Uric acid and survival in chronic heart failure

From the Division of Applied Cachexia Research, Department of Cardiology, Charité, Campus Virchow-Klinikum, Berlin, Germany, and the Department of Clinical Cardiology, Imperial College, National Heart and Lung Institute (NHLI), London, United Kingdom

Clinical parameters of prognostic significance are crucially important for risk stratification and treatment planning in chronic diseases such as heart failure. In recent years, numerous parameters have emerged that can predict prognosis in chronic heart failure patients. Problematically, however, many of these parameters are assessed only in research settings and with tests that are not suitable for routine clinical use. Simple parameters that can be measured anywhere and at low cost, but with sufficient prognostic power, are needed for improved clinical applicability in determining chronic heart failure prognoses.

In chronic heart failure, hyperuricemia is a common finding (independent of kidney function and diuretic dose) that can be used as a marker of impaired oxidative metabolism and hyperinsulinemia,1,2 inflammatory cytokine activation,3 as well as impaired vascular function.4,5 It has been shown that in chronic heart failure, levels of uric acid (UA) reflect the activity of xanthine oxidase,6 an enzymatic system that is an important source of oxygen free radicals.7 Moreover, through the degradation of accumulated purines, UA is a general marker of cell death and hypoxia. We therefore hypothesized that UA might be a potent prognostic marker in chronic heart failure, and that high levels of this acid in combination with a low left ventricular ejection fraction and peak oxygen consumption (VO2) would indicate a particularly poor prognosis.

Patients and methods

To test our hypotheses, we assessed the relationship between survival and UA level in 112 chronic heart failure patients who had been extensively evaluated and prospectively recruited into a long-term metabolic study program from 1992 to 1997. To validate our results, we used a second patient sample, consisting of chronic heart failure patients (n = 182) who were treated on an outpatient basis at our hospital. We used this latter patient sample to test the principal results and best cutoff values in receiver-operator curve analyses for UA levels as prognostic markers as derived from the first,

112-patient sample. All patients were clinically stable and were being treated with an individually optimized treatment regimen of standard heart failure medications, including diuretics, angiotensin-converting enzyme (ACE) inhibitors, beta blocking agents, digitalis, nitrates, calcium antagonists, aspirin, and warfarin. As of September 2001, follow-up data for survival based on all-cause mortality was available from the United Kingdom’s Office of National Statistics.

All parameters needed for calculating the seven-parameter heart failure survival score (HFSS)8 were available for 194 patients from both cohorts, and we compared this with the prognostic power of the UA

level. We also assessed the survival of chronic heart failure patients with a three-risk-factor model including the three simple parameters of a high UA level as a marker of metabolic status, peak VO2 as a marker of low exercise capacity (peak VO2 ≤ 14 mL/kg per minute),9 and a low left ventricular ejection fraction (≤ 25%) as a marker of impaired hemodynamic and cardiac function.10 We accordingly defined this three-risk-factor model as a metabolic, functional, and hemodynamic (MFH) staging system.

We then retrospectively analyzed data from 120 consecutive patients who underwent full assessment of eligibility for heart transplantation at the German Heart Institute in Berlin from January to June 1999. On the basis of the results of this assessment, we calculated the MFH score for each patient and compared it retrospectively with the decision of the responsible physicians in each case. The decision for or against listing a patient as eligible for heart transplantation was based solely on the clinical findings for the patient, because the responsible physicians were unaware of the patient’s MFH score or of our data on UA level and heart failure prognosis.

Results

In the 112-patient derivation sample, UA levels ranged from 187 to 930 µmol/L. A normal level was considered a concentration below 400 µmol/L, which was observed

in 34 patients (30%). During a follow-up of 51 ± 39 months, 69 patients died (12-month mortality, 24%; 95% confidence interval [CI], 16%—32%). Survivors had significantly lower mean UA levels than did nonsurvivors at 438 ± 108 µmol/L versus 544 ± 163 µmol/L,

respectively (P = .0003).

Established prognostic markers, such as urea, age, peak VO2 (all

P < .001), left ventricular ejection fraction (P = .003), creatinine, sodium, and furosemide equivalent dose (all P < .03), predicted mortality. UA as a continuous variable significantly predicted mortality (hazard ratio [HR], 1.00498; 95% CI, 1.00322-1.00674). Additionally, impaired survival was predicted by UA levels above the median (> 484.5 µmol/L; relative risk [RR], 2.6; 95% CI, 1.6—4.3; P < .0001), in the highest tertile (≥ 565 µmol/L; RR, 3.9; 95% CI, 2.4–6.4; P < .0001), and in the highest quartile (≥ 595 µmol/L; RR, 3.2; 95% CI, 1.9–5.4; P < .0001). In receiver-operator curve analyses, a UA level of 565 µmol/L was the best predictor of survival status at both 12 and 18 months.

The 182 outpatients of the validation sample were older, on average, than the derivation study patients (63 ± 12 years versus 59 ± 12 years), and they had a somewhat better left ventricular ejection fraction (32% ± 15% versus 26% ± 15%; both P < .05). The two samples were similar in terms of New York Heart Association (NYHA) functional class, peak VO2, sex and etiology distribution, UA levels, and creatinine levels (all P > .05). Mean UA levels were 466 ± 155 µmol/L (range, 106—1,251 µmol/L). A UA level below 400 µmol/L was seen in 35% of patients (n = 64) in the validation sample. During a follow-up of 41 ± 23 months, 59 patients died (12-month mortality, 15%; 95% CI, 10%–20%). Survivors had significantly lower mean UA levels than nonsurvivors at 418 ± 102 µmol/L versus 565 ± 195 µmol/L, respectively (P < .0001). UA levels also predicted mortality in the validation sample (RR, 1.00532; 95% CI, 1.00400–1.00666; P < .0001). For every 100 µmol/L increase in UA, the risk of death increased by 53%.

When we used the data from the derivation sample, we found that

a UA level of 565 µmol/L or more (RR, 7.14; 95% CI, 4.20—12.15) was strongly related to increased mortality (P < .0001). For patients with

such UA levels, survival at 12 and

24 months was 52% (34%—70%)

and 36% (18%—53%), respectively.

In patients with a UA level below 565 µmol/L, survival at 12 and

24 months was 92% (88%—96%) and 86% (80%–92%), respectively. In

a multivariate Cox model with

seven variables (n = 113), only UA

(P < .0001), left ventricular ejection fraction (P < .04), and peak VO2

(P < .005) predicted prognosis. Creatinine, sodium, urea, and age were the variables that did not predict prognosis.

When we evaluated all 294 chronic heart failure patients from both cohorts, we found a graded rela-tionship between serum UA and mortality in heart failure (P < .0001;

figure 1). There was a stepwise increase in mortality risk paralleling higher UA levels (figure 2).

UA, HFSS, and MFH score. Sufficient data were available for 194 patients to calculate the seven-pa-rameter HFSS. The mean HFSS in these patients was 8.56 ± 1.24. Both at its three risk levels (chi-square test, 39.1; P < .0001) and as a con-

tinuous variable (chi-square test, 45.4; P < .0001), the HFSS had prognostic power that was similar to

that of UA (chi-square test, 54.0;

P < .0001). Both the HFSS and the UA level predicted prognosis independently of each other, regardless of whether the parameters were treated as continuous or categorical variables (all P < .0001).

We were able to calculate the three-risk-factor MFH score for

212 patients (figure 3). Survival was best for those patients without any risk factors (3-year survival, 91%; 95% CI, 85%—98%). With each additional risk factor, we observed a stepwise impairment in survival. Three-year survival with one and two risk factors was 67% (57%–77%) and 33% (19%–47%), respectively. Survival was lowest for patients with all three risk factors (3-year

survival, 12.5%; 95% CI, 0%—29%).

MFH staging and heart transplantation (German Heart Institute, Berlin). For the 120 chronic heart failure patients (105 men and 15 women) undergoing evaluation for heart transplantation (clinical characteristics: age, 53 ± 8 years; NYHA functional class, 2.5 ± 0.7; treadmill peak VO2, 14.3 ± 4.4 mL/kg per minute; left ventricular ejection fraction, 23% ± 8%; UA, 469 ± 145 µmol/L), we calculated the three-risk-factor MFH score retrospectively. Of these patients, 20 had an MFH score of 0 and another 35 had an MFH score of 1. None of these 55 patients was listed for heart transplantation. Of 47 patients with an MFH score of 2, there were 24 patients (51%) listed, whereas 16 of 18 patients with an MFH score of 3 (89%) were listed for heart transplantation. The positive predictive value of an MFH score of 0 or 1 for nonlisting for heart transplantation was 100%.

Discussion

Our study demonstrated that high serum UA levels are a strong independent marker of impaired prognosis in patients with moder-ate to severe heart failure. A stepwise, graded relationship exists between the serum UA level and survival in these patients. Assessing the UA level provides information that is independent of and prognostically stronger than that of other well-established parameters, such

as clinical status, exercise capacity, and parameters of kidney function.

Several factors support the concept that serum UA is a valid metabolic marker in chronic heart failure. First, chronic heart failure is a condition of several abnormally activated metabolic and hormonal regulatory feedback loops. In this context, several pathways, such as those of hypoxia, cell death, immune activation, and insulin resistance, may contribute to increased concentrations of purine metabolites, leading to the activation of xanthine oxidase and increased production of UA. Hyperuricemia has been found to reflect increased xanthine oxidase activity in chronic heart failure.6 Second, this enzymatic system is an important source of oxygen free radicals,7 which provide the pathophysiologic link between high UA levels and a wide variety of detrimental processes including increased cytokine production, cell apoptosis, and endothelial dysfunction, all of which occur in patients with chronic heart failure. In a prospective series of studies of chronic heart failure patients, we have previously shown that hyperuricemia relates to many of these abnormalities independently of diuretic treatment and of markers of kidney function.2-5

In heart failure patients, treatment must be individualized to achieve an optimal outcome. This tailoring of treatment requires a re-liable assessment of individual

patient prognosis. One prognostic stratification system, the seven-pa-rameter HFSS,8 has been previously validated for this purpose, but is computer-based and may not be simple enough for routine use. We have now shown that the UA level predicts prognosis as well as the HFSS and independently of it.

Three main areas of relatively independent prognostic importance in heart failure have emerged: (1) hemodynamic factors, such as the left ventricular ejection fraction; (2) the patient’s functional status, such as peak VO2; and (3) metabolic aspects including neuroendocrine and immunologic processes. The MFH score appears to be a helpful and easy-to-use tool for improving clinical and prognostic assessment in chronic heart failure. We therefore recommend the use of MFH assessment for predicting prognosis in these patients.11 In such MFH staging, we suggest that measurement of UA as a straightforward and easily assessable metabolic marker should become a routine procedure in the evaluation and follow-up of heart failure patients.

Peak VO2 at maximum exercise is widely accepted as an ideal parameter for objectively assessing the functional capacity of chronic heart

failure patients. Combining a threshold for this latter parameter (peak VO2 < 14 mL/kg per minute) with the functional cardiac assessment (left ventricular ejection fraction

≤ 25%) and UA as a metabolic pa-rameter (UA ≥ 565 µmol/L) permits four risk groups to be distinguished, with mortality ranging from very low (MFH score, 0; mortality at

3 years, 9%) to extremely high (MFH score, 3; mortality at 18 months, 87.5%). In that it involves an easily applicable method with wide availability at very low cost, the MFH score appears to be a meaningful tool for improving the stage-related, tailored treatment of chronic heart failure. Besides benefiting patients with moderate, compensated chronic heart failure, use of the MFH score may also benefit patients with severe heart failure by providing additional information for determining their eligibility for heart transplantation.

To extend its use beyond chronic heart failure to the general medical community, the MFH score needs to be tested in further studies to validate its prognostic power and possibly to simplify it to a greater extent. The assessment of peak VO2, for example, may be replaceable by simpler measures, such as the 6-minute walk test. Similarly, some clinicians may want to use higher left ventricular ejection fraction cutoff levels or to replace these levels with brain natriuretic peptide levels.

It has been shown that changes

in therapy for chronic heart failure may affect the prognostic value

of parameters used for its assessment.12 Consequently, the studies that we report here may have had

a limitation in that relatively few patients (6%) were receiving beta blockers. Preliminary analyses also indicate that in the population of

the Second Evaluation of Losartan in the Elderly (ELITE II), UA was a strong, independent prognostic factor that was independent of concomitant beta blocker therapy. Whether therapeutically targeting UA may have beneficial effects in chronic heart failure has not been assessed in detail; however, emerging data suggest that inhibiting the production of UA through blockade of xanthine oxidase with allopuri-

nol may be beneficial.13,14

Elevated serum UA levels are common in chronic heart failure and are a strong, independent marker

of impaired prognosis in patients with moderate to severe heart failure. As an easily measurable parameter with wide availability, the UA level may be useful for providing ad-ditional prognostic information. Whether UA predicts the development of chronic heart failure or the mode of death in any given individual is unknown. Fang and Alderman15 showed that increased serum UA levels in the general population are independently associated with all-cause, total cardiovascular, and ischemic heart disease mortality. Therefore, the assessment of UA may be of general value in both healthy and unhealthy individuals. The MFH score also appears to be a helpful, easy-to-use tool to improve clinical and prognostic assessment in chronic heart failure.