Article Text

Download PDFPDF

Original article
Hepatocyte growth factor is a strong predictor of mortality in patients with advanced heart failure
  1. Kathrin Rychli1,
  2. Bernhard Richter1,
  3. Philipp J Hohensinner1,
  4. Kariem Mahdy Ali1,
  5. Stephanie Neuhold1,
  6. Gerlinde Zorn1,
  7. Rudolf Berger1,
  8. Deddo Mörtl1,
  9. Kurt Huber2,
  10. Richard Pacher1,
  11. Johann Wojta1,
  12. Alexander Niessner1,
  13. Martin Hülsmann1
  1. 1Division of Cardiology, Department of Internal Medicine II, Medical University of Vienna, Vienna, Austria
  2. 2Department of Cardiology and Emergency Medicine, Wilhelminen Hospital, Vienna, Austria
  1. Correspondence to Alexander Niessner, Department of Internal Medicine II, Division of Cardiology, Waehringer Guertel 18-20, 1090 Vienna, Austria; alexander.niessner{at}meduniwien.ac.at

Abstract

Objective To assess the prognostic value of the mitogenic, antiapoptotic, angiogenic and antifibrotic hepatocyte growth factor (HGF) in heart failure (HF).

Design Prospective cohort study.

Setting/patients Assessment of HGF levels at inclusion in 351 patients with advanced HF (median 75 years, interquartile range (IQR) 63–82, 66% male).

Main outcome measures All-cause mortality, cardiovascular mortality.

Results During a median follow-up of 16 months, 26% of patients died. HGF tertiles were associated with an increasing risk for all-cause mortality (p<0.001) with a hazard ratio (HR) of 3.06 (95% confidence interval (CI) 1.69 to 5.53) for the third compared with the first tertile. This association remained significant after multivariable adjustment for B-type natriuretic peptide (BNP) and other risk factors (p=0.002). Subgroup analysis revealed that HGF was a strong predictor of the secondary end point cardiovascular mortality in ischaemic HF (p=0.009) with an adjusted HR of 6.2 (95% CI 1.76 to 21.8) comparing the third with the first tertile but not in non-ischaemic HF (HR=1.47, 95% CI 0.48 to 4.49, p=0.5). Patients with high HGF but low BNP had a comparable survival rate to those with elevated BNP but low HGF (p=0.66). Of note, the dose of angiotensin converting enzyme (ACE) inhibitors inversely correlated with HGF concentrations (r=−0.25, p<0.001).

Conclusions HGF is a strong and independent predictor of mortality in advanced HF and, in particular, in ischaemic HF. These results indicate that HGF with its multiple effects on myocardial function exerts an overall deleterious effect in advanced HF. HGF may be of special interest for risk prediction and tailoring of HF treatment.

  • Hepatocyte growth factor
  • angiogenesis
  • heart failure
  • cardiomyopathy
  • prognosis
  • cardiac remodelling
  • left ventricular hypertrophy
  • aortic valve disease
  • gene expression

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Despite recent advances in its diagnosis and management, heart failure (HF) continues to be a major public health problem and is among the most significant causes of morbidity and mortality in older adults in developed countries.1 Risk stratification is of utmost importance to identify patients with HF with a high risk of future events and to prevent these events by appropriate therapeutic measures. B-type natriuretic peptide (BNP) and, to a lesser extent, some other prognostic parameters such as left ventricular ejection fraction (LVEF) and New York Heart Association (NYHA) classification, have clearly improved risk stratifications during the past decade.2 However, an improvement of risk prediction is still desirable. New prognostic biomarkers could also help to better understand the pathophysiology of HF and possibly point towards new therapeutic options. Cytokines may be a useful adjunctive treatment.3

Hepatocyte growth factor (HGF), initially recognised as a potent mitogen for hepatocytes, has subsequently been shown to have angiogenic, mitogenic, antiapoptotic and antifibrotic activities in various cell types.4–6 HGF and other angiogenic markers were found to be elevated in patients with congestive HF7 8 and during acute myocardial infarction.9 In diverse HF animal models strong expression of HGF resulted in increased angiogenesis and decreased apoptosis and fibrosis.4 6 These actions might primarily be responsible for the cardioprotective effects observed after administration or gene transfer of HGF.6 HGF treatment resulted in attenuation of chronic cardiac remodelling and dysfunction and improved cardiac perfusion and regional systolic function in diverse animal HF models.5 6 10

Though there is a growing body of clinical and experimental evidence, studies on the prognostic value of HGF in patients with HF are scarce and have assessed only stable, ambulatory patients with a rather low mortality rate.11 The need for data on HGF and cardiac repair in humans has been noted in a recently published review article by Beohar et al dealing with cytokines as a promising treatment option for ischaemic heart disease.3

The aim of this study was to evaluate the prognostic value of HGF in patients with advanced HF. Additionally, we compared its prognostic value between patients with an ischaemic and non-ischaemic aetiology of HF. Furthermore, we performed stratified analyses to evaluate the combined risk prediction by HGF and BNP levels.

Methods

Study population

The study population has previously been described in detail.12 In summary, patients with advanced systolic HF were from six cardiology departments in Vienna and included in this study according to the following criteria: (1) current hospitalisation due to clinical signs and symptoms of cardiac decompensation; (2) NYHA class III or IV at time of admission; and (3) cardiothoracic ratio >0.5, and/or LVEF <40%. Exclusion criteria were non-cardiac diseases with a life expectancy <1 year (eg, neoplasia) and refusal to sign informed consent. Ischaemic HF was defined as congestive HF due to significant coronary artery disease proven by coronary angiography (coronary stenosis >70%) and/or a history of prior myocardial infarction. No patient had an acute coronary syndrome leading to hospitalisation. All other causes of HF were classified as non-ischaemic HF. The study was approved by the hospital's ethics committee and written informed consent was given by the patients.

Follow-up visits were arranged 1, 3, 4, 6, 12 and 24 months after hospital discharge. The primary end point was all-cause mortality. The secondary end point was cardiovascular mortality. Mortality was confirmed by reviewing the death registry of the ‘Oesterreichisches Melderegister’. The cause of mortality was confirmed by postmortem examination in 43% of patients. A sample size of 350 patients with 25% of patients experiencing the primary end point all-cause mortality allowed detection of a RR of ≥1.7 (α=0.05, power=80%).

Blood sampling and laboratory analysis

Plasma HGF was analysed by ELISA (R&D Systems, Minneapolis, Minnesota, USA) from venous blood samples obtained in the morning of the day of discharge before intake of medication. At that time point patients were in a stable, already compensated condition. This was considered important as HGF levels have been shown to increase during acute exacerbation of HF followed by a decrease during recompensation.8 EDTA plasma samples were centrifuged (2800 rpm, 20 min) and stored at −80°C in multiple aliquots until analysed. Intra- and interassay coefficients for HGF were 4.1% and 5.4%. The minimum detectable dose was <0.04 ng/ml. In healthy subjects HGF levels are expected to range from 0.47 to 1.11 ng/ml (mean 0.79 ng/ml) according to the manufacturer. BNP was determined by a commercially available specific test (Viva, Bayer Health Care, Leverkusen, Germany).

Statistical methods

Continuous data are presented as median (IQR). Correlations between HGF and continuous variables were assessed using Spearman ρ correlation coefficient. Categorical data were analysed using a test for linear association (Mantel–Haenszel χ2 test). Kaplan–Meier curves (log-rank test), univariate and multivariable Cox proportional hazard regression models were used to evaluate the predictive value of tertiles of HGF for all-cause mortality and for cardiovascular mortality. The multivariable model encompassed demographics (age, sex) and known predictors of clinical end points in patients with HF (BNP, NYHA class, coronary artery disease, diabetes mellitus, estimated glomerular filtration rate, body mass index, atrial fibrillation). In an additional multivariable model we adjusted for novel biomarkers as listed in table 1. In a second step, we built separate Cox regression models for the subgroups of ischaemic HF and non-ischaemic HF. Receiver operating characteristic (ROC) analysis was used to characterise the predictive value of HGF. Interaction terms were used to assess interactions between HGF and the aetiology of HF. A p value ≤0.05 (two-sided) was considered statistically significant. SPSS 15.0 was used for statistical analysis (SPSS Inc).

Table 1

The effect of hepatocyte growth factor (HGF) on survival of patients with heart failure (HF) in Cox proportional hazards models

Results

Patients and baseline characteristics

A total of 441 out of the 462 screened patients met inclusion criteria of advanced HF. Among them, 360 patients consented to participate in the trial. HGF values were available in 351 patients, with a median concentration of 2.46 ng/ml (IQR 1.62–4.49 ng/ml). Two hundred and twenty-one of these patients (63%) were categorised as having ischaemic HF including 158 patients (45%) with prior myocardial infarction. HGF concentrations were slightly, but not significantly, lower in patients with ischaemic HF (2.39 ng/ml, IQR 1.57–4.06 ng/ml) than in the remaining 130 patients with non-ischaemic HF (2.87 ng/ml, IQR 1.76–5.5 ng/ml), p=0.066). The median BNP level in the entire cohort was 441 pg/ml (IQR 231–842 pg/ml). Table 2 depicts baseline characteristics stratified by tertiles of HGF. HGF was correlated with age (r=0.12, p=0.028) and BNP levels (r=0.14, p=0.009). Furthermore, HGF was negatively related to body mass index (r=−0.12, p=0.028). Additionally, the prevalence of atrial fibrillation increased across tertiles of HGF (p=0.028). Of note, treatment with renin–angiotensin–aldosterone-system (RAAS) inhibitors, including ACE inhibitors and angiotensin II receptor blockers, was associated with lower HGF levels (p=0.017). HGF concentrations were inversely correlated with the dose of RAAS inhibitors (r=−0.21, p<0.001) and ACE inhibitors (r=−0.25, p<0.001, table 2).

Table 2

Baseline characteristics

Follow-up and univariate survival analysis

During a median follow-up time of 16 months (IQR 12–19 months), 93 patients (26%) reached the primary end point all-cause mortality. Sixty-six patients (19%) died owing to cardiovascular diseases.

HGF significantly predicted all-cause mortality (for trend, p<0.001, table 1, figure 1A) with a HR of 3.07 (95% CI 1.69 to 5.59, p<0.001) for the second tertile and a HR of 3.06 (95% CI 1.69 to 5.53, p<0.001) for the third tertile compared with the first tertile. ROC analysis yielded an area under the curve (AUC) of 0.63 (p<0.001) for all-cause mortality.

Figure 1

Survival curves according to tertiles of hepatocyte growth factor (HGF) and combined strata of HGF and B-type natriuretic peptide (BNP). Kaplan–Meier plots showing the crude cumulative survival free from all-cause mortality (A) and cardiovascular mortality (B) according to tertiles of HGF as well as the crude cumulative survival free from all-cause mortality (C) and cardiovascular mortality (D) stratified for HGF and BNP. +/− BNP, above versus below the median of BNP (441 pg/ml); +/− HGF, above versus below the median of HGF (2.46 ng/ml).

Stratification for HF aetiology revealed that HGF significantly predicted all-cause mortality only in patients with ischaemic HF (for trend, p=0.001) with a HR of 4.27 (95% CI 1.95 to 9.35, p<0.001) for the second tertile and a HR of 4.35 (95% CI 1.96 to 9.65, p<0.001) for the third tertile compared with the first tertile (table 1). In patients with non-ischaemic HF, however, there was no significant association between HGF levels and all-cause mortality (for trend, p=0.45, table 1). Interaction term analysis indicated a trend towards a difference in the prognostic value of HGF between patients with ischaemic and non-ischaemic HF (p=0.091). With respect to patients with ischaemic HF, ROC analysis revealed an AUC of 0.67 (p<0.001) for all-cause mortality. In patients with non-ischaemic HF, ROC analysis did not yield significant results (p=0.32).

Additionally, HGF concentrations were positively associated with the secondary end point cardiovascular mortality (for trend, p=0.001, table 1, figure 1B) with a HR of 3.83 (95% CI 1.81 to 8.09, p<0.001) for the second tertile and a HR of 3.52 (95% CI 1.66 to 7.47, p=0.001) for the third tertile compared with the first tertile. A particularly strong association of HGF with cardiovascular mortality was found in the subgroup of ischaemic HF (for trend, p=0.004) with a HR of 6.03 (95% CI 2.07 to 17.6, p=0.001) for the second tertile and a HR of 5.49 (95% CI 1.83 to 16.4, p=0.002) for the third tertile compared with the first tertile. ROC analysis evaluating the predictive value of HGF for cardiovascular mortality resulted in an AUC of 0.62 (p=0.003) for the entire study cohort and an AUC of 0.65 (p=0.003) for the subgroup of ischaemic HF.

To determine a potential additive prognostic value of HGF and BNP, we assessed survival in the combined strata of HGF and BNP. Patients with HF with both markers below the median (figure 1C) had a survival of 90% at 12 months. A survival rate of 85% and 78% at 12 months was seen in patients with either HGF or BNP above the median (log-rank test: p=0.655 between strata with one marker elevated; p=0.052 between strata with high HGF and low BNP vs no elevated marker and p=0.022 for low HGF and high BNP compared with no elevated marker). Sixty-three per cent of patients with HF with HGF and BNP above the median survived at 12 months (p=0.002 compared with high HGF and low BNP; p=0.007 compared with low HGF and high BNP). Of note, patients with both biomarkers above the median experienced a 4.7-fold higher hazard of death (95% CI 2.4 to 9.2, p<0.001) than those with both markers below the median. Similar combined predictive values of HGF and BNP were found for the secondary end point cardiovascular mortality (figure 1D).

Multivariable analysis

HGF remained a strong predictor of all-cause mortality (for trend, p=0.002) and cardiovascular mortality (for trend, p=0.004) after adjustment for demographics and prognostic markers including BNP (table 1). This was particularly the case for the subgroup of ischaemic HF (table 1). The adjusted HR for cardiovascular mortality was 6.64 (95% CI 1.94 to 22.7, p=0.003) for the second tertile and 6.2 (95% CI 1.76 to 21.8, p=0.004) for the third tertile of HGF compared with the first tertile. Also after adjustment for a set of novel biomarkers, as listed in table 1, HGF remained a significant predictor of survival. Adjustment for treatment did not significantly change results (data not shown). Of note, BNP was also a significant independent predictor of all-cause mortality (for trend, p=0.003) and cardiovascular mortality (for trend, p=0.015).

Discussion

This study demonstrates that the risk of all-cause mortality increases with endogenous HGF concentrations in patients with advanced HF with a 3.1-fold higher risk in the third tertile compared with the first tertile. Interestingly, additional subgroup analysis stratifying by the aetiology of HF showed that the prognostic value of HGF was only present in patients with ischaemic HF and not in those with HF of other aetiology. In patients with ischaemic HF we observed a 4.4-fold higher risk in the third tertile compared with the first tertile. As depicted in figure 1A, the main increase of risk was between the first and the second tertile of HGF. Therefore, it might be speculated that a certain threshold of HGF has to be exceeded to initiate mechanisms linked with a poor survival.

Additional analysis evaluating the predictive potential of HGF for the secondary end point cardiovascular mortality yielded similar results as reported for all-cause mortality. In patients with ischaemic HF the adjusted hazard for a cardiovascular death was 6.2-fold higher in the third tertile of HGF compared with the first tertile. The predictive value of HGF was independent of BNP and other potential predictors of outcome in patients with HF. Stratified analyses evaluating the combined risk prediction by HGF and BNP levels showed that high HGF indicates a poor prognosis even in patients with low BNP. This subgroup of patients had a comparable risk to those with elevated BNP, but low HGF. As expected, the greatest risk was found when both factors were raised. This additive prognostic value of HGF might help to identify patients at high risk who would benefit from intensive treatment.

These findings extend data published by Lamblin et al,11 which showed a moderate, but significant association of HGF with HF severity and cardiovascular mortality (HR=1.85) in a cohort of stable, ambulatory patients with HF. In contrast to this previous study, we included only patients with a more advanced stage of HF characterised by the requirement for hospitalisation for cardiac decompensation and a thereby markedly higher cardiovascular mortality rate (1-year cardiovascular mortality rate 16% vs 9%). Consequently, BNP levels were higher in this study, with a median of 441 pg/ml compared with a median of 325 pg/ml (equivalent to 94 pmol/l) in the older study. Apart from these differences, the study patients of our study were older and had higher HGF levels. The difference in HGF levels might not be surprising since HGF levels showed an association with age and BNP levels. Despite the association of HGF levels with these two well-known prognostic variables in HF, HGF remained a strong and significant predictor of all-cause morality after adjustment for age, BNP and other established risk factors. Similar to other cytokines, high HGF concentrations were found to be associated with a low body mass index and may therefore contribute to cachexia in patients with HF.13 Additionally, HGF levels were higher in patients with atrial fibrillation. This finding is in line with a previous study reporting elevated HGF levels in patients with atrial fibrillation.14 Of note, we found reduced HGF levels in patients treated with RAAS inhibitors and an inverse correlation of HGF concentrations with the dose of RAAS inhibitors. Published data on the effect of RAAS inhibition on HGF concentrations are contradictory. An experimental study indicated that inhibition of angiotensin II in hamsters results in increased HGF concentrations.15 On the contrary, Yasuda et al found no changes in HGF levels after a 4-week treatment with ACE inhibitors in humans.16 In accordance with our findings, a recently published study showed that angiotensin II receptor blockers decrease HGF concentrations.17 In conclusion, the reduction of endogenous HGF concentrations may be another beneficial effect of RAAS inhibitors and may be helpful in titrating the dose of RAAS inhibitors.

The role of HGF in the pathophysiology of HF has not yet been elucidated. Several triggers of HGF activation have been described including various types of tissue injury, ischaemia, numerous proteases involved in coagulation and fibrinolysis, adrenergic stimulation, but also cytokines such as tumour necrosis factor α, interleukin 1 and HGF activator.6 18–20 Several of these factors also play a role in HF. HGF upregulation may occur in a variety of tissues including liver, lung, kidney, blood vessels and muscle.6

In relation to the potential cardiovascular effects of HGF, experimental data suggest the presence of angiogenic effects in diverse HF models, resulting in the development of functional arterioles and collateral artery growth after HGF gene transfer.6 10 Moreover, HGF is thought to directly affect cardiomyocytes via antiapoptotic and antifibrotic properties.6 These antifibrotic actions may in part be mediated by its effects on the production of transforming growth factor β and metalloproteinases.15 Blockage of HGF by specific antibodies resulted in an increase of myocyte cell death. Additionally, HGF is thought to be involved in the recruitment of stem cells into ischaemic myocardium21 and to suppress oxidative stress.22 These cardioprotective effects were also associated with improvement of functional parameters such as perfusion and left ventricular function.4 6 Additionally, HGF gene transfer has been shown to attenuate unfavourable cardiac remodelling.6

The association of HGF with an unfavourable clinical outcome observed in this and a previous study11 might, at first glance, contradict these experimental data. However, it should be kept in mind that HGF is a pleiotropic cytokine displaying both potential beneficial and detrimental effects. In this respect, it should be mentioned that HGF is also thought to worsen plaque formation and even destabilise atherosclerotic plaques.3 23 This could promote cardiovascular events. Accordingly, HGF was a particularly strong risk factor for cardiovascular death in patients with ischaemic HF. Moreover, one might speculate that progressive HGF upregulation in the course of HF development is an endogenous protective mechanism which might no longer be effective. In particular, myocardial scars in patients with ischaemic HF might not be susceptible to the beneficial proangiogenic actions of HGF. This could also explain why the association between HGF and worse survival was only found in patients with ischaemic HF. Similar to HGF, the natriuretic peptides are considered to exert cardioprotective effects, on the one hand, but are positively associated with HF severity and mortality, on the other.24–26

A limitation of the study is that assessment of left ventricular function differed between patients because six different cardiovascular centres were involved. At study entry, quantitative echocardiographic data for LVEF were only available in 158 patients. However, as there was no significant association between the echocardiographic LVEF and HGF (table 2), we do not believe that LVEF confounded the association between HGF and outcome. Furthermore, the importance of LVEF for the assessment of the severity of HF has declined owing to the use of BNP with a better predictive value for the clinical course of HF.

The definite exclusion of coronary artery disease in non-ischaemic patients with HF may be another limitation of this study. In patients without any clinical sign of coronary artery disease and low pre-test probability according to cardiac risk factors, angiography was not performed routinely. Therefore, the possibility that a small proportion of patients with clinically non-evident coronary artery disease were wrongly classified as non-ischaemic HF cannot be excluded.

Conclusion

Endogenous HGF was shown to be a strong and independent predictor of all-cause mortality in patients with advanced HF. This association was only present in patients with ischaemic HF and not in those with HF of other aetiology, suggesting discrete pathogenic pathways determining the course of disease. The distinctive strength of the observed association suggests that this pluripotent growth factor is of specific relevance in the complex pathogenesis and progression of ischaemic HF. The particularly strong predictive value of HGF for cardiovascular mortality further emphasises its potential role in atherosclerotic disease. Furthermore, HGF improved risk prediction beyond BNP and could therefore help to identify high-risk patients.

These data extend previous results showing that angiogenic molecules such as pigment epithelium-derived factor and vascular endothelial growth factor have a role in HF.7 27 This is of special interest, since cytokine treatment targeting angiogenesis, progenitor cells and/or cardiac repair is regarded as a promising non-invasive treatment option for cardiovascular disease.3 Moreover, endogenous HGF concentrations may also help tailor RAAS inhibitor therapy, as HGF levels may be modifiable by RAAS inhibitors and could therefore be useful for guiding HF treatment. These results together with the evidence on a myriad of other predictive biomarkers in HF support the notion that HF is a multifactorial disease. This is emphasised by the concept of metabolic, functional and haemodynamic staging proposed by Anker and Coates28 including different prognostic factors and also biomarkers to assess the prognosis of patients with HF. Compared with other novel biomarkers previously assessed in our centre and specified in table 1,12 27 29 30 HGF was the strongest predictor of HF outcome.

Future clinical studies are warranted to validate these results and interventional studies have to clarify whether therapeutic modulation of HGF levels would be helpful for patients with HF.

References

Footnotes

  • KR and BR contributed equally to this study.

  • Funding This work was supported by the Association for the Promotion of Research in Arteriosclerosis, Thrombosis and Vascular Biology (Vienna, Austria) and by the Ludwig Boltzmann Foundation for Cardiovascular Research (Vienna, Austria).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval This study was approved by the ethics committee of the Medical University of Vienna, Austria.

  • Provenance and peer review Not commissioned; externally peer reviewed.