Author + information
- Received April 29, 2011
- Revision received September 5, 2011
- Accepted October 3, 2011
- Published online January 24, 2012.
- Bård Waldum, MD⁎,†,⁎ (, )
- Arne S. Westheim, MD, PhD‡,
- Leiv Sandvik, PhD§,
- Berit Flønæs, RN∥,
- Morten Grundtvig, MD¶,
- Lars Gullestad, MD, PhD#,
- Torstein Hole, MD, PhD⁎⁎,†† and
- Ingrid Os, MD, PhD†
- ↵⁎Reprint requests and correspondence:
Dr. Bård Waldum, Department of Nephrology, Oslo University Hospital, Ullevål, N-0407 Oslo, Norway
Objectives The aim of this study was to evaluate the prognostic impact of anemia in outpatients with chronic heart failure attending specialized heart failure clinics and specifically to investigate its prognostic utility in patients with severe renal dysfunction or advanced heart failure.
Background Anemia is an independent prognostic marker in patients with heart failure. The effect of anemia on mortality decreases with increasing creatinine levels.
Methods Multivariate Cox regression analyses were used to investigate the prognostic effect of anemia in 4,144 patients with heart failure from 21 outpatient heart failure clinics in Norway. Severe renal failure was defined as estimated glomerular filtration rate ≤45 ml/min/1.73 m2 and advanced heart failure as New York Heart Association functional classes IIIb and IV.
Results Baseline anemia was present in 24% and was a strong predictor of all-cause mortality (adjusted hazard ratio [HR]: 1.30, 95% CI: 1.09 to 1.56, p = 0.004). Baseline anemia did not predict mortality in the 752 patients with severe renal dysfunction (adjusted HR: 1.08, 95 % CI: 0.77 to 1.51, p = 0.662) and the 528 patients with advanced heart failure (adjusted HR: 0.87, 95% CI: 0.56 to 1.34, p = 0.542). In the 1,743 patients who attended subsequent visits, sustained anemia independently predicted worse prognosis (adjusted HR: 1.47, 95% CI: 1.10 to 1.94, p = 0.008), whereas transient and new-onset anemia did not.
Conclusions According to our study, baseline anemia was not an independent predictor of all-cause mortality in outpatients with heart failure and accompanied severe renal dysfunction or advanced heart disease. Sustained anemia after optimizing heart failure treatment might imply worse prognosis independently of renal function and New York Heart Association functional class.
Anemia is prevalent among patients with heart failure (1). Renal dysfunction, activation of neurohormonal and inflammatory responses, drug effects, and bone marrow hyporesponsiveness all seem to contribute to the development of anemia in patients with heart failure (2,3). Estimates of the actual prevalence of anemia among patients with heart failure vary widely due to differences in the definition of anemia and in investigated patient populations. Most studies have identified prevalence of anemia >20% (4).
Numerous studies have demonstrated a strong relationship between anemia and mortality in populations with heart failure (5,6). It is apparent from a recent review (5) that most studies were performed in selected study populations or among hospitalized patients. Epidemiological data suggest an inverse relationship between serum creatinine level and the influence of anemia on mortality in heart failure patients, because the effect of anemia on mortality declines with increasing serum creatinine (5). This might indicate that anemia is not an independent risk factor of mortality in all subgroups of heart failure patients.
Our aims were to evaluate the prognostic impact of anemia both at baseline and sustained during follow-up on all-cause mortality in patients with chronic heart failure attending specialized outpatient heart failure clinics. Furthermore, we wanted to evaluate the prognostic effect of anemia in subgroups of high-risk heart failure patients.
The Norwegian Heart Failure Registry
The Norwegian Heart Failure Registry was initiated in October 2000 with the intent to collect data on outpatients attending office visits in heart failure clinics (7). The registry originally included 10 hospitals. By March 2008 the number of participating units had gradually increased to 24 clinics situated in all regions in Norway; at that time 5,030 patients were included. Cardiologists in cooperation with specially trained nurses run the heart failure outpatient clinics. Patients with heart failure of any etiology, New York Heart Association (NYHA) functional class I to IV, diagnosed clinically according to guidelines from the European Society of Cardiology (8), were enrolled consecutively in the heart failure clinics. At the first visit (baseline) medical history, physical examination, echocardiography, laboratory results, and the medical management of heart failure were registered. Visit 2 was recorded, after adjustment of medical therapy and having patients undergo an educational program. If required, additional clinical visits could be scheduled to ensure optimization of therapy before Visit 2. Finally, Visit 3 was planned to occur 6 months after Visit 2. Mortality data were retrieved from the Norwegian death registry kept by Statistics Norway; by March every year, the current database was updated with respect to mortality data by March 2008. Both baseline hemoglobin level and mortality data were available in 4,144 patients from 21 hospitals, who were included in the baseline analyses. Less than 50% (n = 1,972) completed all planned visits in the heart failure registry, of whom 1,743 patients had registration of hemoglobin levels at both baseline and Visit 3.
All participants provided written informed consent before inclusion in the database. Only unidentifiable data were entered in the database. Permission for this analysis was granted from the National Data Inspectorate and the Regional Committee of Medical and Health Research Ethics.
Anemia was defined according to the World Health Organization (WHO) guidelines as blood hemoglobin below 13.0 g/dl in men and below 12 g/dl in women (9). Hemoglobin levels were analyzed at the local heart failure clinics. Baseline anemia was defined as anemia at Visit 1. Sustained anemia was defined as anemia at both Visit 1 and Visit 3. Transient anemia was defined as anemia at baseline but not at the last visit, whereas patients with new-onset anemia were not anemic at baseline but anemic at the last visit.
Heart failure was categorized as NYHA functional class I to IV on the basis of symptoms during varying activities. The NYHA functional class III was divided into class IIIa and IIIb; patients in NYHA functional class IIIb were not able to cover 325 m during the 6-min walk test performed at the local heart failure clinics. Advanced heart failure was defined prospectively as NYHA functional class IIIb or IV.
Renal function was assessed as glomerular filtration rate was estimated on the basis of the simplified Modification of Diet in Renal Disease prediction equation and expressed as estimated glomerular filtration rate (eGFR) as follows:
Renal function was stratified according to glomerular filtration rate limits in the Kidney Disease Outcome Quality Initiative guidelines (10). Patients with eGFR >60 ml/min/1.73 m2 were stratified into chronic kidney disease (CKD) stage 1 and 2, despite lack of data on renal damage. Severe renal dysfunction was defined arbitrarily as eGFR ≤45 ml/min/1.73 m2, because the number of patients with CKD stage 4 and 5 were low. This was done prospectively to add statistical power to the severe renal failure group.
A high proportion of the population was already treated at baseline with diuretics, renin-angiotensin-aldosterone system (RAAS) blocking drugs, and beta-blockers. To differentiate between treatment intensity, daily dose equivalents were calculated. Daily doses of loop diuretics were expressed in furosemide equivalents (bumetanide 1 mg = furosemide 40 mg). Doses of angiotensin-converting enzyme inhibitors (ACEI) were expressed as enalapril equivalents/day (captopril 5 mg = ramipril 0.5 mg = lisinopril 1 mg = enalapril 1 mg). The ACEIs were more frequently used than angiotensin receptor blockers (ARBs), and patients using ARBs were excluded when analyzing ACEI daily dose in the baseline characteristics. Daily dose of beta-blocker was expressed as metoprolol equivalents (bisoprolol 1 mg = carvedilol 5 mg = atenolol 10 mg = metoprolol 20 mg).
Vascular disease was defined as previous stroke and/or peripheral arterial disease. The diagnosis of ischemic heart disease (IHD) as the cause of heart failure was based on clinical evaluation at the time of inclusion.
Continuous variables were presented as mean ± SD. Categorical data were presented as percentages. Student t test was used for continuous data when comparing the anemia and non-anemia groups. For the same purpose, chi-square test was used to compare categorical data. Analysis of variance was used to compare hemoglobin levels in different stages of CKD and NYHA functional class. Cox regression analyses were used to determine the association of anemia with all-cause mortality. Univariate Cox regression analyses were performed, and statistically significant predictors of mortality at baseline were entered into the multivariate Cox regression model to measure the multiple adjusted hazard ratio (HR) of baseline anemia on all-cause mortality. In addition to baseline anemia, statistically significant independent predictors of mortality at baseline were male sex, age, NYHA functional class, eGFR, body mass index (BMI), diabetes mellitus, and loop diuretic daily dose. These variables were entered in the following Cox models for multivariate adjustment, baseline values were used when analyzing baseline anemia, and last visit values were used when analyzing follow-up data. All included variables in the multivariate analyses are reported in the Results section.
A multivariate Cox regression model with hemoglobin as a continuous variable instead of the dichotomous variable anemia was performed in patients completing all visits to evaluate the prognostic implication of degree of anemia.
Kaplan-Meier plots were used to describe survival during the observation period in different patient groups. Log-rank statistics was used to identify differences between groups.
A logistic regression model was used to investigate differences between patients who attended all planned visits with those who did not. Independent predictors of mortality at baseline were entered in the model to differentiate the 2 groups.
Brain natriuretic peptide (BNP) and pro-BNP were not included in the multivariate analyses due to low number of valid data. IHD was the dominant cause of heart failure and was evaluated against all other etiologies of heart failure as a group.
The assumption of linearity of the logit of the outcome for continuous variables entered into the logistic models was checked (11). Serum sodium and systolic blood pressure did not meet the assumption of linearity; they were dichotomized with cutoffs near mean values. Likewise, ejection fraction did not meet the assumptions; here we dichotomized with a natural cutoff of ejection fraction <40%. Enalapril equivalent dose did not meet the assumption of linearity of the logit to outcome. The dichotomous variable use of ACEI or ARB, both blocking the RAAS, was entered in the multivariate logistic analysis. Spearman's correlation test was used to check for colinearity; no pair of independent variables was found to correlate to such a degree that they could not be simultaneously entered together in the regression analyses (all coefficients <0.43). For each Cox model the proportional hazard assumption was checked and found to be adequately met (11). Differences in HR between patients with severe renal failure and advanced heart failure compared with the rest of the heart failure population were checked with interaction analysis by product terms. Analyses on other potential interactions in the Cox analyses were performed with respect to age, sex, and ejection fraction; no significant interactions were found. Level of significance was set at 0.05. Analyses were performed with SPSS statistical software (version 15.0, SPSS, Inc., Chicago, Illinois).
A total of 4,144 patients were included in the analyses of the baseline characteristics (Table 1). The majority of patients were men, and 55% were older than 70 years. At baseline mean hemoglobin concentration was 13.8 ± 1.7 g/dl. Frequency of anemia defined by WHO criteria was 24.0%. Severe renal dysfunction (eGFR ≤45 ml/min/1.73 m2) was present in 752 patients (18.2%), of whom 31.0% had anemia (mean hemoglobin 12.9 ± 1.7 g/dl). Severe heart failure (NYHA functional class IIIb or IV) was present in 528 patients (13.0%), of whom 32.8% had anemia (mean hemoglobin 13.3 ± 1.8 g/dl).
Multivariate logistic regression model with baseline anemia as the dependent variable identified a number of variables independently associated with anemia (Table 2). Anemia was strongly associated with both renal function (p < 0.001) and heart failure severity defined by NYHA functional class (p < 0.001). Figure 1 describes the additive effect of worsening renal function and worsening NYHA functional class on hemoglobin levels.
During the median follow-up time of 28 months (range 0 to 94 months), 1,166 patients (28.1 %) died. Baseline anemia was a strong predictor of all-cause mortality (Fig. 2). Median survival in the whole population was 82 months. Patients with anemia had a median survival of 50 months versus 89 months in patients without anemia. Crude HR of anemia was 1.87 (95% confidence interval [CI]: 1.66 to 2.11; p < 0.001). After multivariate adjustment anemia remained a strong predictor of all-cause mortality, with HR of 1.30 (95% CI: 1.09 to 1.56; p = 0.004) (Table 3).
Patients with severe renal dysfunction (eGFR ≤45 ml/min/1.73 m2) and advanced heart failure (NYHA functional class IIIb or IV) were analyzed separately to investigate the prognostic impact of baseline anemia in patients with the highest mortality risk. By interaction analysis, HR of baseline anemia was significantly lower in patients with advanced heart failure compared with the rest of the heart failure population (HR: 1.23 vs. 1.96, p = 0.002), and anemia was not an independent prognostic variable of all-cause mortality in the 528 patients with NYHA functional class IIIb or IV (adjusted HR: 0.86; 95% CI: 0.56 to 1.33, p = 0.495) (Table 4). Likewise, the 752 patients with eGFR ≤45 ml/min/1.73 m2 had significantly lower HR of anemia compared with patients with higher eGFR (HR: 1.39 vs. 1.80, p = 0.022), and anemia was not an independent predictor of all-cause mortality in these patients (adjusted HR: 1.08; 95% CI: 0.77 to 1.51; p = 0.662) (Table 4).
Further interaction analyses did not show any difference in the prognostic information of baseline anemia in patients with mild renal dysfunction (eGFR >45 ml/min/1.73 m2 and ≤60 ml/min/1.73 m2) compared with patients without renal dysfunction (eGFR >60 ml/min/1.73 m2) (HR: 1.50 [95% CI: 1.20 to 1.86] vs. 1.87 [95% CI: 1.52 to 2.30]; p = 0.167). Likewise no interaction was found between NYHA functional classes IIIa and I to II (HR: 1.92 [95% CI: 1.55 to 2.38] vs. 1.85 [95% CI: 1.55 to 2.20]; p = 0.836). The prognostic impact of baseline anemia did not differ with respect to age, sex, or ejection fraction.
Less than 50% (n = 1,972) completed all 3 visits in the heart failure registry, of whom 1,743 patients had valid data on hemoglobin levels at the last visit. Baseline anemia was present in 361 patients (21.3%) in this subgroup of patients. Sustained anemia defined as anemia both at baseline and the last visit was present in 211 patients (12.1% of patients completing all visits), whereas in 41.8% of patients with initial anemia (n = 151), anemia had resolved at Visit 3. New-onset anemia was observed in 195 patients (14.2% of patients who were non anemic at baseline). Prevalence of anemia at Visit 3 was 23.3%.
In the patients who completed all visits, baseline anemia continued to be an independent predictor of mortality (adjusted HR: 1.33, 95% CI: 1.07 to 1.66, p = 0.011) and provided the same prognostic information as anemia at the last visit (adjusted HR: 1.31, 95% CI: 1.03 to 1.68, p = 0.030).
Completing all visits was associated with better survival from the date of the last visit (multiple adjusted HR: 0.67, 95% CI: 0.59 to 0.76, p < 0.001). A multivariate logistic regression model (R2 = 0.018, p < 0.001) identified patients completing all 3 visits to have less baseline anemia (odds ratio [OR]: 0.81, p = 0.013), lower NYHA functional class (OR: 0.81, p < 0.001), higher BMI (1 kg/m2 increase, OR: 1.03, p < 0.001) and higher age (1 year older, OR: 1.01, p = 0.008) compared with patients who did not complete all planned visits.
Risk of mortality was not significantly different in patients with transient anemia compared with patients who never presented with anemia (crude HR: 0.97, 95% CI: 0.66 to 1.43, p = 0.890). New-onset anemia was related to worse prognosis (crude HR: 1.61, 95% CI: 1.20 to 2.16, p = 0.002). After multivariate adjustment new-onset anemia did not predict worse outcome compared with patients with no anemia at any point (adjusted HR: 1.03, 95% CI: 0.73 to 1.46, p = 0.848).
Hemoglobin level predicted mortality only in patients with sustained anemia (HR: 0.51/1-g/dl increase, 95% CI: 0.41 to 0.64, p < 0.001), in contrast to patients without sustained anemia (HR: 0.97/1-g/dl increase, 95% CI: 0.88 to 1.06, p = NS), after adjustment for the variables used in Table 5.
Baseline anemia was common in this outpatient cohort of Norwegian patients with heart failure. The anemia prevalence of 24% is concordant with reports from other outpatient heart failure populations (6). Previous estimates have varied widely, ranging from 4% up to 61% (6). This variation is primarily due to different definitions of anemia and differences in study populations. Hospitalized patients seem to have the highest prevalence of anemia, whereas clinical trial cohorts have the lowest. The WHO definition of anemia used in the present study is the most frequently used definition of anemia in heart failure patients in recent years and is the recommended definition (4). Anemia in women was defined as hemoglobin levels below 12 g/dl, compared with 13 g/dl in men, although the great majority of female patients were post-menopausal. By interaction analyses, the prognostic impact of anemia did not differ between men and women, despite the difference in definition of anemia.
Baseline anemia was strongly associated with mortality in our population. Median survival in patients without anemia was more than 3 years longer than in anemic patients. Our crude 87% increased all-cause mortality risk among patients with anemia was in agreement with findings of a recent meta-analysis on the topic (5). Thus, anemia—independent of cause—identified patients with a poor prognosis and should alert the clinician to the need for intensified care. After adjustment for a large range of confounders, anemia remained a strong independent prognostic risk factor.
The most striking finding in our study was that baseline anemia was not an independent predictor of all-cause mortality in the patients with severe renal dysfunction or the most advanced heart failure. Impaired renal function and increasing NYHA functional class both independently predicted more serious prognosis in heart failure patients (12) and were closely associated with anemia. Furthermore, these conditions share many potential etiologies for anemia (1,13). Groenveld et al. (5) found an inverse relationship between serum creatinine levels and the effect of anemia on mortality with meta-regression analysis. The effect on mortality declined with higher serum creatinine levels; this might reflect that—at higher creatinine levels—renal disease accounts for a greater proportion of the mortality risk than anemia (6). However, Go et al. (14) identified hemoglobin level as an independent predictor of outcomes at all levels of kidney function. The uncertainty of the prognostic impact of anemia in heart failure patients with renal dysfunction motivated us to perform selected analyses in patients with the most severe renal dysfunction and most advanced heart failure. Unfortunately, data regarding an eventual progression of comorbid conditions were not available; therefore the reason why baseline anemia did not predict all-cause mortality in the patients with severe renal dysfunction and advanced heart failure was not addressed.
Although baseline anemia did not affect mortality in the patients with advanced heart failure and severe renal dysfunction in our study, our data did not preclude that they suffer a high symptom burden from their anemia, and specific anemia treatment could be beneficial with regard to symptom relief (15–17).
Patients who completed all planned visits had a better overall prognosis, and our results indicate that these patients had less-serious disease beyond NYHA functional classification, anemia state, age, BMI, and the other prognostic variables corrected for in our statistical models. Furthermore, the interaction with respect to mortality between anemia and advanced heart failure and severe renal function was not present in the patients who completed all planned visits in the registry. Therefore, patients completing Visit 3 should be regarded as a selected population, not necessarily representative of outpatients with heart failure. Medication adherence is shown to be related to survival in heart failure patients (18), and part of the bad prognosis in patients not completing all visits in our study might be explained by nonadherence. Alternatively, they were deprived of the opportunity to have the treatment of heart failure optimized, because they did not complete all clinical visits.
Anemia was not necessarily persistent; a high proportion of patients who were anemic at baseline had normal hemoglobin levels at follow-up. Likewise, new-onset anemia was prevalent. Patients with transient anemia did not have significantly higher mortality rates compared with patients without anemia at any visit. Our data are supported by Tang et al. (19); initial anemia in heart failure patients resolved in a large proportion of the patients, and these patients did not pose greater long-term risk compared with patients without anemia at baseline. However, new-onset anemia is shown to predict outcome in outpatients with heart failure (19,20). Komajda et al. (20) found that a large decrease in hemoglobin level during follow-up independently predicted mortality. Our data confirmed that patients with new-onset anemia had poorer prognosis than patients without anemia at any time, but after multivariate correction new-onset anemia was not an independent predictor of all-cause mortality. This could be due to lack of statistical power, because the number of patients was low in this subgroup.
Sustained anemia, however, involving 12.1% of our patients completing all visits, independently predicted mortality, and no interaction with severe renal dysfunction and advanced heart failure were found in this selected subgroup of patients who completed all planned visits. This might reflect that the cut point for anemia in heart failure patients should not follow the WHO criteria but rather be slightly lower, but alternative explanations might exist.
Lower baseline hemoglobin levels in the patients with sustained anemia were associated with increased mortality; thus the degree of anemia would probably identify patients at especially high risk of mortality. This association was not found in the nonanemic patients and among patients with transient anemia. Go et al. (14) identified increased risk of death in heart failure patients with hemoglobin levels below 13 g/dl. Thus, our data did not support specific anemia treatment goal to exceed 13 g/dl.
The clinical implications of our findings might be important. Baseline anemia was not an independent predictor of all-cause mortality in heart failure patients with the worst prognosis (i.e., advanced NYHA functional classes and severe renal failure). In the selected population who were able to attend subsequent visits, sustained anemia after optimizing conventional heart failure treatment seemed to predict mortality independently of renal function and NYHA functional class. Such patients would be suitable for randomized controlled trials that examine specific anemia treatment in addition to conventional optimal treatment and intense follow-up. However, if these trials would show improved survival of specific anemia treatment, it is uncertain whether the results would be valid and applicable for an unselected outpatient population with heart failure.
This study presented data from an existing heart failure registry. We were not able to evaluate the quality of the data, and analyses were restricted to existing data in the registry. The number of patients was large, but subgroup analyses might have been underpowered. Less than 50% of patients had evaluable follow-up data. Conclusions on cause and effect were not possible with this study design. Data on hospital stay would have given important additional outcome information. Such analyses were not possible, due to the low number of valid data on hospital stay. Furthermore, adequate data on follow-up symptoms, BNP/pro-BNP, and device therapy would have supplemented our data concerning outcome. Arbitrary categorization of advanced heart failure and severe renal function were used, because a low number of patients presented with NYHA functional class IV or CKD class 4 and 5. Continuous worsening prognosis with decreasing eGFR and increasing NYHA functional classes in our material supported the arbitrary classification, which was undertaken to increase statistical power in the subgroup analyses.
Baseline anemia was predictive of all-cause mortality in this study of outpatients with chronic heart failure; however, this was not seen in those with advanced symptoms of heart failure or severe renal failure at baseline. In the selected population with valid follow-up data, sustained anemia after optimizing heart failure treatment implied worse prognosis independently of renal function and NYHA functional class.
For acknowledgments of participating institutions, please see the online version of this article.
All authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- angiotensin-converting enzyme inhibitor
- angiotensin receptor blocker
- body mass index
- brain natriuretic peptide
- confidence interval
- chronic kidney disease
- estimated glomerular filtration rate
- hazard ratio
- ischemic heart disease
- New York Heart Association
- odds ratio
- renin-angiotensin-aldosterone system
- World Health Organization
- Received April 29, 2011.
- Revision received September 5, 2011.
- Accepted October 3, 2011.
- American College of Cardiology Foundation
- Anand I.S.
- van der Meer P.,
- Lipsic E.,
- Westenbrink B.D.,
- et al.
- Groenveld H.F.,
- Januzzi J.L.,
- Damman K.,
- et al.
- Grundtvig M.,
- Gullestad L.,
- Hole T.,
- Flonaes B.,
- Westheim A.
- Remme W.J.,
- Swedberg K.
- World Health Organization
- Katz M.H.
- Kazory A.,
- Ross E.A.
- Go A.S.,
- Yang J.,
- Ackerson L.M.,
- et al.
- Ponikowski P.,
- Anker S.D.,
- Szachniewicz J.,
- et al.
- Tang W.H.,
- Tong W.,
- Jain A.,
- Francis G.S.,
- Harris C.M.,
- Young J.B.
- Komajda M.,
- Anker S.D.,
- Charlesworth A.,
- et al.