Author + information
- Received March 4, 2013
- Revision received April 29, 2013
- Accepted May 3, 2013
- Published online October 1, 2013.
- Kevin Damman, MD, PhD∗∗ (, )
- Serge Masson, PhD†,
- Hans L. Hillege, MD, PhD∗,‡,
- Adriaan A. Voors, MD, PhD∗,
- Dirk J. van Veldhuisen, MD, PhD∗,
- Patrick Rossignol, MD, PhD§,
- Gianni Proietti, MD‖,
- Savino Barbuzzi, MD¶,
- Gian Luigi Nicolosi, MD#,
- Luigi Tavazzi, MD∗∗,
- Aldo P. Maggioni, MD†† and
- Roberto Latini, MD†
- ∗Department of Cardiology, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands
- †Department of Cardiovascular Research, IRCCS - Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy
- ‡Department of Epidemiology, University Medical Center Groningen, Groningen, the Netherlands
- §INSERM, Centre d'Investigations Cliniques, and Université de Lorraine, INSERM U961, Nancy, France
- ‖UO Cardiologia Territoriale, Ospedale di Terni, Terni, Italy
- ¶Servizio di Cardiologia, Ospedale di Venosa, Venosa, Italy
- #Azienda Ospedaliera Santa Maria degli Angeli, UO Cardiologia, Pordenone, Italy
- ∗∗GVM Care and Research, Ettore Sansavini Health Science Foundation–Maria Cecilia Hospital, Cotignola, Italy
- ††ANMCO Research Center, Florence, Italy
- ↵∗Reprint requests and correspondence:
Dr. Kevin Damman, Department of Cardiology, University Medical Center Groningen, Hanzeplein 1, 9700 RB, Groningen, the Netherlands.
Objectives This study sought to investigate the relationship between tubular damage and worsening renal function (WRF) in chronic heart failure (HF)
Background WRF is associated with poor outcome in chronic HF. It is unclear whether urinary tubular markers may identify patients at risk for WRF.
Methods In 2,011 patients with chronic HF, we evaluated the ability of urinary tubular markers (N-acetyl-beta-d-glucosaminidase (NAG), kidney injury molecule (KIM)-1, and neutrophil gelatinase-associated lipocalin (NGAL) to predict WRF. Finally, we assessed the prognostic importance of WRF.
Results A total of 290 patients (14.4%) experienced WRF during follow-up, and WRF was a strong and independent predictor of all-cause mortality and HF hospitalizations (hazard ratio [HR]: 2.87; 95% CI: 2.40 to 3.43; p < 0.001). Patients with WRF had lower baseline glomerular filtration rate and higher KIM-1, NAG, and NGAL levels. In a multivariable-adjusted model, KIM-1 was the strongest independent predictor of WRF (HR: 1.23; 95% CI: 1.09 to 1.39 per log increase; p = 0.001).
Conclusions WRF was associated with strongly impaired outcome in patients with chronic HF. Increased level of urinary KIM-1 was the strongest independent predictor of WRF and could therefore be used to identify patients at risk for WRF and poor clinical outcome. (GISSI-HF–Effects of n-3 PUFA and Rosuvastatin on Mortality-Morbidity of Patients With Symptomatic CHF; NCT00336336)
Renal dysfunction, one of the most common and most important comorbidities in chronic heart failure (HF) is associated with unfavorable outcome, including mortality. Additionally, the development of worsening renal function (WRF) further impairs clinical outcome, including mortality and HF hospitalizations (1,2). Therefore, identifying patients at risk for WRF might lead to better use of life-saving therapies, less frequent hospital visits, and maybe even improved outcome.
Recently, more interest has been shown in markers of tubular injury than serum creatinine because these markers are more sensitive and specific to early changes in renal function, whereas serum creatinine is slowly affected and is dependent on muscle mass and age. We recently showed that these tubular damage markers confer excess risk in patients with chronic HF (3). However, the hypothetical potential of these markers to predict occurrence of deterioration of renal function over time has not been evaluated in patients with chronic HF.
In the present substudy of the GISSI-HF (Gruppo Italiano per lo Studio della Sopravvivenza Nell'Infarto Miocardico) trial, we set out to determine the prevalence of WRF, predictors of the occurrence of WRF and clinical significance of tubular markers in predicting WRF.
The design and main results of the GISSI-HF trial have been described in detail elsewhere (4,5). The institutional review committee at each participating center approved the study, and all patients gave informed consent. A total of 2,130 patients provided a first-morning spot urine at any of the clinical visits scheduled in the trial. The mean time to urine sampling from randomization date in the trial was 1.2 years (range 0 to 3.5 years). Urine samples were stored at −70°C before analysis.
Assessment of renal function, WRF, and tubular markers
Serum creatinine was measured, per protocol, at all clinical visits in local laboratories. Glomerular filtration rate (eGFR) (ml/min/1.73 m2) was estimated using the simplified modification of diet in renal disease formula. Baseline GFR was set at the time of urine collection. WRF was defined as ≥0.3 mg/dl and ≥25% increase in serum creatinine at any moment during follow-up. For sensitivity analysis, a reduction in eGFR >20% and an absolute increase >0.3 mg/dl in creatinine (both separately) were also considered.
N-acetyl-beta-d-glucosaminidase (NAG), kidney injury molecule (KIM)-1, and neutrophil gelatinase-associated lipocalin (NGAL) levels were determined in urine (3). In short, NGAL was determined by means of a commercially available enzyme-linked immunosorbent assay (ELISA) kit from Antibody Shop (Gentofte, Denmark). NGAL was expressed in ng/ml (limit of detection 0.093 ng/ml). The enzymatic activity of NAG was evaluated using the substrate p-nitrophenyl N-acetyl-beta-d-glucosaminide (Sigma, St. Louis, Missouri; limit of detection 0.113 U/ml). Urinary KIM-1 measurements were performed using an ELISA (R&D Systems, Minneapolis, Minnesota; limit of detection 0.125 ng/ml). All urinary tubular marker concentrations were normalized to urinary creatinine concentrations to allow for correction for dilution and concentration of urine.
The primary analysis of this study was the ability of tubular markers to predict WRF. The outcome measure for this analysis was the first occurrence of WRF after the day of urine collection. Secondary outcome was the combined endpoint of the first occurrence of either death or HF hospitalization. The latter was defined as any hospitalization for HF induced by infection, supraventricular or ventricular arrhythmias, acute coronary syndromes, or renal dysfunction due to drug effects or worsening cardiac function. All events of this combined endpoint were centrally validated by an ad hoc committee blinded to study treatments.
Data are given as mean ± SD when normally distributed, as median and interquartile range when distribution is skewed, and as frequencies and percentages for categorical variables. Student t test or Mann-Whitney U test was used to determine significant differences of variables between patients with and without WRF. We used a Cox proportional hazards model to estimate hazard ratios (HRs) with 95% CI for the occurrence of WRF. In multivariable analysis, all univariately associated variables were introduced in a multivariable model. Interaction terms with p value <0.1 were included in the multivariable model when considered clinically relevant. In a secondary model, stepwise backward Cox regression analysis was carried out with all variables. After this model was constructed, optimal cutoff points were determined by the highest sum of specificity and sensitivity calculated using receiver-operator curve analysis. Cutoffs for the model were then chosen at clinically relevant or numerically easy cutoff points. In sensitivity analysis, bootstrapped forward Cox regression was carried out via the swboot syntax in STATA to investigate the relative importance of each included variable. Internal validation of additive power of KIM-1 to the model was carried out by comparison of Harrell C indices in 2 randomly created data sets from the entire study population by using the Somersd syntax in STATA. Afterward, a risk score was constructed (Appendix). For the combined endpoint of death and HF hospitalization, another Cox regression analysis was carried out using all univariate-associated variables. In this model, WRF was treated as a time-varying covariate. All reported probability values are 2-tailed, and a p value <0.05 was considered statistically significant. Statistical analyses were performed using SPSS, version 12.0 (Chicago, Illinois) and STATA, version 11.0 (College Station, Texas).
A total of 2,011 of 2,130 patients had at least 1 serial serum creatinine measurement available and formed the study population (mean number of serum creatinine measurements available 3.8 ± 1.3). Baseline variables are presented in Table 1. During follow-up, 290 patients (14.4%) developed WRF, with a mean time to WRF of 571 ± 270 days. Patients who developed WRF had relatively more serum creatinine values available (4.1 ± 1.4 vs 3.8 ± 1.3; p < 0.001). Furthermore, they were older and had higher body mass index scores, more severe HF with higher heart rates, lower left ventricular ejection fraction, more symptoms, and more often increased jugular venous pressure. Baseline GFR was lower in patients with WRF (60 vs. 69 ml/min/1.73 m2; p < 0.001). They displayed increased albuminuria (14.4 vs. 7.7 mg/g Cr; p < 0.001). All tubular marker concentrations were increased in patients with WRF: NAG 15.9 (9.9 to 26.9) versus 12.6 (7.4 to 20.1) U/g Cr, p < 0.001; KIM-1 2599 (1,246 to 5,076) versus 1783 (615 to 3,493) ng/g Cr, p < 0.001; and NGAL 50 (20 to 126) versus 32 (13 to 84) μg/g Cr, p < 0.001. Loop diuretic agents, digitalis, amiodarone, and nitrates were more frequently prescribed in patients with WRF, whereas the use of aspirin and beta-blockers were less frequently prescribed in patients with WRF.
eGFR for patients who had low tubular marker levels (none in the highest quartile) showed a mean decrease of −2.6 ± 15 ml/ min/1.73 m2 versus patients in the highest quartile (NAG −4.1 ± 16 min/1.73 m2; NGAL −2.9± 16 min/1.73 m2; or KIM-1 −5.5 ± 16 min/1.73 m2), but only the latter showed a significant difference from patients who had no increased tubular marker levels (p = 0.003). Patients without WRF had stable eGFR, whereas patients with WRF had a significant decrease in eGFR across the whole study period (−1.3 ± 15 vs. −12 ± 19.8 ml/min/1.73 m2, after 839 ± 140 days, respectively).
Prediction of WRF
In univariate analysis, KIM-1 (hazard ratio [HR]: 1.42; 95% CI: 1.28 to 1.58 per log increase; p < 0.001) was one of the strongest predictors of WRF. NAG and NGAL were also significant predictors of WRF in univariate analysis (Table 2). Figure 1 shows the cumulative hazard for WRF with quartiles of KIM-1, NAG, and NGAL, as well as stages of chronic kidney disease and albuminuria, showing an increase in the incidence of WRF with increasing concentrations of tubular markers and albuminuria, or decreasing eGFR. In multivariable Cox regression, KIM-1 remained an independent predictor of the occurrence of WRF (HR: 1.23; 95% CI: 1.09 to 1.39 per log increase; p = 0.001), even after adjustment for NAG and NGAL (Table 2). Both NAG and NGAL did not predict WRF after adjustment. Using different definitions of WRF gave similar results for KIM-1, and NGAL also predicted WRF when an absolute increase in serum creatinine was used as the definition. We found 2 interaction terms that remained significant in multivariable analysis. KIM-1 predicted WRF better in patients with more preserved left ventricular ejection fraction, whereas higher body mass index predicted WRF better in patients without loop diuretic therapy. We found no interaction between time to urine collection or number of serum creatinine measurements available and relationship between KIM-1 and WRF. In a stepwise Cox regression analysis (Table 3), the presence of chronic obstructive pulmonary disease was a significant predictor of WRF as well. In sensitivity analysis, KIM-1 was the most important and most robust determinant of WRF, being selected in 85% of all models, with eGFR being second most important (Online Table 1). NAG and NGAL were only selected in a minority of models. Addition of KIM-1 to the entire multivariate model (without KIM-1) resulted in significant improvement of the C-statistic (0.695 to 0.703; p = 0.005). In reclassification analysis, addition of KIM-1 to this model resulted in significant improvement in reclassification: net reclassification index (NRI) = 0.18; 95% CI: 0.12 to 0.25; p < 0.001 (cutoff 5%, 10%, 20%), integrated discrimination improvement (IDI) = 0.011; 95% CI: 0.005 to 0.016; p < 0.001 (Online Table 2). Continuous NRI without cutoff points also showed significant improvement (33%). In comparison, adding albuminuria (either micro or macro) to the model instead of KIM-1 resulted in NRI = 0.12; 95% CI: 0.06 to 0.17; p < 0.001 and IDI = 0.006; 95% CI: 0.002 to 0.011; p = 0.004. Internal validation of the additive power of KIM-1 revealed significant improvement of the Harrell C index (0.684 [0.642 to 0.726]) vs. 0.657 (0.613 to 0.701); p < 0.001).
Risk score for WRF
On the basis of the final stepwise Cox regression model, we constructed a risk model for WRF (Online Appendix). A high-risk score of >28 carried the highest risk for WRF (HR: 6.08; 95% CI: 4.09 to 9.04; p < 0.001) versus low-risk score of <13.
Mean follow-up time was 2.7 (2.5 to 3.1) years. A total of 563 patients reached the combined endpoint of all-cause mortality or HF hospitalization. In univariate analysis, the development of WRF was associated with increased risk of the primary endpoint (HR: 3.63; 95% CI: 3.07 to 4.30; p < 0.001). Table 4 presents the multivariable analysis, showing that the occurrence of WRF remained independently associated with impaired outcome, whereas baseline urine albumin to creatinine ratio did not. Also in this subanalysis, NAG showed prognostic information, independent from the occurrence of WRF, whereas eGFR showed a trend toward an effect with outcome; NGAL and KIM-1 were not independently associated with outcome in this analysis.
This is the first study to assess urinary tubular proteins as predictors of deterioration of renal function in chronic HF. The urinary tubular protein KIM-1 was a strong and independent predictor of WRF in this large cohort of patients with chronic HF.
Incidence of WRF and clinical outcome
Depending on the definition, the occurrence of WRF has been reported to be approximately 25% in acute and chronic HF (1). In this study, WRF was present in a smaller number of patients. The lower incidence might be explained by a more stable chronic HF population and by the fixed time interval of serum creatinine determinations, in contrast to observational studies, in which creatinine is often only measured when deemed necessary by investigators. This increases the likelihood of finding WRF in patients monitored more closely. In our analysis, patients with WRF had more than double the chance of the combined endpoint in adjusted analysis, which corresponds with findings of a meta-analysis of WRF in HF (1). Recent studies have suggested that the reason why WRF develops may be more important than the occurrence of WRF itself. Findings from RALES (Randomized Aldactone Evaluation Study) and SOLVD (Studies of Left Ventricular Dysfunction) suggest that WRF that develops during initiation of renin-angiotensin system inhibition may not be associated with poor outcome (6,7). This may also be the reason why KIM-1 predicts WRF, but not clinical outcome; it may identify a deterioration in renal function but not necessarily poor outcome. Our present results suggest that at least in this population of patients with stable chronic HF, WRF was strongly associated with poor outcome. In addition, we showed that patients who experienced WRF also experienced an accelerated decline in eGFR over time.
Prediction of WRF—traditional factors
Several studies have examined factors associated with the occurrence of WRF in HF (8–12). Overall, the most consistent and important predictor of WRF is impaired baseline GFR. This may be attributed to a reduction in functioning nephron units, and thereby a reduction in residual renal function. Because many studies used absolute increases in serum creatinine as the definition of WRF, this may also be a reflection of regression to the mean. However, our present analysis further supports a functional association between WRF and baseline GFR.
Other factors associated with WRF in HF are the presence of hypertension and diabetes, severity of HF (ejection fraction), the use of diuretic agents (loop and mineralocorticoid receptor antagonists), and age (8–13). Importantly, in our present analysis, some of these entities reappeared, whereas others did not. Although diabetes was associated with WRF in univariate analysis, it was not an independent predictor of WRF, possibly due to the important effect of KIM-1. Diuretic agent therapy has been associated with WRF in other populations as well (9,14–16). It is unclear via which mechanism diuretic agents may impair WRF. Diuretic agents may cause intravascular volume depletion and activation of the renin-angiotensin system, which could have detrimental effects on GFR and prognosis. Surprisingly, we found that lack of aspirin use was associated with more frequent WRF. We can only speculate on the mechanism behind this association. One reason may be the presence of renal artery stenosis (either significant or nonsignificant), a situation in which aspirin may favorably affect renal perfusion (17). On the other hand, it may be a reflection of treatment bias, when secondary reasons to administer aspirin could have affected the frequency of WRF, but it is unclear from our analysis which confounder this could have been.
Prediction of WRF—tubular damage markers
In acute HF, plasma NGAL levels have been associated with the occurrence of acute kidney injury, which is consistent with findings in intensive care patients (18,19). In chronic HF, KIM-1 and NAG levels are susceptible to diuretic-induced subtle changes in volume status (20). Our present study extends findings of a prognostic importance of tubular markers in chronic HF (21–23). Urinary KIM-1 was not only the strongest tubular marker to predict WRF, but they were also by far the most prominent predictor of WRF among all associated variables. Patients with a high KIM-1 level showed a significantly faster decline in eGFR over time. Numerically, it predicted WRF better than either eGFR or albuminuria.
There may be several explanations for this finding. In the acute setting, tubular injury/damage markers are thought to appear in urine and plasma long before creatinine is affected because hypoxic tubular damage is instant, instead of the gradual reduction in creatinine clearance (24). Therefore, urinary tubular damage may be a more sensitive marker of deterioration of renal function. However, how this translates into deterioration over a longer period of time is difficult to interpret. The presence of tubular injury could be a marker of chronic renal hypoxia, which makes the kidney more vulnerable to hemodynamic changes and therefore WRF (25). Also, because tubular damage is linked to interstitial fibrosis and inflammation, the presence of increased tubular protein concentrations can be a representation of ongoing ischemic/neurohormonal damage, which causes a gradual decline in renal function. Finally, tubular markers may be a more sensitive marker of renal congestion as NAG and KIM-1 respond to diuretic therapy (26).
It is unclear why KIM-1 is the most prominent predictor and NAG and NGAL are more modest predictors. KIM-1 and NAG are strongly related to outcome in chronic HF, whereas NGAL shows additive prognostic value next to natriuretic peptides in acute HF (27). All proteins are specific tubular markers, but their site of production varies within the nephron. Both KIM-1 and NAG are secreted from proximal tubular cells, whereas NGAL is more distally formed in the loop of Henle and collecting ducts. There are no data whether a specific region within the kidney is more sensitive or susceptible to renal damage in HF, and we cannot derive a definite explanation from our data.
This study was a retrospective analysis of a randomized clinical trial. Not all patients had serum creatinine measurements available at all time points, and therefore bias could exist for patients with a higher number of serum creatinine values available. Indeed, patients who experienced WRF had slightly, but significantly, higher number of serum creatinine measurements available. These patients could have been monitored more closely for numerous reasons, one of which could be higher risk. WRF could also have occurred in the setting of a hospitalization for HF, but we were unable to account for this in our analysis. Interaction analysis showed no bias with respect to the main results. Also, we excluded patients who died before a first serial serum creatinine measurement was available, and therefore the prevalence and possible association with poor outcome of WRF could be underestimated. Natriuretic peptide levels were only available at the time of urine collection in a limited number of patients, and we were therefore unable to adjust the models for this important marker. Because of a lack of studies in HF with urinary collections available, we were not able to validate our risk score in a second cohort. Although we used rigorous internal validation, this risk score should be prospectively validated in future risk score–guided therapy studies.
WRF in chronic HF is associated with impaired outcome. Among other covariates, urinary KIM-1 level is the most important predictor of the occurrence of WRF. Future studies should address the clinical applicability of our findings.
For an expanded Methods section, and supplemental tables and figure, please see the online version of this article.
The GISSI-HF trial is endorsed by the Associazione Nazionale Medici Cardiologi Ospedalieri, Florence, Italy, Istituto di Ricerche Farmacologiche Mario Negri, Milan, Italy, and Consorzio Mario Negri Sud, Santa Maria Imbaro, Italy. SPA, Pfizer, Sigma Tau, and AstraZeneca contributed to fund the main study. AstraZeneca provided separate support for the microalbuminuria substudy. Drs. Masson, Maggioni, and Latini received institutional grants from AstraZeneca, SPA, Sigma Tau, and Pfizer. Dr. Maggioni received honoraria for lectures from AstraZeneca. Dr. Tavazzi served on the speakers bureaus for the companies that financed the GISSI-HF trial. Dr. van Veldhuisen received board membership fees from Vifor, Amgen, BG Medicine, Johnson & Johnson, and Sorbent. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- estimated glomerular filtration rate
- heart failure
- kidney injury molecule
- neutrophil gelatinase-associated lipocalin
- worsening renal function
- Received March 4, 2013.
- Revision received April 29, 2013.
- Accepted May 3, 2013.
- American College of Cardiology Foundation
- Hillege H.L.,
- Girbes A.R.,
- de Kam P.J.,
- et al.
- Damman K.,
- Masson S.,
- Hillege H.L.,
- et al.
- Vardeny O.,
- Wu D.H.,
- Desai A.,
- et al.
- Testani J.M.,
- Kimmel S.E.,
- Dries D.L.,
- Coca S.G.
- Damman K.,
- Jaarsma T.,
- Voors A.A.,
- Navis G.,
- Hillege H.L.,
- van Veldhuisen D.J.
- de Silva R.,
- Nikitin N.P.,
- Witte K.K.,
- et al.
- Rossignol P.,
- Cleland J.G.,
- Bhandari S.,
- et al.
- Damman K.,
- Ng Kam Chuen M.J.,
- MacFadyen R.J.,
- et al.
- Damman K.,
- van Veldhuisen D.J.,
- Navis G.,
- et al.
- Maisel A.S.,
- Mueller C.,
- Fitzgerald R.,
- et al.