Hepatocyte Growth Factor and Incident Heart Failure Subtypes: The Multi-Ethnic Study of Atherosclerosis (MESA)

Abstract

Background:

Hepatocyte Growth Factor (HGF) is a cytokine and marker of cardiovascular disease (CVD) risk. Less is known about HGF and incident heart failure (HF). We examined the association of HGF with incident HF and its subtypes in a multi-ethnic cohort.

Methods:

We included 6,597 participants of the MESA cohort, free of clinical CVD and HF at baseline, with HGF measured at baseline. Incident hospitalized HF was assessed and adjudicated for HF with preserved ejection fracture (HFpEF) versus HF with reduced ejection fraction (HFrEF). Cox regression models estimated hazard ratios (HR) and 95% confidence intervals (CI) for HF risk by HGF levels, adjusted for socio-demographics, CVD risk factors and N-terminal pro-B-type natriuretic peptide.

Results:

Mean (SD) for age was 62 (10) yrs. Median (IQR) for HGF level was 950 pg/mL (758–1086); 53% were women. Over 14 (IQR, 11.5–14.7) years, there were 324 cases of HF (133 HFpEF and 157 HFrEF). For the highest HGF tertile compared to lowest, adjusted HRs were 1.59 (95% CI: 1.10, 2.31), 1.90 (1.03, 3.51), and 1.09 (0.65, 1.82) for overall HF, HFpEF and HFrEF, respectively. For continuous analysis per 1 SD log-transformed HGF, adjusted HRs were 1.22 (1.06, 1.41), 1.35 (1.09, 1.69), and 1.00 (0.81, 1.24) for HF, HFpEF, and HFrEF, respectively.

Conclusions:

HGF was independently associated with incident HF. HGF remained significantly associated with HFpEF but not HFrEF upon subtype assessment. Future studies should examine mechanisms underlying these associations and evaluate whether HGF can be used to improve HF risk prediction or direct therapy.

Graphical Abstract (Visual Take Home Graphic):

Included are hazard ratios (HR) for HGF and subsequent heart failure comparing the top tertile to bottom tertile for heart failure subtypes of heart failure with reduced ejection fraction (HFrEF) and heart failure with preserved ejection fraction (HFpEF).

Introduction

The global burden of heart failure (HF) is immense, encompassing an estimated prevalence of 37.7 million worldwide and expanding alongside an aging population.^{1, 2} HF has been recognized as not a single entity but a collection of subtypes with differing causative etiologies. Two subtypes of HF are dominant in the current clinical paradigm: HF with reduced ejection fraction (HFrEF), defined as an EF <50%, and HF with preserved ejection fraction (HFpEF), defined by the clinical symptoms of HF along with a “normal” EF of ≥50%.³ Though both lead to similar symptoms of HF, such as dyspnea, orthopnea, lower extremity swelling, and numerous others, HFrEF and HFpEF remain disparate disease processes with differing diagnostic and treatment considerations. While HFrEF and HFpEF have some risk factors in common, these risk factors are not necessarily overlapping.⁴ For example, obesity is a more significant risk factor for HFpEF than HFrEF,⁵ and other studies have found a stronger association of inflammation with HFpEF than with HFrEF.^6–9 As there are currently no therapies to effectively treat HFpEF, prevention of HFpEF from developing in the first place remains imperative.¹⁰

A number of studies have attempted to identify early patients at risk for HF, so that targeted lifestyle and other preventive therapies may be implemented. One important tool in doing so is the use of biomarkers to signal risk for and/or subclinical signs of disease prior to the development of clinical HF. Hepatocyte Growth Factor (HGF) is a mesenchymal cytokine important to the development of many endothelial and epithelial cells.¹¹ HGF is released in response to endothelial injury, and studies have suggested its elevation in serum acts as an important risk marker of cardiovascular disease (CVD).^12–14 Recent data have shown that increased levels of HGF are an independent predictor of incident coronary heart disease (CHD),¹⁵ stroke,¹⁴ and progression of atherosclerosis.^{16, 17} These associations may differ by race/ethnicity, with stronger associations of HGF and CHD in Blacks.¹⁵ Additional work has further linked higher HGF levels to increased mortality among patients with advanced HF.¹⁸ Less is known about the association of HGF with incident HF in a community population. Furthermore, while there are data exhibiting an association of elevated HGF with increasing severity of HFrEF, studies on HFpEF are much more limited.¹⁹ Whether there is heterogeneity in risk of HGF with incident HF by sex across demographic subgroup has not been well studied.

The aims of this study are 1) to evaluate the association of baseline HGF levels with incident HF in a multi-ethnic cohort free of CVD and HF at baseline, 2) to examine the association of baseline HGF with incident HFrEF versus HFpEF, and 3) to determine if the associations of HGF with incident HF and its subtypes differ by sex or race/ethnicity.

Methods

Transparency and Openness Policy

The Multi-Ethnic Study of Atherosclerosis (MESA) data are available through the National Heart, Lung, Blood Institute (NHLBI) Biologic Specimen and Data Repository Coordinating Center (BioLINCC). Requests for access to the data can be made through their website: https://biolincc.nhlbi.nih.gov/studies/mesa/. Through the methods described in our paper, our findings should be easily reproducible.

Study Population

Data from MESA, a multi-center prospective study of 6,814 participants aged 45–84 years old across six institutions and a wide range of demographics, were utilized for this study.²⁰ All participants were free of clinical CVD or HF at baseline (2000–2002), and were followed for up to 5 additional exams after the baseline visit (Exam 2: 2002–2004, Exam 3: 2004 −2005, Exam 4: 2005–2007, Exam 5: 2010–2011, and Exam 6: 2016–2018). Institutional review board agreement at each study site was required, and written consent was obtained for all participants.

In an ancillary study of the MESA cohort, circulating HGF protein was measured in participant blood samples. There were 6,597 participants of the MESA cohort who had HGF measured at initial examination (2000–2002) and were followed prospectively for incident HF with no missing covariates or HF status (Figure 1).

Figure 1: Heart Failure Incidence and Subtypes in Study Participants:

Participant flow chart through study. Missing HF status was classified as unknown HF status for these individuals secondary to loss to follow-up.

HGF Measurement

Peripheral blood samples were obtained from all participants, and stored at −70°C in serum separated by centrifugation within 30 minutes of sample collection.¹⁷ HGF protein levels were assessed via sandwich enzyme-linked immunosorbent assay (R&D Systems, Minneapolis, MN).¹⁶ At mean concentrations of 687, 2039, and 4080 pg/mL, the interassay coefficients of variations were 12%, 8%, and 7.4%, respectively, for manufacturer controls. Pooled serum controls exhibited a mean concentration of 688 pg/mL and a coefficient of variation of 10.4%.¹⁷

Event Adjudication

Every 9–12 months following enrollment, MESA participants or their next of kin were contacted by telephone to ask about interim hospitalizations, and medical records for those hospitalizations were obtained. As previously described, adjudication for hospitalized HF events was conducted by a physician review committee using standardized criteria.^{5, 9, 21, 22} Hospitalized HF was classified by reviewer as ‘absent,’ ‘probable,’ or ‘definite’ based upon data and test results available, and we included probable and definite hospitalized HF events in this analysis. A classification of ‘probable’ HF was ascertained by a diagnosis of HF alongside supporting symptoms and medical treatment in accordance with an HF diagnosis. Classification of ‘definite’ HF required confirmed hospitalization with heart failure and at least one further objective measures via diagnostic tests such as by chest x-ray, echocardiography, or ventriculography.²¹

Participants were followed from the baseline visit until the development of an incident HF event, death, drop-out, or until December 31, 2015. HF events were classified by EF reported in the medical records at the time of hospitalization, and for this analysis was defined as HFpEF if EF ≥50% (n=133) and HFrEF if EF <50% (n=157). Adequate information was not available in medical records to determine subtype for 34 cases (Figure 1). Furthermore, there were not enough HF events in the mid-range of EF (40–49%) to consider three categories of HF subtypes.

Covariate Assessment

Participants underwent standardized questionnaires, physical exam, and laboratory testing at the baseline examination. For this analysis we considered demographic factors (age, sex, race/ethnicity, MESA site, health insurance status, and education), lifestyle risk factors (smoking status, pack-years smoking, alcohol use, body mass index (BMI), and physical activity), and cardiovascular risk factors (systolic blood pressure, total cholesterol, HDL-cholesterol, and diabetes status), kidney function (eGFR), and modifying agents (anti-hypertensive and lipid-lowering medications)). We further included N-terminal pro-B-type natriuretic peptide (NT-pro-BNP) in an additional model to determine if the association of HGF and incident HF is independent of NT-pro-BNP.

Medication use was captured by an inventory. Education was classified into 2 categories as no bachelor’s degree, and bachelor’s degree or higher. Health insurance was classified as a binary yes/no based on insurance status. Physical activity was quantified as the total amount of moderate and vigorous physical activity in metabolic equivalent minutes per week obtained from a typical week Physical Activity questionnaire.²³ Height and weight were measured on exam in standardized fashion, and BMI was calculated as the weight divided by the height squared (kg/m²). Systolic and diastolic blood pressure were measured while participants were seated using a Dinamap automated device, and the 2^nd and 3rd measurements were averaged. EGR was calculated by the chronic kidney disease epidemiology (CKD-Epi) collaboration formula.²⁴ Diabetes was classified as present if the fasting blood glucose level was ≥126 mg/dL, if there was a self-reported diagnosis of diabetes, or use of diabetes medications. NT-proBNP was measured by an Elecsys proBNP immunoassay (Roche Diagnostics Corporation, Indianapolis, IN).²⁵

Statistical Analysis

For our primary analysis, HGF was divided into tertiles of its distribution, in keeping with prior studies analyzing associations of HGF and cardiac disease and/or heart failure.^15,18 Baseline characteristics of the study population were described for each tertile of HGF. Unadjusted incident rates of HF were calculated per 1000-person years. Unadjusted Kaplan-Meier curves by HGF tertiles for each HF outcome were generated.

We tested and confirmed that the proportional hazards assumption was not violated using Schoenfeld residuals for the association of HGF and all three outcomes. We then used multivariable-adjusted Cox regression models to generate hazard ratios (HR) and 95% confidence intervals (CI) for the association of HGF with incident hospitalized HF and its subtypes (HFpEF and HFrEF) as separate outcomes. HGF was modeled both as a categorical variable using tertiles of HGF, with the lowest tertile as a reference group, and as a continuous variable after natural log transformation and compared per 1 standard deviation (SD) increment in ln(HGF). In additional continuous analysis, we used restricted cubic splines for a more flexible distribution to assess association and evaluate any non-linearity or threshold effects.

For all analyses, we used progressively adjusted models to account for potential confounding factors as follows: Model 1 adjusted for age (continuous), sex, race/ethnicity, and MESA site. Model 2 adjusted for Model 1 + education (2 categories), insurance (2 categories), BMI (continuous), smoking status (never, former, current), pack years of smoking, alcohol use (never, former, current), and physical activity (continuous, MET-min/week) categories. Model 3 adjusted for Model 2 + systolic blood pressure (continuous), anti-hypertensive medication (yes/no), total cholesterol (continuous), HDL-cholesterol (continuous), lipid-lowing medication (yes/no), diabetes mellitus (yes/no) and estimated glomerular filtrate rate (continuous). Model 4 adjusted for Model 3 + NT-proBNP (continuous).

We examined for multiplicative interaction between HGF and age (dichotomized at 65 years), sex, BMI, and race/ethnicity in association with incident HF, HFrEF and HFpEF. If an interaction with a subgroup was found to be present, we planned to stratify analysis by subgroup; however, no significant interactions were found. In a sensitivity analysis, we also adjusted for incident CHD events to determine if the association of HGF and incident HF is independent of incident CHD.

We used STATA version 15.0 (StataCorp LP, College Station, TX) for the analysis. P values were two-sided, with statistically significant values considered at p<0.05.

Results

Participant Characteristics

Among the 6597 participants, the mean (SD) for age was 62 (10) yrs. Median (IQR) for HGF level was 950 pg/mL (758–1086); 53% were women. Per self-reported race/ethnicity groups, 39% of participants were White, 27% Black, 22% Hispanic, and 12% Asian American adults. Participant characteristics described by HGF tertile are shown in Table 1. Participants in higher tertiles of HGF were more likely to be prescribed anti-hypertensive medications (27% in tertile 1 versus 47% in tertile 3), and average systolic blood pressure was higher across increasing HGF tertiles (median 122 mmHg, 127 mmHg, and 131 mmHg for tertiles 1, 2, and 3, respectively). Participants in the lower tertiles were also more likely have obtained higher education (47% with ≥bachelor’s degree in tertile 1 versus 25% in tertile 3). The highest proportion of Asian participants were in tertiles 1 (18% versus 6% of tertile 3), while the highest proportion of Hispanic participants were in tertile 3 (12% of tertile 1 versus 32% of tertile 3). All other racial/ethnic participants were relatively evenly distributed across tertiles.

Table 1:

Participant Demographics and Risk Factors by Hepatocyte Growth Factor Tertiles: The Multi-Ethnic Study of Atherosclerosis (2000–2002)

	Total	Tertile 1 [701 (624,757)] ‡	Tertile 2 [905 (857,958)] ‡	Tertile 3 [1171 (1086,1324)] ‡	p-value
N	6,597	2,199	2,199	2,199
† Age, years	62 (10)	59 (9)	62 (10)	65 (10)	<0.001
Male	3,116 (47%)	1,129 (51%)	997 (45%)	990 (45%)	<0.001
Race/ethnicity
White	2,541 (39%)	954 (43%)	802 (37)	785 (36%)	<0.001
Asian American	794 (12%)	400 (18%)	258 (12%)	136 (6%)
Black	1,802 (27%)	586 (27%)	638 (29%)	578 (26%)
Hispanic	1,460 (22%)	259 (12%)	501 (23%)	700 (32%)
Education
≥ bachelor’s degree	2,351 (36%)	1,028 (47%)	779 (35%)	544 (25%)	<0.001
< bachelor’s degree	4,246 (64%)	1,171 (53%)	1,420 (65%)	1,655 (75%)
† BMI, kg/m2	28 (5)	27 (5)	28 (5)	30 (6)	<0.001
Smoking status
Current smoker	848 (13%)	203 (9%)	261 (12%)	384 (17%)	<0.001
Former smoker	2,397 (36%)	805 (37%)	768 (35%)	824 (38%)
Never smoker	3,352 (51%)	1,191 (54%)	1,170 (53%)	991 (45%)
Alcohol use
Never	1,365 (21%)	433 (20%)	466 (21%)	466 (21%)	<0.001
Former	1,572 (24%)	462 (21%)	505 (23%)	605 (28%)
Current	3,660 (55%)	1,304 (59%)	1,228 (56%)	1,128 (51%)
‡ Pack-years of smoking if >0	16 (6–33)	14 (5–28)	17 (6–31)	19 (7–38)	<0.001
‡ Physical activity, MET-minutes/week	4020 (1973–7513)	4500 (2280–8190)	4073 (1980–7455)	3515 (1583–6900)	<0.001
† Systolic blood pressure, mmHg	127 (22)	122 (20)	127 (21)	131 (22)	<0.001
† eGFR, ml/min per 1.73m2	78 (16)	80 (15)	78 (15)	75 (18)	<0.001
† Total cholesterol, mg/dL	194 (36)	194 (35)	196 (36)	192 (36)	<0.001
† HDL-C, mg/dL	51 (15)	53 (16)	51 (15)	49 (14)	<0.001
Diabetes mellitus	821 (12%)	138 (6%)	254 (12%)	429 (19%)	<0.001
Antihypertensive medication	2,452 (37%)	602 (27%)	825 (37%)	1,025 (47%)	<0.001
Lipid-lowering medication	1,081 (16%)	303 (14%)	362 (16%)	416 (19%)	<0.001
‡§ NT-proBNP, pg/mL	55 (24–113)	46 (21–90)	54 (23–108)	67 (31–147)	<0.001

^†Data are presented as mean (standard deviation) for continuous variables and as count (percentages) for categorical variables, unless otherwise specified

^‡Data are presented as median (IQR)

^§Sample size for NT-proBNP = 5,446

Over a median follow-up over 14 (IQR, 11.5–14.7) years, there were 324 cases of incident HF. After adjudication, 133 were designated as HFpEF and 157 HFrEF (34 could not be classified as HFrEF versus HFpEF). Figure 2 shows the unadjusted Kaplan-Meier curves for incident HF (Panel A), HFpEF (B), and HFrEF (C) by tertiles of HGF, with P<0.001 for all three Kaplan-Meier curves).

Figure 2: Kaplan-Meier Curves for HGF Tertiles by HF Subtypes:

Kaplan-Meier survival estimates for unadjusted association of Hepatocyte growth factor (HGF) with Heart failure (HF), preserved ejection fraction HF (HFpEF) and reduced ejection fraction HF (HFrEF). Log-rank test for all 3 Kaplan-Meier curves: P <0.001.

Table 2 shows the adjusted associations of HGF with all incident hospitalized HF and its subtypes. In a model adjusted for demographics-only (model 1), hazard ratios (HR) of the highest HGF tertile compared to the lowest was associated with increased risk of overall incident HF (2.29 (95% CI: (1.67, 3.13)), incident HFpEF (2.52 (1.52, 4.19)), and incident HFrEF 1.74 (1.13, 2.67)). After further adjustment for CVD risk factors (model 3) and NT-proBNP, HGF remained statistically significantly associated with any HF and with HFpEF, but the association was attenuated for HFrEF. For the highest tertile of HGF compared to the lowest, the HR of overall HF was 1.59 (1.10, 2.31) after full multivariable adjustment (model 4), and 1.90 (1.03, 3.51) and 1.09 (0.65, 1.82) for HFpEF and HFrEF subtypes, respectively. For continuous analysis, the fully adjusted HR was 1.22 (1.06, 1.41), 1.35 (1.09, 1.69), and 1.00 (0.81, 1.24) per 1 SD increment in HGF (0.3 pg/mL) for overall incident HF, incident HFpEF, and incident HFrEF, respectively. In sensitivity analyses, further adjustment for incident CHD did not attenuate the significant associations between higher HGF levels and incident HF and HFpEF risk. Figure 3 of the restricted cubic spline of the HRs adjusted for model 4 variables shows that the association of HGF with incident HF (Panel A), HFpEF (B), and HFrEF (C) appears approximately linear.

Table 2:

Incidence rates (95% CI) and Hazard ratios (95% CI) of HF (n=324) by HGF: N= 6,597 The Multi-Ethnic Study of Atherosclerosis (2000 to 2015)

HGF in pg/mL [Median (IQR)]	Tertile 1 [701 (624,757)]	Tertile 2 [905 (857,958)]	Tertile 3 [1171 (1086,1324)]	*Per 1 SD ln(HGF) (0.3 pg/mL) increment
A. Incidence rates (95% CI) and Hazard ratios (95% CI) of any HF (n=324) by HGF: N=6,597
N	2,199	2,199	2,199	6,597
Cases	60	96	168	324
Incidence rate	2.1 (1.6, 2.6)	3.5 (2.9, 4.3)	6.8 (5.9, 7.9)	4.0 (3.6, 4.5)
Hazard Ratio
Model 1	1 (reference)	1.40 (1.01, 1.94)	2.29 (1.67, 3.13)	1.47 (1.32, 1.63)
Model 2	1 (reference)	1.27 (0.92, 1.77)	1.84 (1.33, 2.53)	1.36 (1.21, 1.53)
Model 3	1 (reference)	1.18 (0.85, 1.64)	1.58 (1.14, 2.19)	1.29 (1.14, 1.45)
Model 4	1 (reference)	1.20 (0.82, 1.75)	1.59 (1.10, 2.31)	1.22 (1.06, 1.41)
B. Incidence rates (95% CI) and Hazard ratios (95% CI) of HFpEF (n=133) by HGF: N=6,597
N	2,199	2,199	2,199	6,597
Cases	22	41	70	133
Incidence rate	0.8 (0.5, 1.2)	1.5 (1.1, 2.0)	2.8 (2.2, 3.6)	1.6 (1.4, 2.0)
Hazard Ratio
Model 1	1 (reference)	1.57 (0.93, 2.66)	2.52 (1.52, 4.19)	1.49 (1.26, 1.77)
Model 2	1 (reference)	1.39 (0.82, 2.36)	1.91 (1.14, 3.22)	1.36 (1.12, 1.65)
Model 3	1 (reference)	1.33 (0.78, 2.26)	1.76 (1.04, 2.97)	1.30 (1.07, 1.59)
Model 4	1 (reference)	1.46 (0.79, 2.70)	1.90 (1.03, 3.51)	1.35 (1.09, 1.69)
C. Incidence rates (95% CI) and Hazard ratios (95% CI) of HFrEF (n=157) by HGF: N=6,597
N	2,199	2,199	2,199	6,597
Cases	35	48	74	157
Incidence rate	1.2 (0.9, 1.7)	1.8 (1.3, 2.3)	3.0 (2.4, 3.8)	1.94 (1.7, 2.3)
Hazard Ratio
Model 1	1 (reference)	1.23 (0.79, 1.91)	1.74 (1.13, 2.67)	1.34 (1.14, 1.57)
Model 2	1 (reference)	1.15 (0.74, 1.79)	1.48 (0.95, 2.30)	1.26 (1.06, 1.49)
Model 3	1 (reference)	1.05 (0.67, 1.64)	1.22 (0.78, 1.90)	1.17 (0.98, 1.40)
Model 4	1 (reference)	0.97 (0.59, 1.62)	1.09 (0.65, 1.82)	1.00 (0.81, 1.24)

Abbreviations: CI, confidence interval; HF, heart failure; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; HGF, hepatocyte growth factor

^*HGF was natural log transformed to calculate 1 SD increment

Statistically significant results at P < 0.05 are in bold font Incidence rates reported are per 1000 person-years

Hazard ratios are adjusted as follows:

Model 1: adjusted for age, sex, race/ethnicity and MESA study site

Model 2: model 1 + education, body mass index, smoking, pack years of smoking, alcohol use, physical activity and health insurance

Model 3: model 2 + systolic blood pressure, anti-hypertensive medication, total cholesterol, HDL-cholesterol, lipid-lowing medication, diabetes mellitus and estimated glomerular filtrate rate

Model 4: model 3 + NT-proBNP Using HGF tertiles, P = 0.58, 0.09 and 0.18 for interaction by sex for HF, HFpEF and HFrEF, respectively; P= 0.51, 0.37 and 0.78 for interaction by age for HF, HFpEF and HFrEF, respectively; P = 0.91, 0.55 and 0.90 for interaction by race/ethnicity for HF, HFpEF and HFrEF, respectively

Figure 3: Adjusted restricted cubic splines of the association of HGF levels (continuous) with outcomes of any HF (panel A), HFpEF (panel B), HFrEF (panel C).

Model is adjusted for age, sex, race/ethnicity, MESA site, education, insurance, body mass index, smoking status, pack years of smoking, physical activity, systolic blood pressure, anti-hypertensive medication use, total cholesterol, HDL-cholesterol, lipid-lowing medication use, diabetes mellitus, estimated glomerular filtrate rate, and NT-ProBNP.

There was no significant effect modification found between HGF and demographic groups defined by age, sex, or race/ethnicity for the risk of HF. Using HGF tertiles, P values were 0.58, 0.09 and 0.18 for interaction by sex for HF, HFpEF and HFrEF, respectively; P=0.51, 0.37 and 0.78 for interaction by age for HF, HFpEF and HFrEF, respectively; and P=0.91, 0.55 and 0.90 for interaction by race/ethnicity for HF, HFpEF and HFrEF, respectively.

Discussion

In this multi-ethnic cohort free of CVD and HF at baseline, we found that the mesenchymal cytokine HGF is independently associated with incident HF after adjustment for other traditional CVD risk factors and NT-proBNP. HGF remained significantly associated with HFpEF but was attenuated for HFrEF upon HF subtype assessment. These associations did not differ by age, sex, or race/ethnicity for any HF subtype, despite potential differences in baseline serum HGF values by race/ethnicity.

The costs associated with HF diagnoses and treatment, already over $20 billion as of 2012, are expected to increase to over $53 billion in the United States alone by 2030.¹ Patients with HFpEF have an older median age and have a higher likelihood of obesity, while those with HFrEF have a younger median age with a greater prevalence of previous myocardial infarction, contributing to ischemic cardiomyopathy.^{26, 27} While the incidence of HFrEF appears to be declining, the incidence of HFpEF is rising, leading to an overall significant increase in total HF prevalence.² Despite this, an understanding of associated and casual mechanisms of HFpEF remain limited, with downstream implications for limited abilities in monitoring and treatment.²⁸

The study here shows that the highest tertile of HGF was significantly associated with a 59% increased risk for incident HF and 90% increased risk of HFpEF, independent of cardiovascular and metabolic risk factors. To our knowledge, this is the first study to demonstrate an association of HGF with HFpEF. While previous studies have linked higher HGF levels to increased mortality in HF and to severity of HFrEF in particular, this is both the first study to associate HGF with incident HF and the largest such study of HGF in HF.

From a pathophysiologic standpoint, it appears the biological function of HGF may be cardio-protective. HGF is believed to have many potentially favorable mechanisms, such as being anti-inflammatory, anti-apoptotic, anti-fibrotic, and pro-angiogenic.²⁹ HGF is released in the circulation in response to endothelial damage, and is linked to less adverse left ventricular (LV) remodeling in HF mouse models following an ischemic insult.^{15, 30} HGF is significantly higher in patients following ST Elevation Myocardial Infarction (STEMI), likely in response to ischemic injury, while simultaneously being strongly negatively associated with IL-6 levels.³¹ Further, HGF has been shown to have a stimulatory effect on CD34+ hematopoietic progenitor cells, and may play an important role in endothelial and myocardial regeneration.^{32, 33}

Despite its favorable role in responding to vascular injury, in the general population, elevated circulating levels of HGF levels are associated with increased CVD risk, rather than lower risk, suggesting that elevated serum levels may reflect compensatory mechanisms that ultimately have failed in disease states such as HF.¹⁶ This is substantiated by work exhibiting higher HGF levels in patients with double and triple-vessel coronary artery disease, in comparison to patients with single-artery coronary disease, suggesting it may be a marker of more severe CVD states.³⁴ A prior study which compared mortality in HF patients found a relationship between HGF levels and mortality in ischemic HF, but not in non-ischemic HF, suggesting disparate triggers for its release into the circulation.¹⁸ Interestingly, in mouse models, HGF-treated mice exhibited reduced infarct size and fibrosis following MI as compared to untreated mice, but also exhibited thicker LV walls at the site of infarct.³⁵ This is further evidence to suggest HGF influences LV remodeling,³⁶ and may underlie the significant association of HGF with HFpEF exhibited in the present study. While HGF may indeed be elevated in response to ischemic stimuli and provide regenerative benefit, long-term HGF stimulation may be deleterious, ultimately leading to HFpEF and its subsequent complications. Additional studies evaluating HGF in association with LV structure and function is required to further understand if this proposed mechanism has merit.³⁶

As HFpEF and HFrEF have distinct etiologies and known differential associations with inflammation,^6–8 we compared the association of this novel biomarker HGF with different HF subtypes with the hopes it would yield valuable information that may lead to the discovery of a better prediction marker for HF and a potential therapeutic target. It is notable that in prior work by Rychli et al exhibiting a significant association between serum HGF and mortality in advanced HF, median HGF levels were significantly higher (2460pg/mL, IQR 1620–4610pg/mL) than those in our study.¹⁸ This may be a function of sample size or varying degrees of illness in the patient populations of different studies – our patient population studying incident HF was free of HF at baseline, compared to a cohort of patients with already known advanced HF. Future studies may help to further clarify this question and add to the clinical utility of HGF.

Despite a rapidly developing field, treatment options for HFpEF remain limited.³⁷ Recent data suggest there might be a role for therapies such as sodium-glucose transport protein 2 (SGLT2) inhibitors in reversing LV remodeling, which may therefore create a treatment avenue for modulating adverse adaptations seen in HFpEF (though the benefits of SGLT2 inhibitors specifically are currently seen most clearly in HFrEF).³⁸ HGF may help us capture changes in cardiac remodeling related to therapeutic interventions in HF, or act as a marker of incident HF risk.

Strengths and Limitations

Our study was intended to be exploratory, to evaluate potential mechanisms linked to HF risk using data from a well-characterized community-based cohort where we could take into account numerous potential confounding factors. Nevertheless, this study does have limitations. As a single observational study, causation cannot be derived from any association of HGF with subsequent HF or its subtypes; residual confounding may explain associations seen. There may be cardiac, metabolic, or alternative factors not included here that act as a driver of both HGF elevation and HF risk. While participants were followed for a median of nearly 15 years, the rate of incident HF is still quite limited among these generally healthy individuals free of CVD at baseline. Furthermore, larger cohorts may more clearly elucidate the significance of these findings. Interaction analyses may have been underpowered to exclude clinically meaningful differences. Finally, for this analysis, we only used a single measure of HGF measured at the baseline exam. Further work is needed with large sample sizes to evaluate the prognostic significance of longitudinal change in HGF levels with HF outcomes.

Conclusion

In sum, this study found that the cytokine HGF, previously linked to atherosclerotic CVD risk, is also independently associated with incident HF after adjustment for other HF risk factors, and is significantly associated with HFpEF but not HFrEF upon subtype assessment. Future studies should examine mechanisms underlying these associations and evaluate whether HGF has utility as a biomarker to improve HF risk prediction or to direct therapy.

Highlights.

• Hepatocyte growth factor (HGF) is an important cytokine in the inflammatory milieu of cardiovascular disease.

• Elevated serum HGF levels are associated incident heart failure.

• On heart failure subtype assessment, HGF was significantly associated with heart failure with preserved ejection fraction and not heart failure with reduced ejection fraction.

• Future studies may examine mechanisms underlying these associations and evaluate whether HGF has utility as a biomarker to improve heart failure risk prediction or to direct therapy.

Abbreviations

BMI

Body Mass Index

CI

Confidence Interval

CHD

Coronary Heart Disease

CVD

Cardiovascular Disease

HGF

Hepatocyte Growth Factor

HF

Heart Failure

HFpEF

Heart Failure with Preserved Ejection Fraction

HFrEF

Heart Failure with Reduced Ejection Fraction

HR

Hazard Ratio

MESA

Multi-Ethnic Study of Atherosclerosis

NT-pro-BNP

N-terminal pro-B-type natriuretic peptide

SD

Standard Deviation

SGLT2

Sodium-glucose Transport Protein 2