Author + information
- Received May 14, 2007
- Revision received July 20, 2007
- Accepted July 23, 2007
- Published online February 5, 2008.
- W.H. Wilson Tang, MD, FACC⁎,1,⁎ (, )
- Wilson Tong, MSc⁎,
- Anil Jain, MD†,‡,
- Gary S. Francis, MD, FACC⁎,1,
- C. Martin Harris, MD, MBA‡ and
- James B. Young, MD, FACC⁎,1
- ↵⁎Reprint requests and correspondence:
Dr. W. H. Wilson Tang, Department of Cardiovascular Medicine, Cleveland Clinic, 9500 Euclid Avenue, Desk F25, Cleveland, Ohio 44195.
Objectives This study sought to determine the characteristics and long-term prognosis of anemia in ambulatory patients with chronic heart failure.
Background Anemia is prevalent in heart failure, and may portend poor outcomes.
Methods We reviewed 6,159 consecutive outpatients with chronic stable heart failure at baseline, short-term (3-month) follow-up, and long-term (6-month) follow-up between 2001 and 2006. Clinical, demographic, laboratory, and echocardiographic data were reviewed from electronic medical records. Mortality rates were determined from 6-month follow-up to end of study period.
Results Prevalence of anemia (hemoglobin [Hb] <12 g/dl for men, <11 g/dl for women) was 17.2% in our cohort. Diabetes, B-natriuretic peptide, left ventricular ejection fraction, and estimated glomerular filtration rate were independent predictors of baseline anemia. Documented evaluation of anemia was found in only 3% of all anemic patients, and better in internal medicine than in cardiology clinics. At 6-month follow-up, new-onset anemia developed in 16% of patients without prior anemia, whereas 43% patients with anemia at baseline had resolution of their hemoglobin levels. Higher total mortality rates were evident in patients with persistent anemia (58% vs. 31%, p < 0.0001) or with incident anemia (45% vs. 31%, p < 0.0001) compared with those with without anemia at 6 months.
Conclusions These observations in a broad unselected outpatient cohort suggest that anemia in patients with heart failure is under-recognized and underevaluated. However, resolution of anemia was evident in up to 43% of patients who presented initially with anemia, and did not pose greater long-term risk for all-cause mortality. However, the presence of persistent anemia conferred poorest survival in patients with heart failure when compared with that of incident, resolved, or no anemia.
Anemia has become increasingly recognized as an important factor in the development and progression of heart failure. Anemia occurs with surprisingly high prevalence and portends a poor prognosis (1–3). In a cohort of 1,061 patients with advanced heart failure referred for transplant evaluation, lower hemoglobin (Hb) was associated with greater functional impairment, worse exercise capacity, and increased mortality (4). However, most analyses have been limited to data sets from hospitalized patients (5), post-hoc analyses of clinical trials with primarily baseline measurements (6), or interrogation of large administrative databases (7). Bias of ascertainment in these populations can limit the applicability of their observations (8). Furthermore, data regarding the changes in Hb levels in patients over time and the impact on prognoses remain limited.
Because of the implementation of an electronic medical record system in our practice since 2001, we have the ability to characterize anemia in an unselected cohort of ambulatory patients with a clinical diagnosis of chronic heart failure seen in internal medicine and cardiology clinics. The objective of this paper is to determine the prevalence and characteristics of anemia in ambulatory patients with chronic heart failure and to identify the impact of changes in anemia status over time to long-term survival across patients with varying degrees of renal insufficiency and left ventricular systolic function.
The Cleveland Clinic Institutional Review Board approved this study protocol. We reviewed our heart failure registry compiled from an electronic medical record system (EpicCare, Epic Systems Corp., Madison, Wisconsin), identified 6,159 consecutive ambulatory patients seen at the Cleveland Clinic internal medicine or cardiology outpatient clinics with a documented clinical diagnosis of heart failure, and performed a complete blood count test between January 2001 and April 2006. The EpicCare system provides a choice of common diagnoses with corresponding International Classification of Diseases-Ninth Revision (ICD-9) codes to document clinical problems in clinical chart notes. We therefore confirmed the diagnosis by performing a search for the ICD-9 codes related to heart failure to establish our study cohort: 242.9, 398.91, 402.01, 402.11, 402.91, 404.01, 404.03-4, 404.13, 404.91, 404.93, 425.1, 425.4-5, 425.7-9, 428.0-1, and 428.9. Patients must also have had documented left ventricular ejection fraction (LVEF) data within 100 days of the baseline Hb blood analysis for inclusion in the study.
We excluded patients with a history of congenital heart disease, malignancies or inherited or acquired anemia (ICD-9 code 282.x to 285.x), or documented active gastrointestinal bleeding at the time of blood draw. If a patient was seen at both internal medicine and cardiology clinics, he or she would be classified as being followed up in the cardiology clinic.
Clinical, demographic, laboratory, and documented primary and secondary diagnoses in the electronic medical record were reviewed. All clinical data were matched to within 100 days of baseline Hb blood analysis, with the first occurrence defined as baseline. Body mass index and body surface area were derived from height and weight measurements. Estimated glomerular filtration rate (eGFR) was calculated through the Modification of Diet in Renal Disease equation (9). Long-term all-cause mortality data were determined from electronic medical records and validated by the social security death index. Death and cardiac transplantation dates were used in survival analysis (censoring transplantation), and all patients were followed up until June 20, 2007.
We defined anemia using a conservative criteria: Hb <12 g/dl for men and <11 g/dl for women. Standard criteria for microcytic (mean corpuscular volume [MCV] <82 fl) and macrocytic (MCV >98 fl) anemia were used. For those with anemia identified by blood work, diagnostic ICD-9 coding for anemia (280.x to 281.x), and standard evaluation tests (including iron studies, reticulocyte count, vitamin B12 and folate levels, and gastrointestinal bleeding workup) were identified.
Short-term follow-up was defined as the next available Hb measurement within 3 months from baseline, which results in a mean 0.6 ± 0.8 months of follow-up. Patients must also have had consecutive Hb measurements within 100 days of the 6-month follow-up to be included in the long-term follow-up (mean follow-up at 5.3 ± 1.8 months). Incidence of (new-onset) anemia was defined as the presence of anemia at short-term (3-month) or long-term (6-month) follow-up in those patients without anemia at baseline. Resolution of anemia was classified as normalization of Hb levels (≥12 g/dl for men and ≥11 g/dl for women) with at least improvement in ≥0.5 g/dl at follow-up in those who presented with anemia at baseline. Persistent anemia was said to be present if patients had anemia at both baseline and at short-term or long-term follow-up. Proportions of incidence, resolution, and persistence of anemia were determined relative to the total number of patients at follow-up. Changes in Hb levels over the course of follow-up within the study period were also determined.
All echocardiograms and laboratory testing were performed at the Cleveland Clinic and reported in EpicCare. Echocardiographic left atrial diameter and area, interventricular septal thickness, posterior wall thickness, left ventricular end-diastolic diameter, and LVEF were collected. The LVEF was determined by the Simpson rule, whereas left atrial area was determined from the area–length method as a gross surrogate of diastolic function (10). The LV mass was calculated using the American Society of Echocardiography equation, defined as 0.8 ([(interventricular septal thickness + posterior wall thickness + left ventricular end-diastolic diameter)3 − left ventricular end-diastolic diameter3] + 0.6) adjusted for gender (10). Standard LV diastolic filling pattern was graded on a scale from I to IV based on mitral inflow E- and A-wave velocity profiles described previously (11,12). Mitral regurgitation and tricuspid regurgitation severity was evaluated on a scale from 0 to 4+, with 0 representing normal and 4+ signifying severe regurgitation by color Doppler inflow patterns. Right ventricular systolic pressure was determined from estimated jugular venous pressure and maximum tricuspid regurgitation velocity using the modified Bernoulli equation.
Results are presented in percentages (for dichotomous variables) or mean ± standard deviation (for continuous variables). Univariable analyses were performed using 2-tailed chi-square and Student t tests, respectively. Multivariable logistic regression with forward stepwise selection was performed to determine the predictors of baseline prevalence of anemia in our cohort (variables included baseline age, gender, hypertension, diabetes, MCV, medications (angiotensin-converting enzyme [ACE] inhibitors/angiotensin receptor blockers, beta-blockers, loop diuretics), B-type natriuretic peptide (BNP), eGFR, LVEF, and left atrial area). Trends for the prevalence of anemia across different predictors were assessed with the Cochran-Armitage test. Linear univariate correlation assessed relations between changes in Hb and changes in clinical status. Multivariate correlation to changes in Hb used a stepwise model followed by standard least-squares fit. Survival was analyzed between patients with incident, resolved, or persistent anemia versus those without anemia using Kaplan-Meier analysis. Long-term mortality data over a mean follow-up of 3.9 ± 2.2 years were used to calculate time to death or end of follow-up (censoring cardiac transplantation) in survival analysis. Multivariate Cox proportional hazards model was used to calculate risk ratios for independent predictors of mortality with incremental increases in continuous variables. A p value of <0.05 was considered statistically significant. All statistical analyses were performed using SAS version 9.1 and JMP version 5.1 (SAS Institute, Cary, North Carolina).
A total of 6,159 outpatients with a clinical diagnosis of heart failure met the study inclusion and exclusion criteria (including 3,360 and 2,796 patients initially presenting in internal medicine and cardiology clinics, respectively). The mean baseline Hb level was 14 ± 2 g/dl in men and 13 ± 2 g/dl in women in the overall study cohort. The majority of our patients in the cohort had been treated with ACE inhibitors, beta-blockers, and loop diuretics at baseline (Table 1). The median LVEF was 30%, and median plasma BNP level was 325 pg/ml for those with available data at baseline (n = 2,128). Among the overall study cohort, 1,393 patients (23%) had complete blood count data available at both short-term (3-month) and long-term (6-month) follow-up time points (Fig. 1).
Prevalence of anemia
At baseline, the prevalence of anemia was 17.2% (17.8% in men, 16.0% in women, n = 6,159) (Fig. 1). At follow-up, the prevalence of anemia at baseline in the subgroup was 23.2% (n = 1,393), similarly when assessed at 3- and 6-month follow-up. Baseline characteristics between patients with versus without anemia are shown in Table 1. Patients with anemia were older, had more prevalent diabetes mellitus and heart failure medication usage; had higher mortality rates, BNP, and blood urea nitrogen; and had lower eGFR, low-density lipoprotein, and more severe diastolic dysfunction (both diastolic stage and left atrial area as surrogate measure). A higher prevalence of anemia was associated with male gender, particularly at age >75 years (Fig. 2). Patients with moderately impaired renal function (eGFR <60 ml/min/1.73 m2), more preserved LVEF above median (≥30%), and above median plasma levels of BNP (>325 pg/ml) had significantly higher prevalence of anemia (Fig. 3). These factors were found to be independent predictors of the prevalence of anemia in heart failure patients after multivariate logistic regression analyses (Table 2).
Patterns of diagnostic evaluation and treatment of anemia
The majority (71.6%) were normocytic (MCV 82 to 98 fl). Microcytic (MCV <82 fl) and macrocytic (MCV >98 fl) anemia were observed in 19.6% and 8.8% of patients, respectively, with more microcytic anemia in women (24.9% vs. 17.0%, p = 0.0001), whereas macrocytic anemia had similar prevalence between genders (6.9% vs. 9.7%, p = 0.082). Using ICD-9 codes and standard laboratory workup orders for anemia, only 3.0% of patients with heart failure had documented anemia evaluation. Of interest, anemia was detected and evaluated more often in patients presenting to internal medicine clinics than cardiology clinics (18.1% vs. 16.1% from laboratory data, p < 0.0001; 3.6% vs. 2.3% from documentation of diagnosis in problem list, p = 0.0002), although patients seen in internal medicine clinics were older, were more often women, and were more likely to have preserved LV function. As expected, patients with anemia were more likely to be treated with iron supplementation (21% vs. 7%, p < 0.0001) or erythropoiesis-stimulating agents (8% vs. 1%, p < 0.0001). Careful chart review showed that those on anemia treatment in the nonanemia group had a prior diagnosis of anemia or renal insufficiency in their medical records (data not shown). However, nearly 70% of patients who fulfilled the criteria for anemia did not have a documented follow-up complete blood count in the electronic medical record within the 6 months of follow-up after the initial low Hb level.
Changes in anemia status and Hb levels over time
Follow-up patients had consecutive Hb data at 3- and 6-month follow-up (n = 1,393). In the 6-month follow-up cohort, 210 patients (15.1% of cohort, or 19.6% of patients with no anemia at baseline) developed new-onset anemia, 143 patients (10.3% of cohort, or 43.0% of patients with anemia at baseline) had resolution of their Hb levels, and 180 patients (12.9% of cohort, or 55.7% of patients with anemia at baseline) had persistent anemia (Fig. 1). Mean changes in Hb levels over 6 months were −0.32 ± 1.4 g/dl for those without anemia, +2.44 ± 1.7 g/dl for those with resolved anemia, −2.59 ± 1.4 g/dl for those with incident anemia, and −0.04 ± 1.4 g/dl for those with persistent anemia. All cohorts, except the persistent anemia cohort (p = 0.77), had significant changes in Hb at 6-month follow-up from baseline (p < 0.0001 for the other 3 subgroups). Changes in clinical and echocardiographic variables identified from Table 1 were not correlated with changes in Hb (r < 0.10 for all). However, patients with LVEF ≥30% (median LVEF), BNP >325 pg/ml (median plasma levels of BNP), eGFR <60 ml/min/1.73 m2, or diabetes mellitus were more likely to have persistent anemia rather than no anemia or transient anemia (Fig. 3). Patients with eGFR <60 ml/min/1.73 m2 also had a higher rate of incident anemia.
During a mean follow-up of 3.9 ± 2.2 years, 1,827 patients (30%) died. Incremental changes in Hb levels were independently attributed to higher likelihood of survival after adjusting for LVEF and eGFR (Table 3). At baseline, patients presenting with anemia had a significantly worse prognosis than those without anemia, with an estimated 18.6% increased absolute risk of death at 3 years (overall mortality 47% vs. 26%, p < 0.0001) (Fig. 4). In the subset of patients with follow-up Hb levels, patients with persistent anemia still had the worst survival relative to those with new-incident, resolved, or no anemia at short-term follow-up (chi-square = 81.2) (Fig. 4). At 6 months, patients with persistent anemia or incident anemia had higher all-cause mortality risk than patients without anemia at baseline (all-cause mortality 58% vs. 45% vs. 31%, respectively) (Table 4). The differentiation in survival between these 4 different cohorts became more significant at 6 months (Fig. 4) (chi-square = 81.2) than at 3 months, further showing the greater differentiation in survival over time (Table 4).
The key finding in our single-center heart failure cohort is that anemia in patients with chronic heart failure has been underrecognized and underevaluated in clinical practice. Meanwhile, careful follow-up in patients with sequential evaluation of their Hb levels showed that a relatively large cohort (up to 43% of patients with anemia detected in a single time point at baseline) may proceed with resolution of their Hb to their nonanemic levels over time. Particularly reassuring was the fact that these patients with so-called transient anemia may show equivalent long-term outcomes when compared with those without any anemia. Our data also confirmed prior studies showing that persistent anemia in patients with mild-to-moderate heart failure confers a higher risk for mortality compared with patients with resolved, new-onset, or no anemia.
Although many heart failure studies have shown that anemia (often determined at a single time point) was associated with worse prognosis (3,13–16) , few have explored the impact of longitudinal changes in anemia status as well as the consequences of these changes on clinical outcomes. Our findings complement prior observations from post hoc analysis of the Val-HeFT (Valsartan in Heart Failure Trial) and COMET (Carvedilol or Metoprolol Evaluation Trial), showing patients with the largest decrease in Hb levels fared the worse in hospitalization, morbidity, and mortality (17). The fact that a substantial proportion of patients had transient anemia (despite the fact that only 21% were on iron supplementation and 8% were on erythropoiesis stimulating agents) may imply the heterogeneity of the patient population when anemia was determined only at a single time point. Furthermore, our data showed that patients with no or transient anemia have survival rates significantly better than those of patients with persistent anemia. This finding may imply that a single-time-point evaluation of Hb may not reflect the true prognostic value of low Hb levels because resolution of anemia may occur especially when the development of anemia at that time point was influenced by other concomitant and potentially reversible conditions. Because only a minority of patients received treatment for anemia, it is conceivable that anemia in a substantial number of patients may have resolved after standard heart failure management, particularly with better control of fluid status (leading to less dilutional anemia) and neurohormonal antagonism. Indeed, a recent publication has also challenged the adverse contribution of a diagnosis of anemia at the time of new-onset diagnosis of heart failure to long-term clinical outcomes because the prognostic value of anemia became less apparent when adjusted for other known prognostic variables (age, diastolic blood pressure, serum creatinine, New York Heart Association functional class, and left ventricular function) (18). We were unable to identify reliable predictors of transient versus persistent anemia, therefore careful monitoring and evaluation of sequential Hb levels over time remains the recommended approach.
In our overall study cohort of 6,159 patients, the 17% to 23% prevalence of anemia is consistent with that reported in the literature (4). The majority of patients with anemia (70%) showed normocytic indices, which may influence the perceived need for a more aggressive workup. Up to one-fifth of our anemia patients had microcytic anemia, and many patients may have specific reasons for anemia (such as iron deficiency anemia, thalassemia, renal insufficiency, or occult gastrointestinal bleeding) even with normal erythrocyte volumes. The most likely explanation is still the fact that Hb levels have not been commonly recognized as a relevant measurement in the management of heart failure until recently. This is particularly apparent in cardiology clinics, where the primary focus of managing heart conditions has distracted from evaluating or even establishing a diagnosis of anemia.
The reported incidence of new-onset anemia (12% to 15% within 6 months) in patients with heart failure deserved further discussion and investigation. Excessive ACE inhibition may suppress the degradation of inhibitors of hematopoiesis, resulting in iatrogenic anemia (19). The majority of our patient population had already been receiving ACE inhibitors for some duration, therefore the impact of an ACE inhibitor-induced reduction in Hb levels may not be as relevant (17,20). Nevertheless, these data were comparable with the 14.1% incidence of anemia at year 1 from COMET using less stringent anemia criteria (Hb <13 g/dl for men and <12 g/dl for women) (21). Regardless, the number of patients who may still have persistent anemia or may have developed new-onset anemia within 3 to 6 months of initial evaluation (approximately 30%) implies that sequential evaluation of complete blood count may be clinically useful. As expected, anemia is also more common in patients with diabetes mellitus, who may also have more cardiac and renal compromise. Patients with higher plasma BNP levels also were expected to have low Hb levels (22,23). Better understanding of the factors leading to the development of anemia in heart failure will likely provide potential insights into disease mechanisms and potential therapeutic targets. In particular, whether specific therapy targeting optimal Hb levels (such as the use of erythropoiesis-stimulating agents) can benefit long-term survival and affect short-term clinical improvement (24–26) remains to be proven in large multicenter randomized controlled trials, particularly with recent studies in the chronic kidney disease population that have failed to show significant benefits with the use of erythropoietin therapy to achieve higher Hb targets (27,28).
The strength of our data from electronic medical records lies in our ability to easily track our patient cohort over time in real clinical practices. This study shows the power of using electronic medical records as a unique and novel way to further our understanding of practice patterns as well as disease characteristics and outcomes. Detailed information such as anemia subtypes (such as microcytic vs. macrocytic), concomitant conditions (such as renal insufficiency), and diagnostic workup patterns (or the lack of follow-up evaluation) can be determined directly from the clinical laboratory data of the electronic medical records. The striking contrast between the documented diagnosis of anemia and the documented Hb levels can illustrate some advantages of this research strategy over administrative database interrogation. In turn, future quality assurance measures can be implemented directly into the electronic medical record system to improve diagnostic and coding accuracy via reminder systems or automatic coding algorithms.
Our study is limited by its retrospective design, and the mandate to have an available complete blood count assessment (at baseline and especially at pre-defined intervals) may not represent the true prevalence of an unselected heart failure population. Furthermore, a single cutoff value (even with gender-specific definitions) may limit the reliability of the estimation of the prevalence and incidence rates by values that were borderline, even when an ≥0.5 g/dl increase in Hb levels was instituted to increase the reliability of assessing the changes in the follow-up Hb levels. The lack of consistent objective measurements of fluid overload as well as physician-assessed or self-reported functional status (such as New York Heart Association functional classification) at specific time intervals may have limited the accuracy of determining concomitant changes in clinical status with changes in Hb over time. The arbitrary criteria set for the follow-up durations may have limited the sample size of the follow-up subgroups, and can introduce selection bias in favor of those for whom follow-up evaluation were deemed necessary (which may have explained the higher prevalence of anemia at follow-up). The practice patterns may underestimate the true rates of evaluation because patients may seek care outside our hospital system, although the majority of the patients in this analysis have received long-term follow-up in internal medicine and cardiology clinics. It is also difficult to adjust for variations in treatment effects based on sequential Hb levels because the small subgroup of patients undergoing iron or erythropoietin treatment were initiated on therapy only after the follow-up time interval, and a definitive diagnosis of iron deficiency was not available in a large number of individuals with anemia at baseline even with microcytic anemia. Anemia with normal versus low iron may have different relationships to concurrent hemodynamic measures. Regression of the mean may have contributed to our observations, particularly with the higher resolution rate in patients with eGFR <60 ml/min/1.73 m2, as the prevalence and incidence of anemia were substantially higher in this subgroup. Further studies are needed to determine how to improve the diagnostic workup of anemia for this patient population and how to identify those who may benefit from specific therapies targeting anemia associated with heart failure.
In our single-center cohort of ambulatory patients with a clinical diagnosis of heart failure with available complete blood count assessment, anemia was underrecognized and underevaluated, especially in cardiology clinics. Resolution of anemia was evident in up to 43% of patients who presented initially with anemia, and did not pose greater long-term mortality risks. However, the presence of persistent anemia conferred poorer survival in patients with heart failure when compared with that of incident, resolved, or no anemia.
↵1 Drs. Tang, Francis, and Young are consultants for Amgen, Inc.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- B-type natriuretic peptide
- estimated glomerular filtration rate
- International Classification of Diseases-Ninth Revision
- left ventricular ejection fraction
- mean corpuscular volume
- Received May 14, 2007.
- Revision received July 20, 2007.
- Accepted July 23, 2007.
- American College of Cardiology Foundation
- Horwich T.B.,
- Fonarow G.C.,
- Hamilton M.A.,
- MacLellan W.R.,
- Borenstein J.
- Ezekowitz J.A.,
- McAlister F.A.,
- Armstrong P.W.
- Mozaffarian D.,
- Nye R.,
- Levy W.C.
- McCarthy P.M.,
- Young J.B.
- Lang R.M.,
- Bierig M.,
- Devereux R.B.,
- et al.
- Garcia M.J.,
- Thomas J.D.,
- Klein A.L.
- Go A.S.,
- Yang J.,
- Ackerson L.M.,
- et al.
- Tang Y.D.,
- Katz S.D.
- Anand I.S.,
- Kuskowski M.A.,
- Rector T.S.,
- et al.
- van der Meer P.,
- Lipsic E.,
- Westenbrink B.D.,
- et al.
- Ishani A.,
- Weinhandl E.,
- Zhao Z.,
- et al.
- Komajda M.,
- Anker S.D.,
- Charlesworth A.,
- et al.
- Silverberg D.S.,
- Wexler D.,
- Sheps D.,
- et al.
- Silverberg D.S.,
- Wexler D.,
- Blum M.,
- et al.
- Mancini D.M.,
- Katz S.D.,
- Lang C.C.,
- LaManca J.,
- Hudaihed A.,
- Androne A.S.