Article ID: CR-25-0067
Background: Systemic and cardiac metabolic disorders play a key role in patients with heart failure (HF). Fibroblast growth factor 21 (FGF21) is mainly secreted from the liver and has various effects on cardiomyocytes, including protection against oxidative stress, cardiac hypertrophy, and inflammation. However, the pathophysiologic and prognostic impact of FGF21 remains unknown.
Methods and Results: Serum levels of FGF21 and echocardiography were performed in patients with compensated HF (n=162) and control patients without HF (n=20). Compared with the control patients, those with HF displayed higher FGF21 levels (100 [76–213] vs. 237 [135–575] pg/mL; P=0.0006). There were no or modest correlations of FGF21 levels with clinical variables and echocardiographic parameters. During a median follow up of 12.0 months, there were 56 primary composite endpoints of all-cause death or HF hospitalization in the HF cohort. The highest FGF21 tertile was associated with a 3-fold increased risk of the composite outcome compared with the lowest tertile. After adjusting for age, sex, and the presence of atrial fibrillation, serum FGF21 remained independently associated with the outcome. Adding FGF21 levels to the model based on N-terminal pro B-type natriuretic peptide levels significantly improved the prognostic value (global chi-square 13.07 vs. 8.65; P=0.04).
Conclusions: Data from the present study demonstrated the importance of FGF21 as a potential biomarker that may reflect a different pathophysiologic implication from natriuretic peptides.
Heart failure (HF) is a significant global public health problem with increasing prevalence and incidence.1–3 Despite the development of outcome-improving pharmacological and device therapies, patients with HF suffer from poor quality of life, high mortality, and morbidity.4 Natriuretic peptides (NPs) are secreted by cardiomyocytes in response to myocardial stretch and are also the most common diagnostic and prognostic biomarkers for HF.5–9 Given that NPs are affected by obesity, atrial fibrillation (AF), and renal function, a multi-biomarker approach has been proposed to improve risk stratification.10–13
Fibroblast growth factor 21 (FGF21) is a pleiotropic endocrine hormone secreted mainly by the liver and plays an important role in regulating glucose, lipid, and energy metabolism.14–16 FGF21 levels have been reported to be elevated in patients with type 2 diabetes or coronary artery disease, which are important comorbidities of HF.17–19 Although several animal studies have revealed various effects on cardiomyocytes, including protection against oxidative stress, cardiac hypertrophy, and inflammation, the pathophysiological role of FGF21 reflects in HF remains unclear.20 Recently, it has been shown that FGF21 may be related to the prognosis of HF,21–26 but meta-analyses have yielded neutral results.27
Accordingly, this study sought to examine whether circulating FGF21 levels would be elevated in patients with HF compared with control participants free of HF, and whether circulating FGF21 levels would correlate with the severity of cardiac remodeling and clinical outcomes. We also elucidated whether they had independent and incremental prognostic value over NPs levels.
Patients with compensated HF who were hospitalized at the Gunma University Hospital between 2015 and 2018 were enrolled in this prospective study. All patients with HF were required to have recent hospitalization for HF treated with intravenous diuretics to ensure the unequivocal presence of HF. We excluded patients with non-alcoholic fatty liver disease, unstable coronary disease or recent revascularization, constrictive pericarditis, or myocarditis. Patients without HF who were referred for coronary angiography were included as a comparison group. The control group, consisting of 20 patients, was selected to match the HF group for age and sex. The control group was required to have no history of HF hospitalization or evidence of elevated left ventricular (LV) filling pressures as determined using echo-Doppler. The Gunma University Hospital Clinical Research Review Board approved this study. The authors confirmed that patient consent forms have been obtained for this article. This study was conducted in accordance with the Declaration of Helsinki. All authors have read and agreed to the manuscript as written.
Biomarker MeasurementsVenous blood samples were obtained early in the morning after an overnight fast in a compensated state. Hemoglobin, creatinine, glucose, lipid profiles, and N-terminal pro-B-type NP (NT-proBNP) were measured by routine automated laboratory procedures. Serum FGF21 levels were measured using commercially available enzyme-linked immunosorbent assay (ELISA) kits (R&D Systems). As specified by the manufacturer, the lower limit of detection of serum FGF21 was 4.67 pg/mL.
Assessment of Cardiac Structure and FunctionTwo-dimensional, M-mode, Doppler and tissue Doppler echocardiography was performed within 2 days of the blood sampling. Echocardiographic measurements were performed according to the American Society of Echocardiography guidelines.28 LV systolic function was assessed using ejection fraction (EF) and systolic mitral annular tissue velocity (mitral s′ velocity). LV diastolic function was assessed using the transmitral velocities (early and late-diastolic inflow velocities [E and A] and deceleration time of the E), early diastolic septal mitral annular tissue velocity (e′), and the ratio of E/e′. Left atrial (LA) volume was calculated using the method of disks. LA volume and LV mass were then indexed to body surface area. Right atrial (RA) pressure was estimated from the diameter of the inferior vena cava and its respiratory change. Right ventricular systolic pressure (RVSP) was then calculated as (4 × peak tricuspid regurgitation [TR] velocity)2 + estimated RA pressure. The H2FPEF score was calculated based on 4 clinical characteristics (age, body mass index, AF, and antihypertensive drugs), E/e′ ratio, and RVSP.29
Outcome AssessmentWe collected follow-up data from the day of blood sampling. The primary endpoint was a composite of all-cause death and HF hospitalization. The secondary endpoint was HF hospitalization, which was defined as acute exacerbation of symptoms of dyspnea and pulmonary edema on chest X-ray requiring intravenous diuretics treatment. Investigators obtained follow-up data from the patient’s medical records, telephone interviews, and notices of death from other hospitals.
Statistical AnalysisData are reported as mean (SD), median (interquartile range [IQR]) or number (%) unless otherwise specified. Unpaired t-test, Wilcoxon rank-sum test, or chi-square test were used to compare between-group differences, as appropriate. Relationships between 2 variables of interest were assessed by using Pearson’s or Spearman’s correlation coefficients, as appropriate. The incidence rate (per 1,000 person-years) was calculated by dividing the number of events by the total number of person-years of follow up. Kaplan-Meier curve analysis was used to assess event-free rates, and univariable and multivariable Cox proportional hazards models were then applied to evaluate the independent prognostic power in the HF population. The incremental prognostic value was assessed by comparing −2 log-likelihood values with and without the parameter to χ2 distribution at a degree of freedom of 1. The association between FGF21 levels and the composite outcome was also assessed using restricted cubic spline curve analysis. Cox proportional analyses with an interaction term were used to assess interaction on the association between log-transformed FGF21 and the composite outcome in each subgroup. All tests were 2 sided, with a P value <0.05 considered significant. Statistical analyses were performed using JMP 17.0.0 (SAS Institute, Cary, NC, USA) and R v4.2.1 (R Foundation for Statistical Computing, Vienna, Austria).
After excluding 17 patients with non-alcoholic fatty liver disease, the final study cohort included 182 patients, of whom 162 were patients with HF and 20 were controls without HF. Age, sex, body mass index, and the prevalence of diabetes and hypertension were similar between the groups (Table 1). Compared with the controls, patients with HF had a higher prevalence of AF and less prevalence of dyslipidemia and coronary artery disease. Blood pressure and heart rate were similar between the groups. As expected, patients with HF were treated with diuretics and mineralocorticoid receptor antagonists more frequently than the controls. Patients with HF had lower hemoglobin levels and estimated glomerular filtration rates and higher NT-proBNP levels than controls.
Baseline Characteristics
| Control (n=20) |
Heart failure (n=162) |
P value | |
|---|---|---|---|
| Age (years) | 70±8 | 71±14 | 0.72 |
| Male | 14 (70) | 89 (55) | 0.19 |
| Body mass index (kg/m2) | 22.7±3.5 | 22.2±4.4 | 0.71 |
| HFrEF/HFmrEF/HFpEF (%) | – | 33/15/52 | – |
| Comorbidities | |||
| Diabetes | 7 (35) | 45 (28) | 0.52 |
| Hypertension | 14 (70) | 99 (61) | 0.45 |
| Dyslipidemia | 13 (65) | 61 (38) | 0.02 |
| Coronary artery disease | 19 (95) | 56 (37) | <0.0001 |
| AF | 2 (10) | 69 (45) | 0.001 |
| Vital signs | |||
| Systolic BP (mmHg) | 121±16 | 122±21 | 0.80 |
| Diastolic BP (mmHg) | 67±8 | 67±13 | 0.97 |
| Heart rate (beats/min) | 70±13 | 72±13 | 0.53 |
| Medications | |||
| ACEI or ARB | 9 (45) | 83 (52) | 0.54 |
| β-blocker | 7 (35) | 90 (57) | 0.07 |
| Loop diuretic | 0 (0) | 127 (84) | <0.0001 |
| MRA | 0 (0) | 85 (56) | <0.0001 |
| Laboratory data | |||
| Hemoglobin (g/dL) | 13.2±1.7 | 12.2±2.2 | 0.04 |
| Total bilirubin (mg/dL) | 0.71±0.24 | 0.76±0.39 | 0.51 |
| AST (U/L) | 23.8±6.7 | 24.6±9.0 | 0.72 |
| ALT (U/L) | 19.3±9.9 | 19.9±14.6 | 0.85 |
| Glucose (mg/dL) | 106 [88–138] | 108 [94–147] | 0.70 |
| eGFR (mL/min/1.73 m2) | 69±15 | 55±28 | 0.002 |
| NT-proBNP (pg/mL) | 194 [103–303] | 1,615 [830–3,615] | <0.0001 |
Data are presented as mean±SD, median [interquartile range], or n (%). ACEI angiotensin-converting enzyme inhibitors; AF, atrial fibrillation; ALT, alanine aminotransferase; ARB, angiotensin-receptor blockers; AST, aspartate aminotransferase; BP, blood pressure; eGFR, estimated glomerular filtration rate; HFmrEF, heart failure with mid-range ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; MRA, mineralocorticoid receptor antagonists; NT-proBNP, N-terminal pro B-type natriuretic peptide.
Compared with the controls, patients with HF had a larger LV end-diastolic volume and mass (Table 2). LV systolic function was impaired in the patients with HF than in controls, with a lower EF and mitral s′ tissue velocity. Patients with HF displayed impaired LV diastolic function, evidenced by higher mitral E-wave, E/e′ ratio, and RVSP, shorter deceleration time, and a larger LA volume index than the controls.
Cardiac Structure and Function
| Control (n=20) |
Heart failure (n=162) |
P value | |
|---|---|---|---|
| LV structure | |||
| LV end diastolic volume (mL) | 71±29 | 121±54 | <0.0001 |
| LV mass index (g/m2) | 89±16 | 122±37 | <0.0001 |
| LV ejection fraction (%) | 66±7 | 48±15 | <0.0001 |
| Diastolic function and PH | |||
| Mitral E-wave (cm/s) | 63±16 | 86±29 | <0.0001 |
| Deceleration time (s) | 259±65 | 206±78 | 0.003 |
| Mitral A-wave (cm/s) | 83±20 | 76±26 | 0.27 |
| E/A ratio | 0.77 [0.59–0.88] | 0.87 [0.70–1.53] | 0.03 |
| Mitral annular e′ (cm/s) | 5.2±1.1 | 4.7±1.6 | 0.07 |
| Mitral annular s′ (cm/s) | 6.7±1.3 | 4.7±1.7 | <0.0001 |
| E/e′ ratio | 12.3±3.4 | 19.9±8.2 | <0.0001 |
| LA volume index (mL/m2) | 25 [18–28] | 52 [38–68] | <0.0001 |
| Estimated RVSP (mmHg) | 22±6 | 31±11 | <0.0001 |
| eRAP (mmHg) | 3±0 | 5±3 | 0.04 |
Data are presented as mean±SD or median [interquartile range]. A-wave, late diastolic mitral inflow velocity; BP, blood pressure; E-wave, early diastolic mitral inflow velocity; e′, early diastolic mitral annular tissue velocity; eRAP, estimated right atrial pressure; LA, left atrial; LV, left ventricular; PH, pulmonary hypertension; RVSP, right ventricular systolic pressure; s′, systolic mitral annular tissue velocity.
Serum FGF21 Levels
Compared with the controls, serum levels of FGF21 were elevated in patients with HF (100 [76–213] vs. 234 [133–564] pg/mL; P=0.0007; Figure 1, Supplementary Figure 1). Levels of FGF21 were mildly but significantly correlated with lower hemoglobin (r=−0.20; P=0.009), NT-proBNP (r=0.34; P<0.0001), and estimated glomerular filtration rate (eGFR; r=−0.20; P=0.008). Serum FGF21 levels remained higher in patients with HF than in controls after adjusting for hemoglobin and eGFR (P=0.007). While circulating FGF21 levels were unrelated to indices of LV structure and function (end-diastolic volume, LV mass index, EF, E- and A-waves, mitral e′ and s′ velocities, and the E/e′ ratio; all |r| <0.2), there was a modest correlation between FGF21 levels and RVSP (r=0.24; P=0.004; Supplementary Table 1).
Compared with the control, patients with heart failure (HF) had higher fibroblast growth factor 21 (FGF21) levels.
Prognostic Impact of FGF21 Levels in HF
Over a median follow up of 12.0 months (IQR 2.0–30.4), there were 56 composite endpoints (13 all-cause deaths and 43 HF hospitalizations) in the HF cohort (n=162). Supplementary Tables 2 and 3 show baseline characteristics and cardiac structure and function according to the tertiles of FGF21 levels. The rates of the primary composite outcome monotonically increased from 145 per 1,000 person-years in the lowest FGF21 tertile to 370 per 1,000 person-years in the highest tertile (Table 3). Kaplan-Meier analysis showed a dose-dependent worsening of event-free survival across the FGF21 tertiles (Figure 2A). The cubic spline analysis also showed a linear association between FGF levels and the composite outcome (Supplementary Figure 2). In an unadjusted Cox model, patients in the highest FGF21 tertile (T3) had a nearly 3-fold increased risk of adverse outcomes compared with those in the lowest quintile (T1; hazard ratio [HR] 2.63; 95% confidence interval [CI] 1.32–5.25; P=0.006; Table 3). After adjusting for age, sex, and the presence of AF, risk estimates for the composite outcome associated with FGF21 tertiles remained significant (T3 vs. T1; HR 2.79; 95% CI 1.32–5.88; P=0.006). Even after further adjusting for NT-proBNP levels, the association remained significant (T3 vs. T1; HR 2.67; 95% CI 1.29–5.53; P=0.008). Furthermore, sequential Cox hazard models revealed that the addition of FGF21 levels significantly improved the model based on either NT-proBNP (global χ2 13.07 vs. 8.65; P=0.04) or the H2FPEF score (global χ2 11.0 vs. 6.5; P=0.04). When patients with HF were classified into 4 groups based on the median values of FGF21 (237 pg/mL) and NT-proBNP (1,615 pg/mL), a group of patients with higher FGF21 and NT-proBNP had higher rates of the composite outcome (Figure 3). The association between FGF21 levels and the primary composite outcome was consistent across subgroups except for age (Supplementary Table 4).
Univariable and Multivariable Cox Proportional Hazard Models for the Association Between FGF21 and a Primary Outcome (All-Cause Death or HF Hospitalization)
| FGF21 serum level tertiles (pg/mL) | |||
|---|---|---|---|
| T1 | T2 | T3 | |
| Patients (n) | 54 | 54 | 54 |
| FGF21 level (pg/mL) | 107 (69–137) | 237 (204–283) | 776 (564–1,204) |
| Events, n (%) | 12 (22) | 19 (35) | 25 (46) |
| Incident rate (per 1,000 person-years) | 145 | 205 | 370 |
| Model | HR (95% CI) | HR (95% CI) | HR (95% CI) |
| 1: Unadjusted | 1 (Ref.) | 1.52 (0.74–3.13) | 2.63 (1.32–5.25)* |
| 2: Age + sex | 1 (Ref.) | 1.29 (0.62–2.70) | 2.49 (1.25–5.00)* |
| 3: Age + sex + any AF | 1 (Ref.) | 1.38 (0.64–2.99) | 2.68 (1.31–5.50)* |
| 4: Age + sex + any AF + NT-proBNP | 1 (Ref.) | 1.37 (0.63–2.99) | 2.67 (1.29–5.53)* |
*P<0.05 vs. reference. AF, atrial fibrillation; CI, confidence interval; FGF21, fibroblast growth factor 21; HF, heart failure; HR, hazard ratio; NT-proBNP, N-terminal pro B-type natriuretic peptide.
Kaplan-Meier analysis showed a dose-dependent worsening for event-free survival among fibroblast growth factor 21 tertiles. (A) All-cause death and heart failure (HF) hospitalization. (B) HF hospitalization.
When patients with heart failure (HF) were classified into 4 groups based on the median values of FGF21 (237 pg/mL) and NT-proBNP (1,615 pg/mL), a group of patients with higher FGF21 and NT-proBNP had higher rates of the composite outcome. FGF21, fibroblast growth factor 21; NT-proBNP, N-terminal pro-B-type natriuretic peptide.
Incident rates of HF hospitalization also monotonically increased from 121 per 1,000 person-years in the lowest FGF21 tertile to 281 per 1,000 person-years in the highest tertile (Table 4). Kaplan-Meier analysis showed a dose-dependent decrease in event-free survival among FGF21 tertiles (Figure 2B). In an unadjusted Cox model, patients in the highest FGF21 tertile (T3) had a 2.4-fold increased risk of HF hospitalization compared with those in the lowest quintile (T1; HR 2.41; 95% CI 1.12–5.18; P=0.03; Table 4). After adjusting for age and the presence of AF, risk estimates for the secondary endpoint remained significant (T3 vs. T1; HR 2.28; 95% CI 1.06–4.92; P=0.04). Furthermore, the association remained significant after further adjusting for NT-proBNP levels (T3 vs. T1; HR 2.32; 95% CI 1.07–5.05; P=0.04).
Univariable and Multivariable Cox Proportional Hazard Models for the Association Between FGF21 and Secondary Outcome (HF Hospitalization)
| FGF21 serum level tertiles (pg/mL) | |||
|---|---|---|---|
| T1 | T2 | T3 | |
| Patients (n) | 54 | 54 | 54 |
| FGF21 level (pg/mL) | 107 (69–137) | 237 (204–283) | 776 (564–1,204) |
| Events, n (%) | 10 (19) | 14 (26) | 19 (35) |
| Incident rate (per 1,000 person-years) | 121 | 145 | 281 |
| Model | HR (95% CI) | HR (95% CI) | HR (95% CI) |
| 1: Unadjusted | 1 (Ref.) | 1.34 (0.59–3.02) | 2.41 (1.12–5.18)* |
| 2: Age | 1 (Ref.) | 1.17 (0.51–2.67) | 2.32 (1.08–5.00)* |
| 3: Age + any AF | 1 (Ref.) | 1.17 (0.50–2.71) | 2.28 (1.06–4.92)* |
| 4: Age + any AF + NT-proBNP | 1 (Ref.) | 1.18 (0.51–2.74) | 2.32 (1.07–5.05)* |
*P<0.05 vs. reference. Abbreviations as in Table 3.
In the present study, circulating FGF21 levels were higher in patients with HF than the controls without HF. Against our hypothesis, there were no or only modest correlations of FGF21 levels with clinical variables and echocardiographic parameters. However, circulating FGF21 is associated with unfavorable outcomes after adjustment for age, sex, AF, and NT-proBNP in patients with HF. Furthermore, circulating FGF21 had an incremental prognostic value over clinical factors and NT-proBNP levels. These results highlight the importance of circulating FGF21 as a potential surrogate marker that may represent different pathophysiologic and prognostic implications from NPs.
Elevation of FGF21 Levels in HFPatients with HF displayed higher FGF21 levels compared with the controls. There are probably multiple mechanisms for increased FGF21 in HF. The first mechanism may be that FGF21 is secreted by the liver in response to hepatic congestion.23 Although FGF21 gene expression was found in the heart in advanced HF, its secretion from the heart was extremely low.23 The elevation of FGF21 levels may be a protective effect against increased pulmonary artery pressure.30,31 Another possible mechanism is that FGF21 is a systemic metabolic abnormality in HF.17–20 Refsgaard Holm et al. reported that FGF21 levels were associated with higher interleukin-6, lower body mass index, and higher total cholesterol levels.25
Prognostic Impact of FGF21 in HFNPs have been extensively studied and are currently utilized in clinical practice for the diagnosis of HF.5 However, NP levels are often elevated in the elderly and in those with anemia, chronic kidney diseases, and other cardiac conditions such as acute coronary syndrome or myocarditis.7,8 This suggests a multi-biomarker approach to improve diagnostic accuracy and risk stratification.10–12
Recent studies have suggested FGF21 as a promising candidate. Prognostic values of FGF21 were first reported in HF with preserved EF by Chou et al.,21 followed by findings in HF with reduced EF. Wu et al. showed that FGF21 levels were a better predictor of outcome after hospital discharge than NT-proBNP in patients with acute HF.26 In the present study, FGF21 levels were associated with an increased risk of adverse outcomes in HF, even after adjusting for age and NT-proBNP. Furthermore, we revealed that FGF21 had an incremental prognostic value over NT-proBNP.
A growing number of studies have investigated the beneficial effects of FGF21 on the cardiovascular system. In animal studies, it has been reported that FGF21 acts in a pleiotropic manner and has novel protective effects.20 However, in clinical studies, elevated FGF21 levels are associated with increased cardiovascular risk and mortality.32 Further investigation is needed to determine whether FGF21 can be a therapeutic target.
Study LimitationsThis was a single center study from a tertiary referral center and as such has inherent flaws related to selection and referral bias. All participants were Japanese. The control group was not healthy, in that they were referred to coronary angiography, and therefore had a higher prevalence of coronary artery disease. However, the fact that the control population was more diseased than a truly normal healthy control population only biases our data towards the null. The number of patients in the control group was small. The lack of external validation is the primary limitation. Further study is warranted to confirm our findings. The H2FPEF score was originally developed for screening HF with preserved EF patients, but this study used it to examine the incremental prognostic ability of FGF21. Previous studies have demonstrated the prognostic value of this score in HF patients, which may support its use for this purpose.33,34
Patients with HF display elevations in circulating FGF21 levels with little association with cardiac function. Importantly, higher FGF21 levels predict worse clinical outcomes in HF, with an incremental prognostic value over NPs. The current data highlight the importance of FGF21 as a potential biomarker with a different pathophysiologic implication from NPs.
This work was supported by MSD Life Science Foundation, Public Interest Incorporated Foundation (to H.S.), and Japan Heart Foundation Research Grant (to H.S.). M.O. received research grants from the Japanese Circulation Society, Japanese College of Cardiology, AMI Inc, Nippon Boehringer-Ingelheim, Janssen Pharmaceutical K.K., JSPS KAKENHI (21K16078), and AMED (23jm0210104h0002). The authors thank Reiko Oda for data collection.
M.O. received speaker honoraria from Novartis, Otsuka Pharmaceutical, and Nippon Boehringer-Ingelheim. H.I. received lecture fees from AstraZeneca Inc., Bayer Pharmaceutical Co., Ltd, Boehringer Ingelheim Japan, Bristol-Myers Squibb Inc., Daiichi-Sankyo Pharma Inc., MSD K. K., Mitsubishi Tanabe Pharma Co., Ltd, Mochida Pharmaceutical Co., Ltd, Novartis Japan, and Pfizer Japan Inc. H.I. is a member of Circulation Reports’ & Circulation Journal’s Editorial Team.
Gunma University Hospital Clinical Research Review Board (HS2018–239).
The deidentified participant data will not be shared.
Please find supplementary file(s);
https://doi.org/10.1253/circrep.CR-25-0067
