The association between neutrophil gelatinase-associated lipocalin and clinical outcome in chronic heart failure: results from CORONA

Authors

  • S. H. Nymo,

    1. From the Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet
    2. Centre for Heart Failure Research
    Search for more papers by this author
  • T. Ueland,

    1. From the Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet
    2. Faculty of Medicine, University of Oslo
    Search for more papers by this author
  • E. T. Askevold,

    1. From the Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet
    2. Centre for Heart Failure Research
    3. Department of Cardiology, Oslo University Hospital Rikshospitalet, Oslo
    Search for more papers by this author
  • T. H. Flo,

    1. Department of Cancer Research and Molecular Medicine, Norwegian University of Technology and Science, Trondheim, Norway
    Search for more papers by this author
  • J. Kjekshus,

    1. Centre for Heart Failure Research
    2. Department of Cardiology, Oslo University Hospital Rikshospitalet, Oslo
    Search for more papers by this author
  • J. Hulthe,

    1. Wallenberg Laboratory for Cardiovascular Research, Sahlgrenska Academy, Gothenburg University, Gothenburg
    2. AstraZeneca, Mölndal, Sweden
    Search for more papers by this author
  • J. Wikstrand,

    1. Wallenberg Laboratory for Cardiovascular Research, Sahlgrenska Academy, Gothenburg University, Gothenburg
    Search for more papers by this author
  • J. McMurray,

    1. BHF Glasgow Cardiovascular Research Centre, University of Glasgow, UK
    Search for more papers by this author
  • D. J. Van Veldhuisen,

    1. Department of Cardiology, Thoraxcenter, University Medical Center Groningen, University of Groningen, The Netherlands
    Search for more papers by this author
  • L. Gullestad,

    1. Centre for Heart Failure Research
    2. Faculty of Medicine, University of Oslo
    3. Department of Cardiology, Oslo University Hospital Rikshospitalet, Oslo
    Search for more papers by this author
  • P. Aukrust,

    1. From the Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet
    2. Faculty of Medicine, University of Oslo
    3. Section of Clinical Immunology and Infectious Diseases, Oslo University Hospital Rikshospitalet, Oslo, Norway
    Search for more papers by this author
  • A. Yndestad

    1. From the Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet
    2. Centre for Heart Failure Research
    3. Faculty of Medicine, University of Oslo
    Search for more papers by this author

  • AstraZeneca, Mölndal, Sweden sponsored the CORONA trial.

Ståle H. Nymo, BSc, Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, 0424 Oslo, Norway.
(fax: +47-23073630; e-mail: snymo@rr-research.no).

Abstract

Abstract.  Nymo SH, Ueland T, Askevold ET, Flo TH, Kjekshus J, Hulthe J, Wikstrand J, McMurray J, Van Veldhuisen DJ, Gullestad L, Aukrust P, Yndestad A (Oslo University Hospital Rikshospitalet; University of Oslo; Oslo; Norwegian University of Technology and Science, Trondheim, Norway; Gothenburg University, Gothenburg; AstraZeneca, Mölndal, Sweden; University of Glasgow, UK; and University Medical Center Groningen, University of Groningen, The Netherlands). The association between neutrophil gelatinase-associated lipocalin and clinical outcome in chronic heart failure: results from CORONA. J Intern Med 2012; 271: 436–443.

Objective.  To study the prognostic value of neutrophil gelatinase-associated lipocalin (NGAL) in chronic heart failure (HF) of ischaemic aetiology.

Background.  Neutrophil gelatinase-associated lipocalin is a marker of kidney injury as well as matrix degradation and inflammation and has previously been shown to be increased in HF. We investigated whether serum NGAL levels could provide prognostic information in chronic HF.

Methods.  We assessed NGAL as a predictor of primary outcomes (cardiovascular death, nonfatal stroke and nonfatal myocardial infarction, = 307) and all-cause mortality (= 321), cardiovascular mortality (= 259) and hospitalization (= 647) as well as the number of hospitalizations during follow-up for all (= 1934) and CV causes (= 1204) in 1415 patients with chronic HF (≥60 years, New York Heart Association class II–IV, ischaemic systolic HF) in the CORONA population, randomly assigned to 10 mg rosuvastatin or placebo.

Results.  Multivariate analysis revealed that NGAL added significant information when adjusting for clinical variables, but was no longer significant when further adjusting for apolipoprotein A-1 (ApoA-1), glomerular filtration rate (GFR), C-reactive protein (CRP) and N-terminal pro-brain natriuretic peptide (NT-proBNP). However, belonging to the highest NGAL tertile was associated with more frequent hospitalization, even after adjusting for clinical variables, GFR and ApoA-1, but not after adjusting for CRP and NT-proBNP. There was no interaction between rosuvastatin treatment and NGAL.

Conclusion.  Neutrophil gelatinase-associated lipocalin added no significant information to NT-proBNP and GFR in a multivariate model for primary and secondary end-points.

Introduction

Chronic heart failure (HF) is a progressive disorder that, despite state of the art treatment, is characterized by high mortality and morbidity rates. Several factors have been implicated in the progression of HF, including persistent inflammation, immune activation and decreasing kidney function [1–3].

Neutrophil gelatinase-associated lipocalin (NGAL) is a 25 kDa protein covalently bound to matrix metalloproteinase-9 (MMP-9), which was first isolated from neutrophils [4]. Since then it has been demonstrated that NGAL is produced by a wide variety of cells including respiratory and intestinal epithelial cells [5–7], endothelial cells, renal tubular cells and cardiomyocytes [8, 9]. It has been demonstrated that NGAL has a role as an iron sequestering molecule in the immune system [10] and is up-regulated in several pathological conditions such as inflammatory bowel disease, atherosclerosis and chronic obstructive pulmonary disease (COPD) [8, 11, 12]. It has also been shown that NGAL levels are markedly increased in tubular epithelial cells during acute kidney injury, and the results of several studies suggest that NGAL is a promising early biomarker for this disorder [13]. Recently, NGAL levels have been found to be increased in the circulation and in urine in clinical HF [8, 14–16], and NGAL is up-regulated within the myocardium in experimental HF as well as in clinical and experimental myocarditis [8, 9].

Based on its relation to inflammation, matrix remodelling and kidney function, NGAL has been proposed as a marker of and mediator in chronic HF [2]. However, only a few relatively small studies have reported circulating NGAL levels in patients with HF [8, 17, 18]. To further elucidate the role of NGAL in chronic HF, we examined NGAL levels in a subpopulation of the CORONA study, consisting of elderly patients with chronic systolic HF of ischaemic cause receiving a standard pharmacological treatment regimen, randomly assigned to receive rosuvastatin or placebo in a double-blind fashion. Our aims were (i) to analyse the relationships between NGAL levels and clinical variables and established biomarkers of HF in a large population of patients with HF; and (ii) to assess the ability of NGAL to predict fatal and nonfatal outcomes in this population.

Methods

Patients

The design and principal findings of the CORONA study have been described previously by Kjekshus et al. [19]. Briefly, elderly patients (>60 years of age) with chronic HF of ischaemic cause, New York Heart Association (NYHA) class II–IV disease and left ventricular (LV) ejection fraction (LV-EF) ≤40% (≤35% for NYHA II) were eligible, as long as the investigator considered that they did not need treatment with a cholesterol-lowering drug. Criteria for exclusion included the following: recent cardiovascular (CV) events; current or planned procedures or surgery; acute or chronic liver disease or alanine aminotransferase more than two times the upper limit of normal (ULN); serum creatinine ≥2.5 mg dL−1; chronic muscle disease, contraindication to statin therapy or an unexplained creatine kinase level more than 2.5 ULN; thyroid stimulating hormone ≥2 ULN; and any condition that substantially reduces life expectancy.

Study procedures

The trial was approved by the ethics committees of the participating hospitals, and all patients provided written informed consent. From 15 September 2003 to 21 April 2005, 5011 patients were allocated, equally, to 10 mg rosuvastatin or matching placebo once daily. This study was a voluntary substudy of the main CORONA trial which included 1464 consecutive patients from centres that were capable of collecting the necessary blood samples. The substudy was designed to analyse plasma/serum levels of cytokines and other inflammatory mediators or markers. There were some modest but significant differences between the CORONA subgroup and the complete study population regarding LV-EF (0.01% higher in the subgroup, = 0.001), age (0.9 years younger in the subgroup, < 0.001) and NYHA class (lower percentage of NYHA class 2 [32% vs. 37%] and higher percentage of NYHA class 3 [67% vs. 62%] in the subgroup, = 0.001). Sex, body mass index (BMI) and estimated glomerular filtration rate (eGFR) were not different between the two groups.

Study outcomes and definitions

The primary predefined outcome was the composite of CV mortality, nonfatal myocardial infarction (MI) or nonfatal stroke, analysed as time to the first event. The secondary predefined outcomes were all-cause mortality, CV mortality, all-cause hospitalization and number of all-cause and CV hospitalizations. The definition and adjudication of all outcomes have been described in detail previously; high-sensitivity C-reactive protein (hsCRP) and N-terminal pro-brain natriuretic peptide (NT-proBNP) data have also been previously reported [19, 20].

Blood sampling and biochemical analyses

All blood samples were collected under nonfasting conditions and all measurements, except for NGAL, were made using fresh samples at a central laboratory (Medical Research Laboratories, Zaventem, Belgium). NT-proBNP was analysed using a commercially available assay (Roche Diagnostics, Basel, Switzerland) [19]. An immunonephelometric high-sensitivity method was used to measure C-reactive protein (Dade Behring BNII instrument with CardioPhase hsCRP reagent from Dade Behring, Newark, DE). eGFR was calculated according to the Modification of Diet in Renal Disease (MDRD) formula [21]. Blood samples for the measurement of NGAL were collected in pyrogen-free tubes without any additives, and serum was stored at −80 °C. All samples were thawed an equal number of times (two times). Although there were some differences in storage time (samples were collected between 2003 and 2005, and analysed in 2010), NGAL has been shown to be remarkably stable during storage at −80 °C, and levels are only influenced to a minor degree by multiple thawing cycles [22]. In this study, we used a noncommercial enzyme-linked immunosorbent assay (ELISA) that has previously been described [23], with a detection limit of 0.06 ng mL−1 and an intra- and interassay coefficient of variation of <6%. Although NGAL is found in complex with MMP-9, there was no cross-reactivity with MMP-9 in the current ELISA. NGAL levels measured using this ELISA showed a strong correlation with levels measured using a commercially available assay (r = 0.935; = 20; R&D Systems, Minneapolis, MN, USA).

Statistical methods

For all baseline variables, trends across NGAL tertiles were tested using the Cuzick extension of the Wilcoxon rank-sum test. All survival analyses were performed using the Cox proportional hazard regression model for different end-points. For the multivariate models, NGAL was included as a log-transformed continuous variable in a version of the three-stage model developed previously for the full CORONA population [20], which included the eight clinical variables most associated with outcome at step 1 [LV-EF, NYHA class (as a categorical variable), age, BMI, diabetes mellitus, sex, intermittent claudication and heart rate]; biochemical variables associated with outcome [eGFR and apolipoprotein A-1 (ApoA-1)] were included in the model at step 2, and finally, the known markers of outcome (i.e. levels of hsCRP and NT-proBNP) were included as log-transformed variables at step 3. All variables were included as continuous or dichotomized variables unless otherwise specified. Net reclassification improvement (NRI) and change in Harrell’s C-index were calculated for the addition of NGAL to all steps [24]. The influence of potential competition between some of the events was tested using the method of Fine and Gray [25]. The assumption of proportional hazards was tested using Schoenfeld residuals as well as visual inspection of a log–log plot. The total number of hospitalizations per 1000 days of follow-up in each tertile was investigated using the Kruskal–Wallis test for any difference between tertiles and Cuzick’s extension of the Wilcoxon rank-sum test for trend across increasing tertile. Because the hospitalization count was a Poisson distribution with overdispersion, further analysis was performed using a negative binomial regression model adjusting for the same variables as in the Cox regression model. A two-sided P-value <0.05 was considered to be significant in all cases except interaction terms, for which P-values <0.1 were accepted as significant. All statistical analyses were performed using STATA version 11 for Windows (Stata Corp LP, TX, USA).

Results

Clinical characteristics of the 1415 patients with NGAL measurements at baseline are shown in Table 1. Higher NGAL levels were associated with older age, lower eGFR and higher NT-proBNP and hsCRP levels. NGAL levels were associated with all measured lipid parameters with higher NGAL levels associated with lower LDL (= −0.05, = 0.04), HDL (= −0.08, = 0.001) and ApoA1 (= −0.08, = 0.002), and higher total triglycerides (r = 0.07, = 0.007). Patients with high NGAL levels were also more likely to be treated with aldosterone antagonists and digitalis glycosides. Although we have previously reported increased NGAL levels in patients with COPD [5], there was no significant trend of COPD prevalence across NGAL tertiles. There was a significant correlation between NGAL and NT-proBNP, hsCRP and eGFR (Fig. 1). When all variables with a P-value for trend ≤0.1 were included in a regression model with log-transformed NGAL as the dependent variable, only eGFR and hsCRP remained significantly associated with NGAL, explaining about 14% of the variation, and eGFR was the strongest predictor (∼10% of the variation).

Table 1. Association between clinical variables at enrolment and baseline NGAL levels
VariablesAllTertile 1Tertile 2Tertile 3P-value for trend across tertiles
  1. Categorical data are reported as n (percentages) and continuous data as mean (SD) except NT-proBNP, hsCRP and NGAL which are given as median (interquartile range). P-values below 0.1 in bold. Conversion factor for NT-proBNP: 1 pmol L−1 = 8.457 pg mL−1.

  2. ACE, angiotensin-converting enzyme; ApoA-I, apolipoprotein A-I; ARB, angiotensin receptor blocker; BMI, body mass index; CABG, coronary artery bypass grafting; COPD, chronic obstructive pulmonary disease; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; hsCRP, high-sensitivity C-reactive protein; LDL, low-density lipoprotein; MDRD, Modified Diet in Renal Disease; NT-proBNP, N-terminal pro-brain natriuretic peptide; NYHA, New York Heart Association; NGAL, neutrophil gelatinase-associated lipocalin; PCI, percutaneous coronary intervention.

Age, years71.8 (6.9)70.8 (6.8)72.1 (6.8)72.6 (6.9)<0.001
Female sex324 (23)119 (25)95 (20)110 (23)0.38
NYHA functional class, n (%)
 II461 (33)150 (32)179 (38)132 (28)0.19
 III938 (66)310 (66)287 (61)341 (71) 
 IV16 (1)8 (2)3 (1)5 (1) 
Ejection fraction0.31 (0.07)0.32 (0.07)0.31 (0.06)0.31 (0.07)0.10
BMI, kg m−227.1 (4.5)27.2 (4.2)27.1 (4.4)27.1 (4.7)0.37
Systolic blood pressure, mmHg129 (16)130 (16)130 (15.4)129 (17)0.55
Diastolic blood pressure, mmHg77 (9)77 (8)77 (9)76 (10)0.07
Heart rate, beats min−171 (11)71 (10)71 (11)71 (11)0.34
Medical history, n (%)
 Myocardial infarction888 (62)285 (61)299 (62)305 (64)0.35
 Angina pectoris1032 (73)352 (75)346 (74)334 (70)0.07
 CABG or PCI300 (21)80 (17)111 (24)109 (23)0.03
 Hypertension974 (69)330 (71)305 (65)339 (71)0.88
 Diabetes mellitus368 (26)123 (26)120 (26)125 (26)0.96
 Atrial fibrillation or flutter566 (40)186 (40)186 (40)194 (41)0.79
 Stroke170 (12)53 (11)54 (12)63 (12)0.38
 Intermittent claudication155 (11)46 (10)51 (11)58 (12)0.25
 COPD115 (8)39 (8)28 (6)48 (10)0.26
Laboratory measures
 Total cholesterol5.2 (1.1)5.3 (1.1)5.2 (1.1)5.1 (1.1)0.10
 LDL cholesterol3.6 (0.98)3.7 (1.1)3.6 (1.0)3.6 (1.0)0.034
 HDL cholesterol1.2 (0.3)1.3 (0.3)1.2 (0.3)1.2 (0.25)0.004
 APO A-1/APO B0.88 (0.25)0.87 (0.25)0.87 (0.25)0.90 (0.24)0.09
 Triglycerides, mmol L−12.0 (1.4)1.9 (1.3)1.9 (1.3)2.1 (1.6)0.023
 eGFRMDRD,mL min−1  1.73−2 BSA57.5 (14.3)62.5 (12.9)58.2 (13.8)51.9 (14.2)<0.001
 ApoA-1149.8 (27.8)152.2 (27.4)149.6 (26.8)147.7 (28.9)0.005
 NT-proBNP, pmol L−1162 (59–345)125 (51–307)152 (55–321)205 (91–442)<0.001
 hsCRP, mg L−13.7 (1.6–7.7)2.9 (1.3–5.6)3.3 (1.5–7.2)5.0 (2.3–10.3)<0.001
 NGAL (μg L−1)298 (213–437)183.5 (152–213)296 (268–335)510 (434–643)
Medication, n (%)
 Loop or thiazide diuretic1226 (87)396 (85)410 (87)420 (88)0.14
 Aldosterone antagonist515 (36)155 (33)164 (35)196 (41)0.01
 ACE inhibitor or ARB1133 (80)373 (80)367 (78)393 (82)0.32
 Beta-blocker1074 (76)354 (76)346 (74)374 (78)0.34
 Digitalis or digitoxin405 (29)116 (25)124 (26)165 (35)0.001
Figure 1.

 Scatter plots showing relationship between log-transformed NGAL and eGFR (n = 1415), log-transformed hsCRP (= 1401) and log-transformed NT-proBNP (n  =  1174).

NGAL levels and association with outcomes

Unadjusted Cox proportional hazard regression models using log-transformed NGAL levels showed a significant association between NGAL at baseline and the primary end-point of death because of CV cause, nonfatal stroke or nonfatal MI (= 307), as well as overall mortality (= 321), death as a result of CV cause (= 259) and hospitalization because of any cause (= 647) (Table 2). The hazard ratio varied from 1.29 for all-cause hospitalization to 1.36 for both overall mortality and CV mortality.

Table 2. Multivariate analysis: NGAL as an independent predictor of outcome
Primary end-point (= 307) HR (95% CI)P-value
  1. The models are adjusted as follows: Step 1: adjusted for ejection fraction, NYHA class, age, BMI, diabetes mellitus, sex, intermittent claudication and heart rate; step 2: All variables from step 1 as well as ApoA-1 and eGFR; step 3: All variables from step 2 as well as hsCRP and NT-proBNP. End-points as end-point (number of events) where number of events is the number of events amongst patients with data for all variables in step 3.

  2. NGAL, neutrophil gelatinase-associated lipocalin; NYHA, New York Heart Association; BMI, body mass index; ApoA-I, apolipoprotein A-I; eGFR, estimated glomerular filtration rate; hsCRP, high-sensitivity C-reactive protein; NT-proBNP, N-terminal pro-brain natriuretic peptide; HR, hazard ratio.

UnadjustedLog NGAL1.35 (1.11–1.63)0.002
Step 1Log NGAL1.24 (1.02–1.51)0.027
Step 2Log NGAL1.03 (0.84–1.27)0.781
Step 3Log NGAL1.06 (0.84–1.35)0.616
Mortality (= 321)
 UnadjustedLog NGAL1.36 (1.14–1.64)0.001
 Step 1Log NGAL1.23 (1.02–1.48)0.033
 Step 2Log NGAL1.03 (0.84–1.27)0.742
 Step 3Log NGAL1.01 (0.80–1.28)0.919
Cardiovascular death (= 259)
 UnadjustedLog NGAL1.36 (1.10–1.67)0.004
 Step 1Log NGAL1.24 (1.00–1.53)0.046
 Step 2Log NGAL1.00 (0.80–1.26)0.965
 Step 3Log NGAL1.04 (0.80–1.35)0.776
Hospitalization for any cause (= 647)
 UnadjustedLog NGAL1.29 (1.13–1.48)0.000
 Step 1Log NGAL1.20 (1.05–1.38)0.008
 Step 2Log NGAL1.07 (0.92–1.24)0.388
 Step 3Log NGAL1.04 (0.88–1.22)0.680

Multivariate analysis

After adjusting for demographic and clinical variables (step 1, see Statistical methods section), NGAL remained significant for the primary and secondary end-points (Table 2); however, the hazard ratio decreased for all end-points. After further adjustment for ApoA-1 and eGFR (step 2), and hsCRP and NT-proBNP (step3), NGAL no longer remained a significant predictor for any of the end-points (Table 2). There was no significant change in the Harrell’s C-index when including NGAL to the variables at steps 1–3, and the NRI was not significant (data not shown).

Frequency of hospitalization

In univariate analyses, NGAL levels showed a particularly strong association with hospitalization, and a higher level of NGAL at baseline was associated with more frequent hospitalization for each patient during follow-up (both for all causes [= 1934] and specifically CV causes [= 1204], Table 3). The association between NGAL levels in the top two tertiles and number of hospitalizations because of all or CV causes remained significant also when adjusting for the demographic variables used in step 1 of the Cox model, as well as when including eGFR and ApoA-1 from step 2, but not hsCRP and NT-proBNP (step 3) (Table 3).

Table 3. Average increase in frequency of hospitalization (per 1000 days of follow-up) because of all causes and cardiovascular (CV) causes with increasing NGAL tertile (1st and 2nd tertile combined vs. 3rd tertile) unadjusted and adjusted for clinical and biochemical variables
ModelVariableNumber of hospitalizations, regression coefficient (P-value)Number of hospitalizations because of CV cause, regression coefficient (P-value)
  1. Number of patients included in the analysis is given for each model type. The models are adjusted as follows: Step 1: adjusted for ejection fraction, NYHA class, age, BMI, diabetes mellitus, sex, intermittent claudication and heart rate; step 2: All variables from step 1 as well as ApoA-1 and eGFR; step 3: All variables from step 2 as well as hsCRP and NT-proBNP.

  2. NGAL, neutrophil gelatinase-associated lipocalin; NYHA, New York Heart Association, BMI, body mass index; ApoA-I, apolipoprotein A-I; eGFR, estimated glomerular filtration rate; hsCRP, high-sensitivity C-reactive protein; NT-proBNP, N-terminal pro-brain natriuretic peptide.

Unadjusted (n = 1415)Top two tertiles of NGAL0.42 (0.00)0.44 (0.00)
Step 1 (n = 1412)Top two tertiles of NGAL0.32 (0.00)0.31 (0.00)
Step 2 (n = 1397)Top two tertiles of NGAL0.21 (0.02)0.22 (0.05)
Step 3 (n = 1167)Top two tertiles of NGAL0.12 (0.23)0.18 (0.01)

Effect of rosuvastatin treatment on NGAL levels

There was no significant change in NGAL levels from baseline to follow-up at 3 months, with no significant difference between the rosuvastatin and placebo groups (data not shown). In contrast to hsCRP [26], an interaction between NGAL and treatment group was not significant at the 0.1 level for any of the end-points.

Discussion

It has previously been shown that patients with both chronic and acute HF have increased circulating levels of NGAL compared with controls [8]. In acute HF, an increased number of CV events were shown to occur in those with higher NGAL levels [8]. Our findings in the CORONA study, the largest study of circulating NGAL levels in chronic HF, extend these earlier results to a population of patients with chronic HF of ischaemic aetiology. We show that circulating levels of NGAL are predictive of total mortality and death from CV causes in this large population of patients with HF, also after adjustment for several clinical variables. However, these findings were no longer significant after adjusting for eGFR, ApoA-1, hsCRP and NT-proBNP. Similar results were recently reported in a smaller study of patients with HF, showing that systemic NGAL levels are largely determined by underlying impairment of renal rather than myocardial function [18]. Our data suggest that the clinical use of circulating NGAL levels in patients with chronic HF is limited.

There have been several reports of urinary NGAL as a predictor of chronic HF [14, 15]; the latest from the GISSI-HF study that is similar to the CORONA study [16]. In the GISSI-HF study, urinary NGAL was significantly associated with primary outcome also after adjustment for clinical and established risk factors, even in the presence of a normal eGFR. Although NT-proBNP was not included in the multivariate analyses and the GISSI-HF study included a more heterogeneous group of patients than the CORONA cohort (i.e. not restricted to elderly patients and those with HF of ischaemic aetiology), these findings suggest that urinary NGAL may be a more suitable biomarker than circulating NGAL levels in chronic HF.

There are some important differences between urinary and serum/plasma NGAL levels that might explain this discrepancy. Most importantly, NGAL mRNA is markedly up-regulated within the distal nephron during ischaemic or toxic renal injury, and the resulting synthesis and secretion of NGAL into the urine appears to comprise the major fraction of urinary NGAL, which is a sensitive marker of renal tubular injury [13, 27]. By contrast, although systemic NGAL levels increase during acute and chronic renal impairment, the direct contribution of the kidney itself to circulating NGAL has been questioned. However, kidney injury and other ischaemic and inflammatory processes cause an up-regulation of NGAL production in other tissues and cells, most probably representing the main source of circulating NGAL [13, 28]. Therefore, whereas urinary NGAL seems to be closely linked to kidney function, circulating NGAL levels may, to a larger extent, also mirror other conditions such as systemic and local inflammation. Indeed, in the present study, we found in the multivariate analyses that eGFR and hsCRP were the most important determinants of serum levels of NGAL, suggesting that both reduced kidney function and inflammation could be important sources of circulating NGAL during HF. The correlation between NGAL and hsCRP may reflect their combined relation to inflammation, further supporting the presence of chronic inflammation in HF. The potential superiority of urinary NGAL as compared to systemic NGAL levels as a biomarker in chronic HF could therefore reflect the fact that urinary NGAL is a more sensitive and direct marker of tubulo-interstitial injury, indicating renal damage even in the presence of normal glomerular filtration. It is interesting that in the GISSI-HF population, other sensitive markers of tubular damage (i.e. N-acetyl-beta-D-glucosaminidase and kidney injury molecule 1) showed comparable prognostic significance with urinary NGAL [16], further underscoring the importance of kidney impairment for prognosis of patients with HF.

It has previous have been suggested that NGAL could contribute to the progression of HF. We have previously reported that NGAL may be up-regulated in cardiomyocytes within the failing myocardium in response to inflammatory cytokines, including interleukin-1β and Toll-like receptor 2 and 4 agonists [8]. In myocarditis, the expression of both NGAL and its specific receptor (24p3R) in cardiomyocytes further indicates the presence of cardiac-specific NGAL expression in response to inflammation [9]. Moreover, it has been suggested that NGAL may promote matrix degradation and myocardial remodelling via its ability to stabilize MMP-9 activity. However, although the lack of a role as a biomarker in HF for a particular molecule does not exclude the possibility that it is a mediator of this disorder, and vice versa, the present study provides no additional evidence for the involvement of NGAL in the progression of HF. Further mechanistic studies are needed to clarify this issue.

In the present study, we examined a large cohort of patients with HF with a considerable number of events. There are, however, some important study limitations that need to be considered. First, there is a lack of data on albuminuria and proteinuria as alternative markers of impaired kidney function. Second, the study was performed in patients more than 60 years of age with systolic HF of ischaemic origin and therefore our findings might not apply to patients with preserved LV-EF or HF of nonischaemic aetiology, or those in other age groups. Third, as we studied several different end-points, there may be a problem with multiple testing, and therefore results close to the significance threshold should be interpreted with caution.

In conclusion, in this study of a large group of patients with HF, we have shown that circulating levels of NGAL are predictive of total mortality and death from CV causes, both before and after adjustment for several clinical variables. However, these findings were not significant after adjustment for additional variables such as hsCRP, Apo-A1, eGFR and NT-proBNP, indicating that biomarkers other than systemic NGAL levels may provide more prognostic information in this population.

Source of funding

This work was supported by grants from the Norwegian Council of Cardiovascular Research, the Research Council of Norway, the Odd Fellow Medical Research Fund, South-Eastern Norway Regional Health Authority and the Center for Heart Failure Research, University of Oslo, Norway.

Conflicts of interest

This study was supported by AstraZeneca. Drs Kjekshus, McMurray and Gullestad have received lecture fees from AstraZeneca, Drs Kjekshus and McMurray have received consulting or advisory board fees from AstraZeneca, Dr Wikstrand has received research grants from AstraZeneca, Dr Hulthe is an employee of AstraZeneca, Dr Wikstrand is a former Senior Medical Advisor to AstraZeneca and Dr Van Veldhuisen has received consultancy fees from Alere.

Ancillary