measured SUA levels prior to discharge9 and the second during the first days of admission.11 The RR of all-cause mortality for acute HF was 2.4 (95% CI, 1.5–3.7) and for chronic HF was 2.1 (95% CI, 1.5–2.9) for SUA >7 mg/dL compared with SUA level <7 mg/dL.
Risk of All-Cause Mortality by Cut-Off of SUA Level. The results of the effects of SUA cut-offs and meta-regression analysis for trend are presented in Table III. There was an excess risk of all-cause mortality for SUA at all cut-off points; however, this association became statistically significant at 7 mg/dL to 8 mg/dL of SUA. The RR of all-cause mortality for SUA levels of 7 mg/dL to 8 mg/dL was 1.6 (95% CI, 1.0–2.4) compared with the reference group.
Table III. Stratified Pooled Relative Risk of Mortality by SUA Cutoff in 6 Studies
The results of the meta-regression analysis of trend found a β-coefficient of all-cause mortality of 0.08 (95% CI, 0.06–0.09; P<.01) by increasing SUA cut-offs. The overall results indicate a linear association between SUA and the RR of mortality (P<.01 for trend).
Our study found that high SUA levels increase all-cause mortality in patients with HF, that this increase is consistent in patients with both acute and chronic HF, and that this increase in risk seems to start at an SUA level of 7 mg/dL. The strengths of this analysis are the inclusion of studies from different countries, the lack of heterogeneity and publication bias, and large sample size, permitting an analysis of dose-dependence.
Several limitations deserve mention. First, the selected studies did not report consistently the same SUA level cut-off; therefore, when combined into the categories created for our analysis, misclassification of event data could have occurred. However, we evaluated the effect of using different cut-offs by using meta-regression methodology, and the difference in cut-offs did not impact the results. Second, we found few studies reporting lower cut-offs of SUA, which could explain our nonsignificant results when evaluating the association between <7 mg/dL of SUA and mortality. Third, the selected cohort studies used different adjustment variables in their multivariate regression analysis, which could produce some variability in the RRs of mortality for uric acid reported by the different studies; however, using meta-regression, we did not find that the number or type of adjustment-confounding variables influenced the results. Finally, we did not include meeting abstracts and citations in languages other than English, because the values required for the analysis were not reported consistently.
Biomarkers (BNP and troponins),17 EF, NYHA,18 and the Seattle HF index19 are important HF prognostic markers. Although these prognostic markers have a good correlation with mortality, their discriminatory ability has been questioned in early stages of HF. Furthermore, the measurements of left ventricular EF and BNP have important cost implications, especially when serial measurements are potentially helpful in guiding therapy. Our results demonstrate that SUA level can be used as a prognostic marker for all-cause mortality in HF, and that this prognostic ability is potentially independent of other well-established measures of HF severity. The RR of all-cause mortality was increased 2-fold in patients who had an elevated SUA level compared with patients with a lower SUA level. Most of the studies included in this analysis adjusted for left ventricular EF, therefore indicating that SUA gives a potential advantage when added to well-known HF prognostic markers.
Uric acid has been previously implicated in the pathogenesis and prognosis of cardiovascular diseases in numerous studies.20,21 The mechanism of increased XO activity in the myocardium of the failing heart is controversial. The increase in uric acid could be a reflection of increased XO activity due to apoptosis, resulting in accumulation of uric acid precursors (hypoxanthine and xanthine). This catalytic reaction produces as an end-product reactive oxygen species that, in turn, cause a vicious cycle resulting in oxidative stress and programmed cell death. The source of uric acid in the failing heart seems to be the failing myocardium itself since it exists widely in the heart muscle, and an increase in endothelium-bound XO activity in HF patients has been documented. This is supported by the fact that SUA level has been found to be elevated in the coronary sinus in HF when compared with control patients.16
In addition, our study found that the cut-off level of SUA that increases mortality in HF was lower than reported previously.10 In our analysis, we found that patients with lower levels (5 mg/dL and 6 mg/dL) had a higher risk of death; however, this association was not statistically significant. In contrast, a clear significant association was seen at 7 mg/dL. Two studies included in our analysis reported on specific SUA cut-offs that might increase death. The first, by Alimonda and colleagues,11 consistent with our findings, found that the hazard ratio of death was significant after crossing 7.7 mg/dL of SUA. The second, by Anker and colleagues,10 reported a graded relationship between SUA quartiles and mortality that was significant in the second quartile of SUA; however, the second quartile of SUA included a wide range of observations (6.9–10 mg/dL of SUA) and therefore created difficulties for an analysis of lower cut-points.
Two randomized trials have used inhibitors of the XO pathway. The first randomized 405 class III or IV HF patients to oxypurinol or placebo and found no significant changes in the risk of a composite end point. However, the study found a trend in patients with SUA levels ≥9.5 mg/dL.22 The second study included 60 NYHA class II or III HF patients randomized to oxypurinol and placebo.23 This study found no improvement in EF or increase in functional capacity after 4 weeks of therapy. Therefore, SUA might not have a causal role in HF but rather is simply a biomarker of adverse prognosis indicative of oxidative damage in the myocyte.
Uric acid is an important prognostic marker for all-cause mortality in HF. Patients with HF have an increased risk of death if they are hyperuricemic when compared with patients with normal uric acid levels. Future studies should evaluate the predictive ability of SUA when added to known predictors of HF mortality and the value of serial measurements of SUA.