Author + information
- Received December 17, 2001
- Revision received May 17, 2002
- Accepted June 12, 2002
- Published online September 18, 2002.
- Niall G Mahon, MD*,
- Eugene H Blackstone, MD, FACC†,‡,§,
- Gary S Francis, MD, FACC*,
- Randall C Starling III, MD, FACC*,
- James B Young, MD, FACC* and
- Michael S Lauer, MD, FACC*,* ()
- ↵*Reprint requests and correspondence:
Dr. Michael S. Lauer, Department of Cardiovascular Medicine, Cleveland Clinic Foundation, Department of Cardiology, Desk F25, 9500 Euclid Avenue, Cleveland, Ohio 44195, USA.
Objectives The goal of this study was to determine the prognostic significance of estimated creatinine clearance (CrCl) in relation to 6-min walk distance in ambulatory patients with congestive heart failure (HF).
Background Although measurement of renal function is integral to the management of chronic congestive HF, its prognostic implications are not well described and have not been formally evaluated relative to measures of functional capacity.
Methods We analyzed outcomes of the 585 participants of the 6-min walk substudy of the Digitalis Investigation Group (DIG) trial. The CrCl was estimated using the Cockcroft-Gault equation. Predictors of all-cause mortality were identified using semiparametric Cox proportional hazards regression and completely parametric hazard analyses.
Results Most subjects (85%) were New York Heart Association functional class II and III. Mean age was 65 (±12) years and mean ejection fraction (EF) 35% (±13%). There were 153 (26%) deaths during a median of 2.6 years of follow-up. Mortality by increasing quartiles of estimated CrCl was 37% (18 to 48 ml/min), 29% (47 to 64 ml/min), 18% (64 to 86 ml/min), and 21% (86 to 194 ml/min) with corresponding hazard ratios (HRs) relative to the top quartile of 2.1 (95% confidence interval [CI], 1.4 to 3.3), 1.6 (95% CI, 1.0 to 2.5), and 0.9 (95% CI, 0.5 to 1.5), respectively. In Cox regression analyses, independent predictors of mortality were estimated CrCl (adjusted HR [quartile 1:quartile 4] 1.5; 95% CI, 1.1 to 2.1), 6-min walk distance ≤262 m [adjusted HR, 1.63; 95% CI, 1.12 to 2.27]), EF, recent hospitalization for worsening HF, and need for diuretic treatment. Parametric (hazard) analysis confirmed consistent effects of estimated CrCl on mortality in several subgroups including that of patients with EF >45%.
Conclusions In ambulatory patients with congestive HF, estimated CrCl predicts all-cause mortality independently of established prognostic variables.
Measurement of renal function is integral to the management of chronic congestive cardiac failure and influences clinical decision-making (1). Moreover, data from the Studies Of Left Ventricular Dysfunction (SOLVD) registry have suggested that renal function may provide independent prognostic information both in symptomatic and asymptomatic heart failure (HF) (2). However, studies systematically evaluating the prognostic significance of renal function in cardiac failure are surprisingly few, and the degree to which this variable has been evaluated in relation to traditional prognostic markers including exercise tolerance is limited.
The 6-min walk test is an objective measure of exercise capacity that can be obtained in a range of settings and is an independent predictor of mortality in cardiac failure (3–5). This test was employed in a substudy of the Digitalis Investigation Group (DIG) trial, a large-scale multicenter study designed to evaluate the effect of digoxin on morbidity and mortality in patients with HF in sinus rhythm (6). The aim of this study was to determine the prognostic utility of estimated creatinine clearance (CrCl) relative to traditional prognostic markers, including exercise capacity, in patients participating in the exercise substudy of the DIG trial.
Full details of the DIG trial have been published elsewhere (6). Briefly, for the main trial (n = 6,800) eligible patients had HF (based on current or past clinical symptoms or signs and/or on X-ray evidence of pulmonary congestion  and an ejection fraction [EF] ≤45% as determined by echocardiography, radionuclide scanning, or ventriculography). Patients with HF and EF >45% were enrolled in an ancillary trial (n = 998). In 39 of 302 participating centers, 585 patients were enrolled in a 6-min walk test substudy designed to measure the effect of treatment on functional capacity. Inclusion criteria were as for the overall trial. Follow-up for up to five years (median 3.2 years) was available on all 7,788 patients who participated in the DIG trial including the 585 patients in the exercise substudy (median follow-up 2.6 years). For this analysis only all-cause mortality was considered as an end point, as, unlike “cardiovascular mortality,” it is clinically relevant, wholly objective, and unbiased (8,9).
The Cleveland Clinic Foundation Institutional Review Board approved analyses based on the publicly available DIG data set.
Calculation of CrCl
All patients had serum creatinine recorded at baseline. Estimated CrCl was calculated for each patient using the Cockcroft-Gault equation: CrCl (men) = ([140 − age in years]·weight in kg) / (72·creatinine in mg/dl) (10). For women the above was multiplied by 0.85.
6-min walk test
The 6-min walk test was performed using standard methods (11). Distance walked at 3 min and 6 min, total duration of walk if <6 min, whether there were any breaks, and reasons for stopping were recorded. Other baseline clinical data on all participants were recorded as documented previously (6). These included EF as measured either by echocardiography (29%), radionuclide angiography (67%), or ventriculography (4%) within six months before enrollment, provided there had been no intercurrent cardiac event.
Patients were ranked according to CrCl, and comparisons of baseline variables between groups were performed using the Wilcoxon rank-sum test for continuous variables and the chi-square test for categorical variables. All analyses were performed with two-sided p values. Survival according to CrCl, 6-min walk distance, and other potential predictors of mortality was assessed using Kaplan-Meier curves (12) and Cox proportional hazards (13) with transformations of all continuous variables tested. Optimal proportional hazard models were identified using the Akaike Information Criterion (14). The Cox proportional hazards model was also used to test for interactions between creatinine quartiles and other baseline variables. The proportional hazards assumption was confirmed by examination of Schoenfeld residuals. In addition, a model incorporating serum creatinine rather than CrCl was tested, with age and gender entered separately, but this had a chi-square inferior to that incorporating CrCl.
Model validation was carried out by means of bootstrap resampling (15,16). A total of 1,000 resampling analyses were performed for variable selection using p = 0.10 for model entry and p ≤ 0.05 for model retention. Those variables that were retained in at least half the models were considered in 1,000 fixed variable resamplings.
In addition, a completely parametric survival analysis was carried out to resolve the number of hazard phases, identify the form of the equation for each phase, and estimate the parameters that characterized the distribution of times until death (17). Only one hazard phase, which followed an exponential distribution, was found. Using the HAZARD procedure, with stepwise variable entry, a multivariable risk factor model was generated with a value of p < 0.05 criterion for retention of variables in the final model. Graphical displays of survival according to CrCl under varying conditions of concomitant variables are also shown using the proc HAZPRED and proc HAZPLOT functions. All analyses were performed using SAS version 8.1 software (SAS Inc., Cary, North Carolina). The HAZARD, HAZPRED, and HAZPLOT macros, and a discussion of parametric survival analysis can be found at www.clevelandclinic.org/heartcenter/hazard.
Baseline characteristics of patients in the exercise substudy were similar to those of the main trial except for the presence of more individuals with EF >45% in the exercising group than the main group (20% vs. 12%, p < 0.0001) with consequent lower follow-up mortality (26% vs. 34%, p < 0.0001). Nonetheless, mortality in the exercise group was substantial and amounted to 153 deaths over the follow-up period. Mean baseline serum creatinine was similar in both groups. Results hereafter refer only to the group participating in the exercise substudy.
Baseline variables in CrCl subgroups
Table 1lists baseline variables according to strata of CrCl. Patients in the lower two quartiles were significantly older than those in the highest quartile, had lower body mass index (BMI), shorter 6-min walk distances, and a higher proportion of female patients. More patients in the lower two quartiles had a recent hospitalization for HF, were in New York Heart Association functional classes III and IV, or required regular diuretic therapy. There was no difference in EF between the ranks even when patients with EF >45% were excluded from analysis. History of previous myocardial infarction, diabetes mellitus, and use of angiotensin-converting enzyme (ACE) inhibitors did not differ between the groups.
CrCl, functional capacity, and survival
Survival according to quartiles of CrCl is shown in Figure 1A. Survival was significantly reduced in patients with CrCl in the lower two quartiles. Figure 1B shows survival according to quartiles of 6-min walk distance. Survival was significantly impaired in patients in the lowest quartile (walk distance ≤262 m).
To evaluate consistency of effect of CrCl on survival, mortality according to CrCl was analyzed in prespecified subgroups. The effect of estimated CrCl was generally consistent across all subgroups inspected (Table 2). However, there was evidence of a weak interaction between CrCl and history of previous myocardial infarction, with a high mortality in patients with no history of myocardial infarction irrespective of CrCl, and between CrCl and walk distance, with low CrCl having less of an adverse effect in individuals with very poor functional capacity.
Multivariable survival analyses
Preliminary Cox proportional hazards analysis was performed with consideration of the following variables: lowest quartile of 6-min walk distance, age, gender, BMI, EF, resting heart rate, previous myocardial infarction, diabetes mellitus, systolic blood pressure, log CrCl, use of diuretics, cardiothoracic ratio, recent hospitalization for worsening heart failure, presence of a third heart sound, and the presence of auscultatory rales. Log CrCl was found to perform better than untransformed CrCl in proportional hazards modeling. No other variable transformation improved the models. Only log CrCl, 6-min walk distance, use of diuretics, recent hospitalization for HF, and EF emerged as independent predictors of mortality.
To create a final model, these 14 variables were entered into 1,000 models based on 1,000 bootstrap resamplings (Table 3). Only five variables entered models more than 50% of the time: 6-min walk distance, EF, diuretic use, hospitalization for worsening HF, and log CrCl. Log CrCl and EF entered the highest number of models. Repeating the resampling with categorization of continuous variables such as age did not affect the results. Similarly, performing the analysis only for patients with EF <45% did not affect the results. These five variables were subsequently considered in 1,000 fixed variable resamplings for calculation of hazard ratios (Table 3).
As expected, parametric analyses yielded the same independent predictors of death as did proportional hazards modeling. There was no interaction noted between CrCl and 6-min walk distance. Plots describing four-year survival according to CrCl for different degrees of exercise limitation and left ventricular dysfunction are shown in Figures 2 and 3. ⇓⇓ In all circumstances there was a continuous decline in survival with falling CrCl, with sharper declines observed as CrCl fell below 40 to 50 ml/min. To determine whether CrCl had similar prognostic implications in the subgroup with HF and EF >45%, a second hazard analysis was performed with EF dichotomized at the 45% level. There were comparable effects of estimated CrCl on survival in both groups. No interaction was found.
Renal function in congestive HF
Few studies have systematically explored the prognostic utility of renal function in cardiac failure. In the Prospective Randomized Milrinone Survival Evaluation (PROMISE) study involving 1,088 patients, a serum creatinine level >1.3 mg/dl was associated with a trend towards increased all-cause mortality (18). Subsequent smaller studies have also suggested that serum creatinine predicts all-cause or cardiovascular mortality (19–21). However, analysis of serum creatinine (rather than CrCl) is potentially problematic because of the confounding effects of age, gender, and BMI. Accordingly, in an analysis from the SOLVD prevention and treatment studies (n = 6,887), CrCl (Cockcroft-Gault) <60 ml/min was shown independently to predict pump-failure deaths and the combined end point of death or hospitalization (2). In the present study the relationship between CrCl and prognosis was confirmed over a wider range of creatinine levels because in the SOLVD trial a creatinine level >2 mg/dl was an exclusion criterion, and among patients whose therapy more closely approximated to current HF management because the majority (93%) was on ACE inhibitors. Furthermore, this study demonstrated a continuous relationship between CrCl and survival. As is clear from the hazard plots, the probability of survival was reduced by even minor reductions in CrCl, levels at which the serum creatinine might still be normal.
Renal function and exercise capacity in HF
In addition, the present study incorporated an objective measure of exercise capacity. While this study confirmed that 6-min walk distance independently predicts mortality in HF patients, of greater interest is the finding that CrCl predicted survival independently of measured functional capacity and that low CrCl was associated with a reduced probability of survival irrespective of effort tolerance. Indeed, CrCl may be a particularly useful marker in individuals with less severely impaired functional capacity, in whom the prognosis may be unclear. Similarly, we demonstrated a consistent effect of CrCl on survival in patients with different degrees of ventricular dysfunction, including patients with HF and normal systolic function, with potential implications for its use as a prognostic marker in individuals with less severe degrees of left ventricular dysfunction.
Impact of renal function on natural history of HF potential mechanisms
The mechanisms by which impaired renal function exerts an adverse prognostic effect remain to be established. As proposed previously (2), renal dysfunction may have a causative role in the progression of cardiac failure, through subtle alterations in the ability to control volume status and/or compromised clearance of cardiac toxins. An active role for renal dysfunction in this regard is supported by published reports of resolution of secondary ventricular dysfunction after renal transplantation in patients with primary renal insufficiency (22–26).
The Cockcroft-Gault formula is based on ideal rather than actual body weight. Because many patients in the DIG trial had clinical congestion, baseline recorded weights may have exceeded ideal body weights. Furthermore, the formula performs less well in subgroups including diabetics and at the extremes of age and body size (27).
With respect to the generalizability of these findings, the DIG trial had no upper age limit, and its participants may more closely resemble the general population than participants in more complex studies. Nonetheless, the mean age of 64 years is lower than in practice (28), and caution must be exercised in extrapolating these findings to older patients with HF. Furthermore, this trial did not include patients with atrial fibrillation, who account for approximately 20% of patients with HF in major HF studies (29,30).
Due to the limited dataset, the prognostic utility of peak oxygen consumption (VO2) (or, indeed, of potential newer markers such as natriuretic peptides) could not be evaluated. The 6-min walk test is less established as a prognostic tool than peak VO2, may be subject to greater degrees of variability, likely measures submaximal rather than maximal exercise capacity, and correlates only modestly with peak VO2(5,31–34). In one study 6-min walk distance was not prognostic if peak VO2 was included in the model (34). However, for the majority of patients with HF, the walk test may be a more practical, albeit cruder, prognostic functional test.
Assessment of prognosis is central to the delivery of patient care and is a basis on which important therapeutic decisions are made, including referral for cardiac transplantation. The use, for prognostic purposes, of a variable that is universally recorded in patients with HF is inherently attractive. Recently, Aaronson et al. (35) have developed predictive models incorporating several commonly measured prognostic variables, not including CrCl (a dichotomous variable “renal disease [moderate or severe]” was recorded but did not enter the final models) (35). In view of the importance of renal function in the assessment and management of HF, and in light of emerging evidence of its prognostic importance, CrCl may be an appropriate variable for inclusion in future prognostic models.
In ambulatory patients with chronic congestive cardiac failure, CrCl predicts all-cause mortality independently of established clinical, structural, and functional prognostic variables.
☆ Dr. Lauer receives support from the American Heart Association (Grant 0040244N) and the National Heart, Lung, and Blood Institute (Grant HL 66004-01). Drs. Young and Starling receive support from the Kauffman Heart Failure Center, Cleveland Clinic Foundation.
- angiotensin-converting enzyme
- body mass index
- confidence interval
- creatinine clearance
- Digitalis Investigation Group
- ejection fraction
- heart failure
- hazard ratio
- Studies Of Left Ventricular Dysfunction
- oxygen consumption
- Received December 17, 2001.
- Revision received May 17, 2002.
- Accepted June 12, 2002.
- American College of Cardiology Foundation
- American College of Cardiology/American Heart Association Task Force on Practice Guidelines
- Dries D.L.,
- Exner D.V.,
- Domanski M.J.,
- Greenberg B.,
- Stevenson L.W.
- Zugck C.,
- Kruger C.,
- Durr S.,
- et al.
- Lauer M.S.,
- Blackstone E.H.,
- Young J.B.,
- Topol E.J.
- Cockcroft D.W.,
- Gault M.H.
- Guyatt G.H.,
- Sullivan M.J.,
- Thompson P.J.,
- et al.
- Cox D.R.
- Blackstone E.H.,
- Naftel D.C.,
- Turner M.E.
- Madsen B.K.,
- Hansen J.F.,
- Stokholm K.H.,
- Brons J.,
- Husum D.,
- Mortensen L.S.
- Cowie M.R.,
- Wood D.A.,
- Coats A.J.,
- et al.
- Riley M.,
- McParland J.,
- Stanford C.F.,
- Nicholls D.P.
- Lipkin D.P.,
- Bayliss J.,
- Poole-Wilson P.
- Aaronson K.D.,
- Schwartz J.S.,
- Chen T.M.,
- Wong K.L.,
- Goin J.E.,
- Mancini D.M.