Prediction of mortality using quantification of renal function in acute heart failure

International Journal of Cardiology 201 (2015) 650–657

Contents lists available at ScienceDirect

International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard

Prediction of mortality using quantification of renal function in acute heart failure☆ Zoraida Moreno Weidmann a,b,1, Tobias Breidthardt a,c,1, Raphael Twerenbold a,b, Christina Züsli a,b, Albina Nowak d, Arnold von Eckardstein e, Paul Erne f, Katharina Rentsch g, Mucio T. de Oliveira Jr. h, Danielle Gualandro h, Micha T. Maeder i, Maria Rubini Gimenez a,b,1, Kateryna Pershyna a,b, Fabio Stallone a,b, Laurent Haas a,b, Cedric Jaeger a,b, Karin Wildi a,b, Christian Puelacher a,b, Ursina Honegger a,b, Max Wagener a,b, Severin Wittmer a,b, Carmela Schumacher a,b, Lian Krivoshei a,b, Petra Hillinger a,b, Stefan Osswald a,b, Christian Mueller a,b,⁎ a

Cardiovascular Research Institute Basel (CRIB), University Hospital, Basel, Switzerland Department of Cardiology, University Hospital, Basel, Switzerland Department of Internal Medicine, University Hospital, Basel, Switzerland d Department of Internal Medicine, University Hospital, Zurich, Switzerland e Department of Laboratory Medicine, University Hospital, Zurich, Switzerland f Department of Cardiology, Spital St. Anna, Luzern, Switzerland g Department of Laboratory Medicine, University Hospital, Basel, Switzerland h Cardiology and Emergency Department, Heart Institute (InCor), University of São Paulo Medical School, Brazil i Department of Cardiology, Kantonsspital St. Gallen, Switzerland b c

a r t i c l e

i n f o

Article history: Received 8 February 2015 Received in revised form 3 July 2015 Accepted 9 August 2015 Available online 11 August 2015 Keywords: Acute heart failure Renal function Mortality

a b s t r a c t Background: Renal function, as quantified by the estimated glomerular filtration rate (eGFR), is a predictor of death in acute heart failure (AHF). It is unknown whether one of the clinically-available serum creatininebased formulas to calculate eGFR is superior to the others for predicting mortality. Methods and results: We quantified renal function using five different formulas (Cockroft–Gault, MDRD-4, MDRD6, CKD-EPI in patients b 70 years, and BIS-1 in patients ≥ 70 years) in 1104 unselected AHF patients presenting to the emergency department and enrolled in a multicenter study. Two independent cardiologists adjudicated the diagnosis of AHF. The primary endpoint was the accuracy of the five eGFR equations to predict death as quantified by the time-dependent area under the receiver-operating characteristics curve (AUC). The secondary endpoint was the accuracy to predict all-cause readmissions and readmissions due to AHF. In a median follow-up of 374days (IQR: 221 to 687 days), 445 patients (40.3%) died. eGFR as calculated by all equations was an independent predictor of mortality. The Cockcroft–Gault formula showed the highest prognostic accuracy (AUC 0.70 versus 0.65 for MDRD-4, 0.55 for MDRD-6, and 0.67 for the combined formula CKD-EPI/BIS-1, p b 0.05). These findings were confirmed in patients with varying degrees of renal function and in three vulnerable subgroups: women, patients with severe left ventricular dysfunction, and the elderly. The prognostic accuracy for readmission was poor for all equations, with an AUC around 0.5. Conclusions: Calculating eGFR using the Cockcroft–Gault formula assesses the risk of mortality in patients with AHF more accurately than other commonly used formulas. © 2015 Elsevier Ireland Ltd. All rights reserved.

☆ Acknowledgment of grant support: Professor Mueller has received research grants from the Swiss National Science Foundation and the Swiss Heart Foundation, the Cardiovascular Research Foundation Basel, 8sense, Abbott, ALERE, Brahms, Critical Diagnostics, Nanosphere, Roche, Siemens, and the University Hospital Basel, as well as travel support or speaker/consulting honoraria from Abbott, ALERE, Astra Zeneca, BG medicine, Biomerieux, Brahms, Cardiorentis, Daiichi Sankyo, Lilly, Novartis, Pfizer, Roche, and Siemens. ⁎ Corresponding author at: Department of Cardiology, University Hospital Basel, Petersgraben 4, CH-4031 Basel, Switzerland. E-mail address: [email protected] (C. Mueller). 1 Both authors contributed equally to this work as the first author.

http://dx.doi.org/10.1016/j.ijcard.2015.08.097 0167-5273/© 2015 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Acute heart failure (AHF) is a current worldwide pandemic with unacceptably high morbidity and mortality [1–3]. The intense interference and crosstalk between cardiac and renal function in AHF, including its potential modification by novel therapies, have attained increasing recognition [1–6]. Therefore, current clinical practice guidelines universally recommend the use of the estimated glomerular filtration rate (eGFR) to quantify renal function in AHF patients.

Z.M. Weidmann et al. / International Journal of Cardiology 201 (2015) 650–657

Four serum creatinine-based formulas, initially derived and validated in patients with chronic kidney disease (CKD), are currently applied in patients with AHF: the Cockroft–Gault formula [7], the Modification of Diet in Renal Disease (MDRD) 6 formula [8], the simplified MDRD-4 formula [9], and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula [10]. These formulas incorporate factors such as age, gender, race, and weight and allow a reasonably accurate estimation of eGFR [11]. A less commonly used creatinine-based equation, the Berlin Initiative Study 1 (BIS-1) formula, seems to be more accurate than the CKD-EPI to estimate GFR in persons aged 70 years or older [12]. While it has been consistently shown that impaired renal function is common in AHF and associated with a higher rate of death, it is largely unknown whether one of the clinically-available serum creatininebased formulas to calculate eGFR is superior to the others to predict death. Recent pilot studies (two in AHF and three in chronic heart failure (CHF)) have begun to shed light on the relevance of the method used to calculate eGFR [13–17]. Based on the observations made in CHF [15], we hypothesized that eGFR, as calculated by the Cockroft–Gault formula, predicts postdischarge outcomes in AHF more precisely than MDRD-4, MDRD-6, CKD-EPI or BIS-1. We aimed to test this hypothesis in a large, international multicenter study.

2. Methods 2.1. Setting and study population We prospectively analyzed unselected patients presenting with AHF at the emergency department (ED) of one of four participating centers in two countries, Switzerland (Basel, Luzern and S. Gallen), and Brazil (Sao Paolo). The diagnosis of AHF was adjudicated by two independent cardiologists based on available medical records and according to the current guidelines of the European Society of Cardiology [1]. The study was carried out according to the principles of the Declaration of Helsinki and approved by the local ethics committees. All patients provided written informed consent to participate. For inclusion in these analyses, patients were required to have serum creatinine quantification and information on body weight at presentation to the ED. We excluded patients with endstage kidney disease requiring dialysis.

2.2. Clinical assessment We prospectively recorded medical history, physical exam, electrocardiography, blood tests, radiographies, echocardiograms and other studies performed during hospital admission. We collected blood samples for serum creatinine at presentation to the ED. All samples were analyzed in the respective central laboratory. Serum creatinine was measured using the enzymatic method (COBAS INTEGRA®, Roche Diagnostics GmbH, at 37 °C, calibrated to IDMS standard).

651

2.4. Follow-up and endpoints The primary endpoint was all-cause mortality. Secondary end-points were all-cause rehospitalization and rehospitalization due to AHF. Clinical follow-up data were obtained by telephone interview with the patient or with the referring physician at 3, 6, 12 and 24 months after presentation at the ED, and from hospital-based reports and administrative databases. 2.5. Statistical analysis Continuous variables are reported as mean (standard deviation [SD]) or median (interquartile range [IQR]), according to their distribution; categorical variables are expressed as numbers and percentages. Continuous variables were compared using the independent Student's t-test or the Wilcoxon–Mann–Whitney U-test and categorical variables using Pearson's chi-square test. The Wilcoxon signed-rank test was used for pairwise comparisons. Spearman's rank-correlation coefficients and a Bland–Altman method with 95% limits of agreement (LoA) were used to describe the pairwise agreement between eGFR calculated with the various formulas. Cohen's Kappa was used to compare agreement in classification into the four defined stages of the NKF K/DIGO or as a dichotomized variable (impaired versus preserved renal function). We compared Spearman's rank correlation coefficient and Cohen's kappa values between pairing methods using a bootstrap method with 500 replicates. The predictive performance of the eGFR equations was calculated constructing a time-dependent receiver operating characteristic (ROC) curve, as proposed by Haegerty [20], comparing the areas under the curve (AUC) with a Wilcoxon test. Prognostic accuracy between predefined subgroups was evaluated constructing a ROC curve and calculating the AUC in a standard fashion, and comparisons were made using a DeLong test. Survival was calculated using Kaplan–Meier analysis and differences between the curves were evaluated using Log-Rank statistics. Cox proportional hazards regression modeling was performed to assess time-to-event associations with mortality. Variables with b10% missing values and a p-value b 0.1 at univariate analysis were entered into the multivariate model using a forward stepwise selection. Covariates with p N 0.1 in a univariate analysis but with theoretically clinical relevance were also included in the model. The improvement in predictive accuracy for the Cockroft–Gault equation over the other equations was tested using the net reclassification improvement (NRI) as continuous NRI, as proposed by Pencina et al. [21]. Significance was defined as two-tailed p value b 0.05. Analyses were performed with IBM SPSS Statistics version 21.0 (SPSS, Chicago, Illinois), R software version 3.0.2 (R Foundation for Statistical Computing, Vienna, Austria) and MedCalc version 14.8.1 (MedCalc Software, Ostend, Belgium).

3. Results 3.1. Baseline characteristics of the cohort A cohort of 1104 patients with AHF was included in the analysis (Table 1): 45% of patients were female, median age was 79 years and the prevalence of comorbidities was high. The most common cause of chronic heart failure was ischemic heart disease and the median left ventricular ejection fraction was 45% (IQR: 30–58%). A history of CKD was present in 43.3% of patients. 3.2. Renal function

2.3. Quantification and interpretation of renal function We calculated eGFR using: 1) the Cockroft–Gault equation [7], ((140 − Age in years) ∗ weight at admission in kg / (72 ∗ serum creatinine (sCr) in mg/dl) ∗ 0.85 (if female)). In a secondary analysis, we adjusted the same equation for body surface area (BSA) according to the Mosteller formula (Height in cm ∗ Weight in kg / 3600)1/2; 2) MDRD-4 equation [18] (186.3 ∗ sCr−1.154 ∗ Age−0.203 ∗ 0.742 (if female)); 3) MDRD-6 equation [8], (170 ∗ sCr−0.999 ∗ Age−0.176 ∗ 0.762 (if female) ∗ serum urea nitrogen in mg/dl−0.170 ∗ Albumin in g/dl0.318; 4) CKD-EPI creatinine equation [10], 141 min(sCr/κ, 1)∝ max(sCr/κ, 1)−1.209 ∗ 0.993Age ∗ 1.018 (if female), where κ is 0.7 for female and 0.9 for male, α is −0.329 for female and −0.411 for male, min is the minimum of serum creatinine/κ or 1, and max is the maximum of serum creatinine/κ or 1; and 5) BIS-1 equation [12], 3637 ∗ sCr in mg/dl−0.87 ∗ Age in years−0.95 ∗ 0.82 (if female). Given that the BIS-1 formula seems to be superior to the CKD-EPI to estimate GFR in elderly patients, for the purpose of the study, we combined the CKD-EPI and BIS-1 formulas in a single equation, hereafter referred to as CKD-EPI/BIS-1, using CKD-EPI in patients b 70 years and BIS-1 in patients ≥ 70 years. Renal dysfunction was defined as a GFR b 60 ml/min/1.73 m2. eGFR was categorized into five groups based on the National Kidney Foundation (NKF) Kidney Disease Outcomes Quality Initiative (KDOQI) stages [19]: ≥90 ml/min/1.73 m2; 89 to 60 ml/min/1.73 m2; 59 to 30 ml/min/1.73 m2; 29 to 15 ml/min/1.73 m2 and b15 ml/min/1.73 m2. Categories 4 and 5 were combined given the small number of patients. A clinical history of CKD was determined by interview or examination of clinical reports.

Fig. 1 shows the estimates of GFR using the four formulas, and Table 2 shows their categorization in four groups across the NKF/DOQI classification. The quantitative distribution (p b 0.01 using a Wilcoxon signed-rank test) and the prevalence of renal insufficiency differed significantly according to the four formulas: 69.4% using the Cockroft– Gault, 56.7% with the MDRD-4, 40.8% with the MDRD-6, and 75.7% with the CKD-EPI/BIS-1 (p b 0.05 using Pearson's χ2). 3.3. Agreement of measurements The four formulas correlated significantly when evaluated using the Spearman correlation test, with r-values between 0.92 and 0.99 (Table 3, online supplement). However, when assessed as a categorical variable using Cohen's Kappa, the correlation was overall poorer (between 0.18 and 0.68 when values were divided into four categories according to the KDIGO classification, and 0.36 and 0.77 when values were divided into two categories according to the presence or absence of renal dysfunction). The correlation analysis using a scatterplot and Bland–Altman method (Fig. 2) showed a good agreement between the

652

Z.M. Weidmann et al. / International Journal of Cardiology 201 (2015) 650–657

Table 1 Baseline characteristics of the study population. All-cause mortality Overall cohort (n = 1104)

No (n = 659)

Yes (n = 445)

p-Value

Male (%) Age (years)

55.3 79 (70–85)

56.1 77 (67–83)

53.9 82 (76–87)

0.47 b0.01

Patients' history Hypertension (%) Diabetes mellitus (%) Hypercholesterolemia (%) Smokersa (%) COPD (%) CAD (%) Atrial fibrillation (%) CKD (%) Peripheral arteriopathy (%) Stroke (%)

76.3 30.0 43.1 60.6 25.0 53.4 42.6 43.3 17.3 17.4

77.3 29.9 42.8 61.3 23.7 49.4 43.1 36.0 16.5 14.6

75.6 30.0 43.1 59.4 27.0 59.3 41.9 54.2 18.6 21.6

0.56 0.86 0.98 0.53 0.16 b0.01 0.88 b0.01 0.53 b0.01

Physical exam at ED Weight (kg) Body mass index (kg/mb) Systolic blood pressure (mm Hg) Heart rate (bpm) Orthopnea (%) Paroxysmal nocturnal dyspnea (%) Rales (%) Leg edema (%)

74.0 (63.0–85.0) 26.0 (23.1–29–7) 136 (120–157) 89 (72–108) 53.4 37.6 61.7 59.1

77.0 (66.1–88.0) 27.0 (23.9–30.4) 139 (123–158) 89 (72–109) 56.0 50.8 61.1 57.1

68.0 (58.0–79.4) 24.2 (21.6–28.2) 131 (112–152) 88 (73–105) 55.0 42.6 64.4 64.1

b0.01 b0.01 b0.01 0.84 0.75 0.08 0.08 0.03

Echocardiography HF-REFb (%) LVEF (%) LVEDD (mm)

48.0 45 (30–58) 52 (46–58)

47.4 45 (30–58) 52 (46–58)

49.0 40 (28–58) 52 (47–59)

0.66 0.14 0.75

Laboratory at admission BNP (ng/l) Creatinine (μmol/l) Urea (μmol/l) Albumin (g/l) Hemoglobin (g/l) Length of hospitalization (days)

938.3 (524.1–1700.6) 105.0 (81.0–146.0) 9.6 (6.6–14.1) 35.0 (32.0–38.0) 128.0 (114.0–142.0) 9 (2–15)

812.9 (440.5–1413.2) 96.5 (78.0–129.8) 8.3 (6.1–12.0) 35.0 (32.0–38.0) 131.0 (117.0–145.0) 9 (2–15)

1290.3 (698.8–2320.7) 125.5 (92.0–169.5) 12.0 (8.7–17.1) 34.0 (31.0–36.0) 122.5 (109.0–137.0) 10 (2–16)

b0.01 b0.01 b0.01 b0.01 b0.01 0.42

Treatment at discharge Beta-blockers (%) ACEI/AT-2 inhibitors (%) Aldosterone antagonists (%) Diuretics (%)

62.8 64.7 15.0 71.1

63.7 69.5 14.5 71.6

50.5 57.3 15.7 65.2

b0.01 b0.01 0.10 0.08

Values are percentages or median (interquartile range). Reduced ejection fraction was defined as b45%. COPD: chronic obstructive pulmonary disease; CAD: coronary artery disease; CKD: chronic kidney disease; LVEF: left ventricular ejection fraction; LVEDD: left ventricular end-diastolic diameter; BNP: B-type natriuretic peptide; ACEI: angiotensin converting enzyme inhibitors; AT-2 inhibitors: angiotensin-2 receptor inhibitors. a Smokers include active and former smokers. b HF-REF (HF with reduced ejection fraction).

four formulas in patients with low eGFR values. The differences between formulas increased in higher eGFR values. 3.4. Mortality During the follow-up (median 374 days (IQR 221–687 days)), 445 patients (40.3%) died. These patients were older, more likely to have coronary artery disease and history of CKD, lower systolic blood pressure, higher BNP levels and worse renal function at presentation. A lower proportion of these patients were treated with β-blockers and angiotensin converting enzyme inhibitors (ACEI) or angiotensin-2 (AT-2) receptor inhibitors at discharge. All formulas effectively predicted death. However, the Cockroft– Gault equation was significantly superior, as evaluated by a timedependent ROC curve (Fig. 3). The quantified area under the ROC (AUC) with a two-year cut-off was 0.7 for the Cockroft–Gault, 0.67 for CKD-EPI/BIS-1, 0.65 for MDRD-4, and 0.55 for MDRD-6 formulas (p b 0.01 for all comparisons of the formulas against Cockroft– Gault using a Wilcoxon test). Adjustment of the Cockroft–Gault formula by BSA did not increase the prediction ability provided by the

unadjusted formula (AUC 0.68, 95% CI 0.64–0.71 for two-year mortality prediction). Fig. 4 shows long-term Kaplan–Meier survival. All four formulas provided prognostic information for long-term mortality (Log-Rank p value b 0.01). The four eGFR groups diverged in a more pronounced way with the Cockroft–Gault formula, reflected in higher chi-square values: 108.3 for Cockroft–Gault formula, 104.4 for MDRD-6, 63.8 for MDRD-4 and 63.4 for CKD-EPI/BIS-1. A multivariate Cox regression analysis for long-term mortality (Table 4) showed that all five formulas remained accurate for predicting long-term mortality after controlling for age, body mass index (BMI), systolic blood pressure and BNP at presentation. When eGFR was entered in a multivariate model as categoric variables, according to the NKF K/DOQI classification, there was an independently graded increase in the risk of mortality at higher stages. Reclassifying patients according to the probability of death at two years showed a net reclassification improvement (NRI) of 58% when comparing the Cockroft–Gault formula with MDRD-4, 56% when comparing it with MDRD-6, and 57% when comparing it with CKDEPI.

Z.M. Weidmann et al. / International Journal of Cardiology 201 (2015) 650–657

653

remained superior for the Cockroft–Gault formula in both subgroups (Table 8). 3.5. Rehospitalization During follow-up, 534 patients (48.4%) were readmitted (195, 17.6% due to AHF). The prognostic accuracy for all-cause rehospitalization (AUC 0.54–0.56) and for readmission due to AHF (AUC 0.53–0.55) was low for the four formulas. 4. Discussion

Fig. 1. Quantitative distribution of eGFR according to the 4 different estimating formulas. y-Axis values are eGFR in ml/min/1.73 m2 for MDRD-4, MDRD-6 and CKD-EPI/BIS-1 and ml/min for Cockroft–Gault.

Table 2 Categoricala distribution of estimated GFR in study population according to the 4 formulas. Formula

eGFR ≥ 90 ml/min/ 1.73 m2

eGFR 89–60 ml/min/ 1.73 m2

eGFR 59–30 ml/min/ 1.73 m2

eGFR b 30 ml/min/ 1.73 m2

Cockroft–Gault MDRD-4 MDRD-6 CKDEPI/BIS-1

91 (8.2) 150 (13.6) 320 (29.0) 79 (7.2)

247 (22.4) 328 (29.7) 333 (30.2) 189 (17.1)

523 (47.4) 466 (42.2) 346 (31.3) 626 (56.7)

243 (21.0) 160 (14.5) 105 (9.5) 210 (19.0)

Values are n (%). a Categories correspond the stages of chronic kidney disease (CKD) in the NKF/KDOQI classification.

All formulas obtained a slightly lower predictive potential in patients with near-to-normal renal function (eGFR ≥ 60 ml/min/1.73 m2 to b60 ml/min/1.73 m2). Nevertheless, the Cockroft–Gault formula was the most accurate in both subsets (Table 5). All four formulas showed significantly lower median eGFR values in the three predefined subgroups: women patients, elderly patients, and patients with severely impaired left ventricular systolic function (Table 6, online only). The Cockcroft–Gault continued to be the most accurate formula when analyzing the performance of the five formulas in the prespecified subgroups of vulnerable patients (Table 7). When patients were dichotomized according to the median BMI, only the Cockroft–Gault formula showed a significant difference in the median eGFR (eGFR for BMI ≤ 25.95 kg/m2: 40.2 ml/min [IQR: 40.2– 57.0 ml/min] vs eGFR for BMI N 25.95 kg/m2: 50.1 ml/min [IQR: 34.7– 71.8 ml/min], p b 0.01). Furthermore, the AUC to estimate death

This large prospective multicenter study was performed to test the hypothesis that calculating eGFR using the Cockroft–Gault formula was more accurate in predicting mortality in AHF than the MDRD-4, MDRD-6, and CKD-EPI/BIS-1 formulas. We report four main findings. First, renal function at ED presentation, as quantified by all four formulas, was a predictor of death both in univariate and multivariate analyses. Second, prognostic accuracy, as quantified by time-dependent AUC and NRI, was significantly higher for eGFR calculated by the Cockroft–Gault formula than by MDRD-4, MDRD-6 or CKD-EPI/BIS-1 formulas. Third, this second finding was consistent in the prespecified vulnerable subgroups — women patients, elderly patients, and patients with severely systolic dysfunction — and also in patients with preserved renal function and in patients with impaired renal function. Fourth, renal function, as quantified by all four formulas, failed to show relevant prognostic value to predict all-cause rehospitalization or AHF rehospitalization. Our findings in patients with AHF extend and corroborate previous work performed mainly in patients with CHF [14–17]. Our hypothesis was largely based on a seminal study in a Spanish cohort of 925 stable ambulatory CHF patients with reduced left ventricular ejection fraction who reported that calculating eGFR using the Cockroft–Gault equation outperformed calculations using other serum-creatinine-based formulas to predict death [15]. Our work extends this observation to AHF and highlights that the Cockroft–Gault equation is also superior in a more heterogeneous population. Like previous researchers [15], we postulate that the inclusion of body weight as a variable in the formula might explain, at least in part, the superiority of the Cockroft–Gault equation. In obese patients (high body weight in the numerator), the Cockroft–Gault equation calculates a higher eGFR for any given serum creatinine, possibly reflecting the obesity paradox in the prediction of mortality risk, described in heart failure and renal insufficiency [22–24]. This argument is supported by the observation that when patients were dichotomized according to BMI (below versus above median BMI), only the Cockroft–Gault formula showed a significant difference in the median eGFR (higher eGFR in patients with higher BMI). It is important to highlight that we used anamnestic body weight recorded at presentation. This method is easy to apply, whereas obtaining actual body weight is cumbersome and therefore not standard practice at ED presentation.

Table 3 Qualitative and quantitative correlations between the eGFR formulas.

Spearman's rho correlation Cohen's Kappa (4 categories) Cohen's Kappa (2 categories)

CGa-MDRD4

CGa-MDRD-6

CGa-CKDEPI/BIS1

MDRD4-MDRD6

MDRD4-CKDEPI/BIS-1

MDRD6-CKDEPI/BIS-1

0.93 (0.92–0.94) 0.49 (0.44–0.53)

0.92 (0.91–0.93) 0.19 (0.15–0.22)

0.93 (0.92–0.94) 0.68 (0.65–0.73)

0.99 (0.98–0.99) 0.49 (0.44–0.53)

0.98 (0.98–0.98) 0.53 (0.49–0.57)

0.97 (0.97–0.98) 0.18 (0.15–0.22)

0.65 (0.61–0.70)

0.45 (0.42–0.53)

0.77 (0.73–0.81)

0.70 (0.65–0.73)

0.58 (0.54–0.62)

0.36 (0.32–0.40)

Values are correlation coefficient and confidence interval. Four categories: eGFR ≥90, 60–89, 30–59 and b30 ml/min/1.73 m2; two categories: eGFR ≥ 60, b60 ml/min/1.73 m2. a Cockroft–Gault.

654

Z.M. Weidmann et al. / International Journal of Cardiology 201 (2015) 650–657

Fig. 2. Pairwise correlation and agreement analyses between the 4 formulas evaluated with scatterplot and Bland–Altman method. Left: scatterplot. Right: Bland–Altman analysis. CG: Cockroft–Gault. x- and y-axis values are eGFR in ml/min/1.73 m2 for MDRD-4, MDRD-6 and CKD-EPI/BIS-1 and ml/min for Cockroft–Gault. eGFR values were similar among patients with an impaired renal function (eGFR b 60 ml/min/1.73 m2), however differences between equations increased with rising eGFR. In eGFR N 60 ml/min/1.73 m2, MDRD-6 showed systematically higher values when comparing with the other 4 formulas and CKD-EPI/BIS-1 showed systematically lower values. When comparing Cockroft–Gault formula with MDRD-4, there were 8.9 points (mean) difference, with lower values when using the Cockroft–Gault formula, but without any clear pattern (95% LoA −16.7 to +34.4). The greater agreement was found between Cockroft–Gault formula and CKD-EPI/BIS-1, with a −0.2 points (mean) of overestimation with Cockroft–Gault formula (95% LoA −14.5 to +14.2).

Our knowledge about the accuracy of eGFR creatinine-based formulas in the acute setting is scarce and derived from studies [17] that compared the more recently developed MDRD-4 and CKD-EPI formulas but did not include the Cockroft–Gault formula. A large Korean cohort of 1111 patients with AHF [16] compared the prognostic accuracy of the CKD-EPI and MDRD-4 and found the former to be superior. The prevalence of patients with renal dysfunction in this study

was lower than in our cohort. This could well explain the superiority of the CKD-EPI over MDRD-4 in their cohort, as CKD-EPI has shown to be a better estimate for filtration rate in patients with normal or only mildly impaired renal function [10]. In AHF patients typically seen in Europe and the US [5,17], with a high prevalence of renal dysfunction, accuracy in predicting death using the CKD-EPI formula seems to be lower.

Fig. 3. 2-year time-dependent ROC curve comparing the death prediction performance for each formula. AUC = area under the curve with a cut-off of 2 years; comparison was made using a Wilcoxon test.

Z.M. Weidmann et al. / International Journal of Cardiology 201 (2015) 650–657

655

Fig. 4. 2-year mortality Kaplan–Meier survival curves for the 4 eGFR formulas stratified by the stages of the NKF K/DOQI classification. p-Values evaluated with Log-Rank test.

The BIS-1 formula was developed in white participants from one of the largest German statutory health insurance schemes [12]. It has been externally validated in large cohorts [25,26], showing a more accurate GFR estimation than the CKD-EPI equation, especially in elderly patients with eGFR ≤ 60 ml/min/1.73 m2. However, there are no studies

evaluating its prognostic ability in our particular setting of patients with AHF. As a first step in our study, we calculated CKD-EPI for all patients, but because almost 76% of these participants were 70 years or older, we decided to perform a second step, adapting the formula by incorporating the BIS-1 equations for elderly patients. Remarkably,

Table 4 Multivariate Cox models for the different equations entered as quantitative and categorical variables.

HRa HRa by categories –eGFR ≥ 90 ml/min/1.73 m2 –eGFR 60–89 ml/min/1.73 m2 –eGFR 30–59 ml/min/1.73 m2 –eGFR b 30 ml/min/1.73 m2

Cockroft–Gault

MDRD-4

MDRD-6

CKD-EPI/BIS-1

0.97 (0.98–0.99)

0.98 (0.98–0.99)

0.99 (0.98–0.99)

0.98 (0.97–0.99)

1 0.99 (0.47–2.06) 1.44 (0.66–2.80) 2.43 (1.16–5.10)

1 0.70 (0.45–1.07) 1.26 (0.84–1.88) 1.94 (1.26–2.99)

1 1.17 (0.82–1.66) 2.23 (1.60–3.10) 3.45 (2.32–5.14)

1 0.79 (0.39–1.61) 1.02 (0.52–2.02) 2.07 (1.03–4.15)

Adjusted for age (years), BMI (kg/m2), systolic blood pressure (mm Hg) and BNP (pg/ml) at presentation. a HR = hazard ratio. Categories correspond to the stages of chronic kidney disease (CKD) in the K/DOQI (Kidney Disease Outcomes Quality Initiative) CKD classification.

656

Z.M. Weidmann et al. / International Journal of Cardiology 201 (2015) 650–657

Table 5 Prediction value of the formulas stratifying by normal/slightly impaired renal function and impaired renal function evaluated with a ROC curve.

eGFR ≥ 60 ml/min/1.73 m2 eGFR b 60 ml/min/1.73 m2

Cockroft–Gault

MDRD-4

MDRD-6

CKD-EPI/BIS-1

0.60 (0.43–0.78) 0.71 (0.61–0.80)

0.50 (0.43–0.56) 0.59 (0.46–0.57)

0.57 (0.51–0.62) 0.51 (0.46–0.57)

0.48 (0.39–0.54) 0.65 (0.61–0.69)

Values are areas under the curve and 95% confidence interval.

Table 6 Median eGFR according to the different formulas in the prespecified subgroups.

Men Women ≤75 years N75 years LVEF ≥ 35% LVEF b 35%

Cockroft–Gault

MDRD-4

MDRD-6

CKD-EPI/BIS-1

48.8 (33.9–40.1) 41.7 (28.7–57.8) 62.1 (41.7–85.8) 39.3 (28.9–52.0) 48.6 (33.2–68.8) 43.5 (29.7–63.2)

58.7 (40.8–82.1) 50.9 (34.8–68.6) 64.7 (40.8–86.0) 51.5 (36.6–68.4) 59.1 (41.5–80.0) 50.5 (34.5–70.4)

72.5 (48.9–100.5) 63.1 (42.4–87.3) 79.6 (49.5–107.1) 63.4 (44.3–85.5) 74.2 (50.6–99.9) 62.8 (43.2–87.4)

49.7 (37.2–66.9) 43.6 (32.0–56.8) 60.6 (41.7–79.0) 42.3 (32.8–52.8) 49.0 (36.5–64.6) 44.2 (33.1–63.5)

Values are median (IQR). p b 0.02 for all comparisons using Wilcoxon–Mann–Whitney test, except for the comparison of Cockroft–Gault and CKD-EPI/BIS-1 among patients with LVEF ≥ 35%.

neither equation showed a better ability to predict mortality than the Cockroft–Gault formula. More accurate risk prediction regarding long-term mortality in AHF has noteworthy clinical consequences. The GFR is widely accepted as the best index of renal function, and its estimation using creatininebased formulas has become a simple decision-making tool for the practicing clinician to determine the prescription, timing and under- or over-dosing of some medications, and other therapeutic strategies that may affect outcome. Patients with impaired function are less likely to be prescribed efficacious treatments, despite evidence of better outcomes if they receive these drugs [27]. This observation was confirmed in our cohort; patients who died had a worse renal function and were less likely to receive CHF-directed treatments (β-blockers and ACEI or AT-2 inhibitors). It is important to highlight that our aim was not to evaluate the efficacy of the four formulas with respect to a gold standard. Our objective

was exclusively to compare the value of the formulas most commonly used in clinical practice to predict death in patients with AHF. The pathophysiology of renal function in AHF is complex and multifactorial: it includes common cardio-renal risk factors, hemodynamic disturbance, fluctuation in creatinine production and excretion, potential druginduced effects of AHF therapy on renal function parameters, and the degree of malnutrition or cachexia [4,28,27]. These factors make it highly challenging to calculate the actual filtration rate, a concern that was beyond the scope of the present study. Some limitations to our study should be mentioned. First, as most of the patients were Caucasian, we were unable to evaluate a possible role of ethnicity on the performance of the formulas to predict death. Second, as we excluded patients on chronic hemodialysis, we cannot comment on the risk prediction in this small high-risk subgroup of AHF patients. Third, we can only hypothesize whether the superior risk prediction achieved with the Cockroft–Gault formula can be used clinically

Table 7 Performance of eGFR formulas for prediction of death in prespecified subgroups.

Men Women ≤75 years N75 years LVEF ≥ 35% LVEF b 35%

Cockroft–Gault

MDRD-4

MDRD-6

CKD-EPI/BIS-1

p-Valuea

p-Valueb

p-Valuec

0.72 (0.68–0.75) 0.69 (0.64–0.73) 0.68 (0.64–0.73) 0.67 (0.63–0.70) 0.67 (0.62–0.71) 0.73 (0.67–0.79)

0.67 (0.64–0.73) 0.63 (0.59–0.67) 0.66 (0.61–0.71) 0.63 (0.59–0.67) 0.62 (0.57–0.67) 0.68 (0.62–0.74)

0.69 (0.63–0.71) 0.65 (0.60–0.69) 0.68 (0.63–0.73) 0.65 (0.61–0.68) 0.63 (0.58–0.66) 0.70 (0.63–0.75)

0.70 (0.65–0.74) 0.66 (0.61–0.71) 0.65 (0.58–0.72) 0.64 (0.60–0.69) 0.64 (0.59–0.69) 0.71 (0.65–0.79)

b0.01 b0.01 0.01 b0.01 b0.01 b0.01

b0.01 b0.01 0.71 b0.01 b0.01 b0.01

b0.01 b0.01 0.02 b0.01 b0.01 b0.01

Values are areas under the curve and 95% confidence interval. Comparisons were done using DeLong test. a p-Value: Cockroft–Gault vs MDRD-4. b p-Value: Cockroft–Gault vs MDRD-6. c p-Value: Cockroft–Gault vs CKD-EPI/BIS-1.

Table 8 Performance of eGFR formulas for prediction of death dichotomizing according to the median BMI.

BMI ≤ 25.95 kg/m2 BMI N 25.95 kg/m2

Cockroft–Gault

MDRD-4

MDRD-6

CKD-EPI/BIS-1

p-Valuea

p-Valueb

p-Valuec

0.68 (0.64–0.72) 0.71 (0.67–0.74)

0.65 (0.60–0.68) 0.67 (0.63–0.71)

0.67 (0.62–0.71) 0.68 (0.64–0.72)

0.66 (0.62–0.70) 0.70 (0.65–0.7471)

b0.01 b0.01

0.03 b0.01

b0.01 0.02

Values are areas under the curve and 95% confidence interval. Comparisons were done using DeLong test. a p-Value: Cockroft–Gault vs MDRD-4. b p-Value: Cockroft–Gault vs MDRD-6. c p-Value: Cockroft–Gault vs CKD-EPI/BIS-1.

Z.M. Weidmann et al. / International Journal of Cardiology 201 (2015) 650–657

to improve patient outcomes. However, we think our findings remind clinicians about the merits of simple, outdated clinical tools such as the Cockroft–Gault formula. Fourth, as cystatin C was not measured in this cohort, our data cannot answer the remaining question of whether the Cockroft–Gault formula is also superior to cystatin C-based formulas [29] in the prediction of death. Cystatin C, a low molecular weight protein filtered at the glomerulus and not reabsorbed, may even be a more accurate predictor of early acute kidney injury and post-discharge outcome in AHF than creatinine-based formulas [30]. However, cystatin C has its own inherent limitations, including the influence of other nonrenal factors such as thyroid disorders, inflammation, fat mass and diabetes [31]. In a study of 526 patients with AHF [17], the authors compared outcome prediction using four formulas: the MDRD4 equation, and the CKD-EPI equation based on serum creatinine alone, based on cystatin C alone, or based on both serum creatinine and cystatin C. The CKD-EPI equation based on cystatin C alone was superior to the other three equations to predict the combined end point of death and/ or readmission due to AHF. Further studies are needed, however, because the CKD-EPI equation based on cystatin C has not been directly compared with the Cockroft–Gault formula and because cystatin C values are influenced by many factors, as mentioned above. Moreover, definitions incorporating the use of cystatin C [32,33] are lacking. Until these data become available, it seems likely that cystatin C will be used almost exclusively in research settings. In conclusion, quantifying renal function at ED presentation using the Cockcroft–Gault formula in patients with AHF assesses the risk of mortality more accurately than other commonly used formulas. In contrast, eGFR does not seem helpful to predict recurrent hospitalization. Conflict of interest All authors declare that they have no conflict of interest with this study. The sponsors had no role in the design and conduct of the study, in the collection, management, analysis, and interpretation of the data, or in the preparation, review and approval of the manuscript. Z.M.W. and C.M. had full access to all the data in the study and are responsible for the integrity of the data and the accuracy of the data analysis. Acknowledgments We wish to thank the patients for their participation in the study, and the staff at the participating EDs, the research coordinators, and the laboratory technicians (particularly Michael Freese, Claudia Stelzig, Irina Klimmeck, Janine Voegele, Beate Hartmann, and Fausta Chiaverio) for their valuable collaboration. References [1] J.J.V. McMurray, S. Adamopoulos, S.D. Anker, A. Auricchio, M. Böhm, K. Dickstein, et al., ESC Guidelines for the Diagnosis and Treatment of Acute and Chronic Heart Failure 2012: The Task Force for the Diagnosis and Treatment of Acute and Chronic Heart Failure 2012 of the European Society of Cardiology. Developed in collaboration with the Heart, Eur. Heart J. 33 (2012) 1787–1847. [2] C.W. Yancy, M. Jessup, B. Bozkurt, J. Butler, D.E. Casey, M.H. Drazner, et al., ACCF/AHA Guideline for the Management of Heart Failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force On Practice Guidelines, Circulation 2013 (2013) 128. [3] A.S. Go, D. Mozaffarian, V.L. Roger, E.J. Benjamin, J.D. Berry, M.J. Blaha, et al., Heart disease and stroke statistics 2014 update, A Report From the American Heart Association, vol. 1292014. [4] C. Ronco, M. Cicoira, P. a McCullough, Cardiorenal syndrome type 1: pathophysiological crosstalk leading to combined heart and kidney dysfunction in the setting of acutely decompensated heart failure, J. Am. Coll. Cardiol. 60 (2012) 1031–1042. [5] T. Breidthardt, T. Socrates, M. Noveanu, T. Klima, C. Heinisch, T. Reichlin, et al., Effect and clinical prediction of worsening renal function in acute decompensated heart failure, Am. J. Cardiol. 107 (2011) 730–735. [6] M. Metra, G. Cotter, B.A. Davison, G.M. Felker, G. Filippatos, B.H. Greenberg, et al., Effect of serelaxin on cardiac, renal, and hepatic biomarkers in the relaxin in acute heart failure (RELAX-AHF) development program: correlation with outcomes, J. Am. Coll. Cardiol. 61 (2013) 196–206.

657

[7] D.W. Cockcroft, M.H. Gault, Prediction of creatinine clearance from serum creatinine, Nephron 16 (1976) 31–41. [8] A.S. Levey, J.P. Bosch, J.B. Lewis, T. Greene, N. Rogers, D. Roth, A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group, Ann. Intern. Med. 130 (1999) 461–470. [9] A.S. Levey, T. Greene, J.W. Kusek, BG, A Simplified Equation to Predict Glomerular Filtration Rate From Serum Creatinine, 2000 (Suppl.:155A). [10] A.S. Levey, L.A. Stevens, C.H. Schmid, Y.L. Zhang, A.F. Castro III, H.I. Feldman, et al., A new equation to estimate glomerular filtration rate, Ann. Intern. Med. 150 (2009) 604–612. [11] H.A. Rolin, P.M. Hall, R. Wei, Inaccuracy of estimated creatinine clearance for prediction of iothalamate glomerular filtration rate, Am. J. Kidney Dis. 4 (1984) 48–54. [12] E.S. Schaeffner, N. Ebert, P. Delanaye, U. Frei, J. Gaedeke, O. Jakob, et al., Two novel equations to estimate kidney function in persons aged 70 years or older, Ann. Intern. Med. 157 (2012) 471–481. [13] T.D.J. Smilde, D.J. van Veldhuisen, G. Navis, A. a Voors, H.L. Hillege, Drawbacks and prognostic value of formulas estimating renal function in patients with chronic heart failure and systolic dysfunction, Circulation 114 (2006) 1572–1580. [14] M. Plischke, S. Neuhold, M. Kohl, G. Heinze, G. Sunder-Plassmann, R. Pacher, et al., Renal function in heart failure: a disparity between estimating function and predicting mortality risk, Eur. J. Heart Fail. 15 (2013) 763–770. [15] E. Zamora, J. Lupón, J. Vila, A. Urrutia, M. de Antonio, H. Sanz, et al., Estimated glomerular filtration rate and prognosis in heart failure: value of the Modification of Diet in Renal Disease Study-4, chronic kidney disease epidemiology collaboration, and Cockroft–Gault formulas, J. Am. Coll. Cardiol. 59 (2012) 1709–1715. [16] J. Oh, S. Kang, N. Hong, J. Youn, S. Han, E. Jeon, The CKD-EPI is more accurate in clinical outcome prediction than MDRD equation in acute heart failure: Data from the Korean Heart Failure (KorHF) Registry, Int. J. Cardiol. 167 (3) (2013) 1084–1087. [17] S. Manzano-Fernández, P.J. Flores-Blanco, J.I. Pérez-Calvo, F.J. Ruiz-Ruiz, F.J. CarrascoSánchez, J.L. Morales-Rull, et al., Comparison of risk prediction with the CKD-EPI and MDRD equations in acute decompensated heart failure, J. Card. Fail. 19 (2013) 583–591. [18] A.S. Levey, J. Coresh, T. Greene, J. Marsh, L.A. Stevens, J.W. Kusek, et al., Expressing the modification of diet in renal disease study equation for estimating glomerular filtration rate with standardized serum creatinine values, Clin. Chem. 53 (2007) 766–772. [19] A. Khwaja, KDIGO clinical practice guidelines for acute kidney injury, Nephron Clin. Pract. 2 (2012). [20] P.J. Heagerty, T. Lumley, M.S. Pepe, Time-dependent ROC Curves for Censored Survival Data and a Diagnostic Marker, 2000 337–344. [21] M.J. Pencina, R.B. D'Agostino, E.W. Steyerberg, Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers, Stat. Med. 30 (2011) 11–21. [22] C.J. Lavie, M.A. Alpert, R. Arena, M.R. Mehra, R.V. Milani, H.O. Ventura, Impact of obesity and the obesity paradox on prevalence and prognosis in heart failure, JACC Heart Fail. 1 (2013) 93–102. [23] M. Noveanu, T. Breidthardt, S. Cayir, M. Potocki, K. Laule, C. Mueller, B-type natriuretic peptide-guided management and outcome in patients with obesity and dyspnea—results from the BASEL study, Am. Heart J. 158 (2009) 488–495. [24] K. Kalantar-Zadeh, Causes and consequences of the reverse epidemiology of body mass index in dialysis patients, J. Ren. Nutr. 15 (2005) 142–147. [25] E. Vidal-Petiot, J.-P. Haymann, E. Letavernier, F. Serrano, C. Clerici, J.-J. Boffa, et al., External validation of the BIS (Berlin Initiative Study)-1 GFR estimating equation in the elderly, Am. J. Kidney Dis. 63 (2014) 865–867. [26] I.M. Alshaer, H.S. Kilbride, P.E. Stevens, G. Eaglestone, S. Knight, J.L. Carter, et al., External validation of the Berlin equations for estimation of GFR in the elderly, Am. J. Kidney Dis. 63 (2014) 862–865. [27] J. Ezekowitz, F. a McAlister, K.H. Humphries, C.M. Norris, M. Tonelli, W. a Ghali, et al., The association among renal insufficiency, pharmacotherapy, and outcomes in 6427 patients with heart failure and coronary artery disease, J. Am. Coll. Cardiol. 44 (2004) 1587–1592. [28] M. Haase, C. Müller, K. Damman, P.T. Murray, J.A. Kellum, C. Ronco, et al., Pathogenesis of cardiorenal syndrome type 1 in acute decompensated heart failure: Workgroup Statements From the Eleventh Consensus Conference of the Acute Dialysis Quality Initiative (ADQI), Contrib. Nephrol. 182 (2013) 99–116. [29] L.A. Stevens, J. Coresh, C.H. Schmid, H.I. Feldman, M. Froissart, J. Kusek, et al., Estimating GFR using serum cystatin C alone and in combination with serum creatinine: a pooled analysis of 3418 individuals with CKD, Am. J. Kidney Dis. 51 (2008) 395–406. [30] J.P.E. Lassus, M.S. Nieminen, K. Peuhkurinen, K. Pulkki, K. Siirilä-Waris, R. Sund, et al., Markers of renal function and acute kidney injury in acute heart failure: definitions and impact on outcomes of the cardiorenal syndrome, Eur. Heart J. 31 (2010) 2791–2798. [31] L.A. Stevens, C.H. Schmid, T. Greene, L. Li, G.J. Beck, M.M. Joffe, et al., Factors other than glomerular filtration rate affect serum cystatin C levels, Kidney Int. 75 (2009) 652–660. [32] S.G. Coca, R. Yalavarthy, J. Concato, C.R. Parikh, Biomarkers for the diagnosis and risk stratification of acute kidney injury: a systematic review, Kidney Int. 73 (2008) 1008–1016. [33] J. Macdonald, S. Marcora, M. Jibani, G. Roberts, M. Kumwenda, R. Glover, et al., GFR estimation using cystatin C is not independent of body composition, Am. J. Kidney Dis. 48 (2006) 712–719.