Author + information
- Received July 26, 2005
- Revision received November 18, 2005
- Accepted November 21, 2005
- Published online May 16, 2006.
- Grace L. Smith, MD, MPH⁎,1,
- Judith H. Lichtman, PhD, MPH†,
- Michael B. Bracken, PhD, MPH†,
- Michael G. Shlipak, MD, MPH§∥,
- Christopher O. Phillips, MD, MPH¶,
- Paul DiCapua, BS⁎ and
- Harlan M. Krumholz, MD, SM, FACC⁎,†,‡,#,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Harlan M. Krumholz, Robert Wood Johnson Clinical Scholars Program, Yale University School of Medicine, Sterling Hall of Medicine (SHM), I-Wing, Suite 456, 333 Cedar Street, New Haven, Connecticut 06520.
Objectives We estimated the prevalence of renal impairment in heart failure (HF) patients and the magnitude of associated mortality risk using a systematic review of published studies.
Background Renal impairment in HF patients is associated with excess mortality, although precise risk estimates are unclear.
Methods A systematic search of MEDLINE (through May 2005) identified 16 studies characterizing the association between renal impairment and mortality in 80,098 hospitalized and non-hospitalized HF patients. All-cause mortality risks associated with any renal impairment (creatinine >1.0 mg/dl, creatinine clearance [CrCl] or estimated glomerular filtration rate [eGFR] <90 ml/min, or cystatin-C >1.03 mg/dl) and moderate to severe impairment (creatinine ≥1.5, CrCl or eGFR <53, or cystatin-C ≥1.56) were estimated using fixed-effects meta-analysis.
Results A total of 63% of patients had any renal impairment, and 29% had moderate to severe impairment. After follow-up ≥1 year, 38% of patients with any renal impairment and 51% with moderate to severe impairment died versus 24% without impairment. Adjusted all-cause mortality was increased for patients with any impairment (hazard ratio [HR] = 1.56; 95% confidence interval [CI] 1.53 to 1.60, p < 0.001) and moderate to severe impairment (HR = 2.31; 95% CI 2.18 to 2.44, p < 0.001). Mortality worsened incrementally across the range of renal function, with 15% (95% CI 14% to 17%) increased risk for every 0.5 mg/dl increase in creatinine and 7% (95% CI 4% to 10%) increased risk for every 10 ml/min decrease in eGFR.
Conclusions Renal impairment is common among HF patients and confers excess mortality. Renal function should be considered in risk stratification and evaluation of therapeutic strategies for HF patients.
Renal impairment in heart failure (HF) patients has become increasingly recognized as an independent risk factor for morbidity and mortality (1–6). However, the precise magnitudes of these risks are not known. Given the growing incidences of HF and chronic kidney disease, both exceptionally costly and morbid conditions, it is important to understand the magnitude of risks in this patient group (7,8). Clarifying the contribution of renal impairment to outcomes in HF could provide more precise risk stratification and prognostication and, ultimately, the development of optimal therapeutic strategies in these patients (6).
We performed a systematic review and meta-analysis of published literature to characterize the prevalence of renal impairment in community-based, clinical trial, and hospitalized HF patients. Additionally, we sought to estimate the all-cause mortality risks associated with renal impairment, as determined by serum creatinine, cystatin-C, calculated creatinine clearance (CrCl), or estimated glomerular filtration rate (eGFR). Furthermore, we evaluated the association between renal impairment and other adverse outcomes such as early mortality (≤6 months), cardiovascular mortality, hospital admissions, and functional status. Finally, we explored whether published studies have identified higher risk subgroups (according to demographic or clinical characteristics) that may have particularly high mortality risks.
To identify eligible studies, we searched MEDLINE (1966 through May 2005) accessed by PubMed, using keywords including (heart failure [MeSH] ± congestive) and ((renal and (impairment or dysfunction or function)) or chronic kidney (disease or failure [MeSH])), limited to English language. Additionally, we identified potentially relevant studies using a manual search of references from all eligible studies, review articles, and Science Citation Index Expanded on Web of Science (1945 to May 2005).
We included cohort studies and retrospective secondary analyses of randomized controlled trials whose primary objective was to analyze the association between renal impairment and mortality in HF patients. Titles and abstracts of all articles identified by the search strategy were evaluated and rejected on initial screen if they: 1) included subjects other than HF patients; 2) did not define HF using a combination of symptoms and signs (dyspnea, peripheral edema, third heart sound, jugular venous distention, limited exercise tolerance, or documented New York Heart Association [NYHA] functional class), imaging (documented pulmonary congestion or impaired ejection fraction), or medical record diagnosis by International Classification of Diseases-9; 3) had no evaluation of renal function; 4) did not include all-cause mortality as a primary outcome; 5) were published only in abstract form, because validity and quality of methods could not be adequately assessed and scored; or 6) included patients ≤18 years old. A Quality of Reporting of Meta-analyses (9) flow diagram illustrates the study selection process (Fig. 1).Excluded were studies with <6 months follow-up (10,11) and a study that defined renal impairment using International Classification of Diseases-9 code but no direct serum measures (12).
Quality assessment and data abstraction
The quality of each study was evaluated according to the guidelines developed by the United States Preventive Task Force (13) and the Evidence-Based Medicine Working Group (14). We assessed for the following seven characteristics: 1) the inclusion of possible confounders in multivariable analysis; 2) recording of loss to follow-up rates; 3) blind and 4) consistent method of assessing renal function and outcomes in all risk groups; 5) follow-up of at least six months; 6) temporality (renal function measured at baseline, not at time of outcomes assessment); and 7) dose-response evaluation (assessment over a spectrum of renal function). Studies were graded as good quality if they met 5 to 7 criteria; fair if they met 2 to 4; and poor if they met ≤1 criterion.
Measures of renal function included serum-measured creatinine and cystatin-C or creatinine-based estimated CrCl using the Cockcroft-Gault equation (15) and eGFR using the Modification of Diet in Renal Disease equation (16). Renal impairment definitions were based on categorizations available in the published studies rather than on empiric re-categorization of raw data, which were not available for this analysis. Any (mild to severe) renal impairment was defined as all categories of renal function other than the baseline category (lowest tertile, quartile, and so on), which corresponded to creatinine >1.0 mg/dl, CrCl or eGFR <90 ml/min, or cystatin-C >1.03 mg/dl. Moderate to severe renal impairment was defined as the worst category of renal function (highest tertile, quartile, etc.), which corresponded to creatinine ≥1.5, CrCl or eGFR <53, or cystatin-C ≥1.56. The definitions based on eGFR are similar to categories proposed by the National Kidney Foundation for stage 2 or worse chronic kidney disease (eGFR <90) and stage 3 or worse disease (eGFR <60). National Kidney Foundation staging does not provide equivalent measures for creatinine and cystatin-C, and thus the “equivalent” definitions for our analysis were based on patient distributions (percentiles), as defined in the individual studies. Additionally, for our analysis, we evaluated worsening renal function in hospital and defined this as a change in creatinine ≥0.3 mg/dl.
The primary outcome was all-cause mortality, determined by individual study methods, including review of Medicare databases, death certificates, hospital records, or follow-up with proxy contacts. Secondary outcomes were also evaluated. Cardiovascular mortality (all cardiovascular mortality and HF or pump failure mortality) was determined according to individual study methods, including committee consensus regarding data from death certificates, hospital records, or follow-up with proxy contacts; hospital admissions/readmissions by review of Medicare databases, hospital administrative data, or by committee consensus regarding data from follow-up patient interviews and review of hospital records; hospital costs by review of Medicare databases or hospital administrative data; and functional decline by validated functional status scales such as NYHA functional class or activities of daily living assessment (17).
Fixed effects meta-analysis was conducted to estimate the magnitude of risk associated with renal impairment and all-cause mortality, as measured by combined crude mortality risks and unadjusted risk ratios (RR), and combined adjusted hazard ratios (HR) with 95% confidence intervals (CI). Adjusted risk estimates included those published in final multivariate models for each study, which considered confounding from sociodemographic and clinical covariates such as age, gender, race, comorbidities, medications, physical exam and symptoms, ejection fraction, electrocardiogram findings, laboratory values, and neurohormonal measures.
Inverse variance-weighted averages of logarithmic RRs and HRs were calculated for all-cause mortality and secondary outcomes. Meta-estimates reflected only the subset of studies reporting combinable risk estimates. Among-study heterogeneity of risk estimates was examined using a standard chi-square test for heterogeneity. A funnel plot based on the primary outcome evaluated for potential systematic bias in studies, including publication bias. Subgroup analyses stratifying by race, gender, etiology of HF, severity of HF symptoms (determined by NYHA functional class), and follow-up period (early mortality, defined as ≤1-year mortality or mean follow-up ≤1 year) were also conducted to identify highest risk patients.
All statistical tests assumed a two-tailed alpha level = 0.05. Studies using overlapping patient populations (multiple secondary analyses of randomized trial data) were included in calculations for combined risk estimates only once, with the study reporting outcomes for the greatest number of patients taking precedence. All analyses were conducted using Review Manager version 4.2 (Cochrane Collaboration) (18).
Description of studies
A total of 16 high-quality studies with 80,098 HF patients were identified from a total of 1,309 potentially relevant studies screened, with 3 studies excluded (10–12) from the initial 19 studies of renal impairment in HF that were identified (Fig. 1). In the included studies, HF patients were derived from randomized clinical trials (2–4,19–24), hospital cohorts (5,25–27), and outpatient HF populations (28–30). Patients had a mean age of 74 years and a wide spectrum of HF severity. Follow-up ranged from a minimum of six months to more than five years (Tables 1 and 2).⇓⇓
In 12 studies reporting crude prevalence data (2–5,19,20,22,26,28–30), 29% of all patients had moderate to severe renal impairment, and more than half of all patients had any degree of renal impairment. Prevalence was lower in outpatients compared with hospitalized patients (Table 3).In the 11 studies reporting all-cause mortality rates for follow-up ≥1 year (range 1 to 11.7 years) (2–5,19,21,26,28,30), 26% without impairment, 42% with any impairment, and 51% with moderate to severe impairment died. This translated into a combined unadjusted mortality risk of RR = 1.48, 95% CI 1.45 to 1.52, p < 0.001 in patients with any renal impairment and RR = 1.81, 95% CI 1.76 to 1.86, p < 0.001 in patients with moderate to severe impairment.
In the nine studies that reported mortality risks adjusted for sociodemographic and clinical covariates (2–5,19,21,26,30), any renal impairment was still associated with higher mortality risk (HR = 1.56, 95% CI 1.53 to 1.60, p < 0.001) (Fig. 2).Significant heterogeneity was present (p < 0.001). An alternative estimate using random effects yielded an estimate of similar effect size (HR = 1.54, 95% CI 1.39 to 1.70). In a sensitivity analysis excluding Hillege et al. (2) (the only study limited to NYHA functional class III to IV patients) and Smith et al. (5) (disproportionate number of hospitalized elderly), no heterogeneity was observed (p = 0.24) and mortality risk remained elevated and significant (HR = 1.39, 95% CI 1.29 to 1.50, p < 0.001). The funnel plot suggested no publication bias (Fig. 3).In the five studies that reported adjusted mortality risk specifically for patients with moderate to severe renal impairment, mortality risk was more than doubled (HR = 2.31, 95% CI 2.18 to 2.44, p < 0.001) (Fig. 4).
When renal function was characterized continuously and linearly, renal impairment generally showed an incremental increase in mortality risk. A 33% increased mortality risk per 1 mg/dl creatinine (31% to 36%, p < 0.001) (or 15% per every 0.5 mg/dl creatinine [14% to 17%]) found in three studies (5,26,29) and a 7% increased mortality risk per 10 ml/min eGFR (4% to 10%, p < 0.001) found in two studies (20,24) suggested a dose-response relationship (Fig. 5).However, one study observed a threshold effect with eGFR, whereby increased risk was observed only with eGFR <50 (25). Additionally, three studies of hospitalized patients (5,22,25) showed significantly increased mortality risk associated with renal impairment, characterized as worsening renal function during hospital admission (HR = 1.47, 95% CI 1.26 to 1.72) (Table 4).
For cardiovascular-specific mortality, risk of pump-failure death was increased in patients with renal impairment, but risk of arrhythmic death was not (22,25,27). For the outcome of hospitalization, worsening renal function was marginally associated with increased hospital readmissions in hospital-derived patients (5,25), whereas baseline renal impairment in non-hospital-derived patients was significantly associated with increased risk of the combined outcome of death or hospitalization for HF (4,19) (Table 3). Only a single study (5) assessed and reported increased hospital admission costs associated with worsening renal function, and only one study (3) reported increased six-month functional decline with severe renal impairment (odds ratio = 1.95, 95% CI 1.16 to 3.28).
One prior study (23) found significantly higher mortality risk in whites versus blacks: severe renal impairment in whites was associated with HR = 2.61 (95% CI 2.44 to 2.80), but in blacks was HR = 1.99 (95% CI 1.62 to 2.45) (p < 0.001 for interaction). Though no prior study stratified the effects of renal impairment on mortality by gender, a single study reported results in a cohort limited to women (19) (HR = 2.40, 95% CI 1.60 to 3.62 for severe impairment). Finally, in an analysis stratified by HF etiology (23), significantly increased mortality risk was reported both for patients with non-ischemic HF and ischemic HF.
Level of baseline symptoms appeared to modify the effect of renal impairment. Asymptomatic patients (NYHA functional class I to II) (31,32) still had significantly increased mortality risk with moderate to severe renal impairment (HR = 1.41, 95% CI 1.15 to 1.73), whereas more symptomatic patients (NYHA functional class III to IV) (1) showed substantially higher mortality risk for any renal impairment (HR = 2.10, 95% CI 1.76 to 2.50) and severe renal impairment (HR = 3.23, 95% CI 2.42 to 4.31). In patients with follow-up for early mortality (<1 year), any renal impairment was associated with a higher adjusted mortality risk than the risk found for ≥1 year (HR = 1.84, 95% CI 1.62 to 2.09, p < 0.001), although the studies contributing to this risk estimate were relatively heterogeneous (Fig. 5).
The majority of HF patients in these studies had some degree of renal impairment, and these patients represent a high-risk group with an approximately 50% increased relative mortality risk compared with patients of normal renal function. Moreover, up to 29% of patients had moderate to severe renal impairment, with more than 100% increased relative mortality risk and absolute mortality rate as high as 51% by five years of follow-up. White patients and symptomatic patients appear to have particularly poor outcomes in the setting of renal impairment. Regardless, the impact of renal impairment in all HF patients is substantial both in prevalence and excess mortality risk, underscoring the importance of incorporating renal function into risk-stratification models and identifying optimal HF therapies for this population.
Though risk-stratification models in HF have increasingly included renal function for mortality risk prediction (33), evidence for therapeutic strategies to actually reduce mortality in HF patients with renal impairment is still lacking. Typically, HF patients with creatinine ≥2.5 have been systematically excluded from therapeutic trials, and thus optimal pharmacotherapy remains poorly defined, especially in patients with severe renal impairment (34). Evidence for therapeutic benefit from angiotensin-converting enzyme (ACE) inhibitors in HF patients with renal impairment is incomplete. One randomized trial of enalapril in patients with severe HF found a significant mortality benefit even with the inclusion of patients with moderate to severe renal impairment (6). An observational study by Masoudi et al. (1,35) recently showed similar or even increased relative survival benefit from ACE inhibitor treatment in HF patients with severe renal impairment compared with treatment in patients with moderate impairment or normal renal function. In contrast, Ezekowitz et al. (4) found no mortality benefit from ACE inhibitors in patients with CrCl <30 and ischemic HF, potentially because of interactions between combined aspirin and ACE inhibitor use. Interestingly, Ezekowitz et al. (4) noted a certain degree of “therapeutic nihilism” toward HF patients with renal impairment, as these patients had lower prescription rates of ACE inhibitors and beta-blockers. Certainly, caution is warranted, given evidence for risk of worsening renal function from spironolactone and nesiritide, as well as ACE inhibitors (5,26,29). The therapeutic dilemma concerning treatment of HF patients with renal impairment supports the need for randomized trials of therapeutic agents that specifically target HF patients with moderate to severe renal function. Indeed, though many of the studies in our meta-analysis adjusted for use of medical therapies, it is still possible that the excess mortality associated with renal impairment reflected, in part, a lack of efficacious treatment options available or possible underutilization of current therapies.
Results from our study also highlight the need to establish the best estimator of renal function and standardized definitions of renal impairment in HF patients. The studies reviewed in this analysis utilized a variety of directly measured serum markers as well as calculated estimates of GFR, but no study has empirically compared these estimators. Yet despite the varied definitions of renal impairment in the studies included in our meta-analysis, the association with mortality remained robust. However, future studies are still required to better define incremental mortality risk, and whether a linear and continuous characterization of the association between renal impairment and mortality is appropriate. Our study found conflicting results, with at least one study finding a threshold effect for eGFR >50, suggesting that excess risk was virtually null above this threshold (30). However, alternative studies suggested incremental risk even at the lowest levels of renal impairment (36). Notably, studies reporting incremental risks assumed a prioria linear relationship between eGFR or creatinine and mortality risk. Although both the categorical and continuous characterizations of renal function appeared to affirm the presence of a dose-response relationship, the validity of the linearity assumption could not be confirmed by our analysis and should be addressed in future studies. Finally, newer measures of renal function, including cystatin-C, may show promise for risk stratification and mortality risk prediction of HF patients although, to date, only one study had been published using cystatin-C to assess renal function in HF patients (37).
A causal pathway?
Our findings support the hypothesis that renal impairment could be a marker for worsening HF. However, some investigators have further postulated that renal impairment could be on the causal pathway to worsening HF. Our findings of the strength, consistency, and dose-response effect may help to support, although certainly not prove, causality. In particular, the temporal relationship between renal impairment and worsening HF could not be established in our study. However, other investigators have found that baseline renal impairment in healthy community subjects appears to increase the risk of developing HF, providing initial evidence to support a temporal relationship. Proposed biological mechanisms also support a potential causal pathway, as investigators have postulated that renal impairment could upregulate the renin-angiotensin-aldosterone system, enhance basal sympathetic nerve discharge, increase pro-inflammatory factors, and exacerbate underlying anemia, worsening left ventricular hypertrophy and myocyte contractility, leading to impaired volume handling, pump failure, and death. If renal impairment in HF is truly causal for mortality, cardio-renal pathophysiologic pathways could be an important target for preventing worsening HF, especially in those patients with the most severe levels of renal impairment.
For mortality risk estimates, our analysis could not rule out residual confounding from individual studies in adjusted analyses, although most studies did adjust for all known major confounders and yielded consistent estimates. However, this limitation emphasizes the inability of our analysis to establish a causal association. Second, sampling strategies varied for the different studies included in our analysis, so our results are not a nationally representative sample for prevalence. However, estimates for the prevalence of any renal impairment, despite the heterogeneity of definitions and populations, remained surprisingly consistent in both hospitalized and non-hospitalized patients. Thus, a conservative estimate of renal impairment in at least half of HF patients is likely valid (particularly given that clinical trial populations usually exclude patients with severe impairment).
Renal impairment confers a clinically significant risk for excess mortality in patients with HF, and the magnitude of this increased mortality risk is comparable to that associated with traditional prognostic indicators in HF, such as ejection fraction. Though more common in patients hospitalized for HF, at least some degree of renal impairment is still present in about half of stable HF outpatients. As renal function is a key independent prognostic factor in HF and relevant for developing optimal therapeutic strategies, a more refined understanding of the best method of estimating and characterizing renal impairment, particularly in the mild to moderately impaired range, is still required.
↵1 Dr. Smith was supported by NIH/National Institute of General Medical Sciences Medical Scientist Training Grant GM07205.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- confidence interval
- creatinine clearance
- estimated glomerular filtration rate
- heart failure
- hazard ratio
- New York Heart Association
- odds ratio
- risk ratio
- Received July 26, 2005.
- Revision received November 18, 2005.
- Accepted November 21, 2005.
- American College of Cardiology Foundation
- Hillege H.L.,
- Girbes A.R.,
- de Kam P.J.,
- et al.
- Bibbins-Domingo K.,
- Lin F.,
- Vittinghoff E.,
- Barrett-Connor E.,
- Grady D.,
- Shlipak M.G.
- Shlipak M.G.,
- Smith G.L.,
- Rathore S.S.,
- Massie B.M.,
- Krumholz H.M.
- Smith G.L.,
- Shlipak M.G.,
- Havranek E.P.,
- et al.
- Ezekowitz J.,
- McAlister F.A.,
- Humphries K.H.,
- et al.
- Sarnak M.J.,
- Levey A.S.,
- Schoolwerth A.C.,
- et al.
- Forman D.E.,
- Butler J.,
- Wang Y.,
- et al.
- Katz S.,
- Downs T.D.,
- Cash H.R.,
- Grotz R.C.
- ↵The Cochrane Collaboration. Available at: www.cochrane.org. Accessed July 20, 2005
- Dries D.L.,
- Exner D.V.,
- Domanski M.J.,
- Greenberg B.,
- Stevenson L.W.
- Al-Ahmad A.,
- Rand W.M.,
- Manjunath G.,
- et al.
- Mahon N.G.,
- Blackstone E.H.,
- Francis G.S.,
- Starling R.C. 3rd.,
- Young J.B.,
- Lauer M.S.
- McClellan W.M.,
- Flanders W.D.,
- Langston R.D.,
- Jurkovitz C.,
- Presley R.
- McAlister F.A.,
- Ezekowitz J.,
- Tonelli M.,
- Armstrong P.W.
- Shlipak M.G.,
- Katz R.,
- Fried L.F.,
- et al.
- Masoudi F.A.,
- Rathore S.S.,
- Wang Y.,
- et al.
- Sackner-Bernstein J.D.,
- Skopicki H.A.,
- Aaronson K.D.