Abstract
Aims
There are only a few studies on novel biomarkers for incident heart failure (HF). We investigated the association of multiple circulating biomarkers with incident HF in a large prospective population‐based study.
Methods and results
Conventional risk factors and inflammatory biomarkers were measured, and systemic metabolic measures determined by a high‐throughput serum nuclear magnetic resonance platform in a population‐based Metabolic Syndrome in Men study including 10 106 Finnish men without HF at baseline. During an 8.8 year follow‐up, 172 (1.7%) participants developed HF. Adiponectin, high‐sensitivity C‐reactive protein (hs‐CRP), glycoprotein acetyls, alanine, phenylalanine, glycerol, and pyruvate were associated with incident HF in unadjusted Cox regression analyses, in addition to age, systolic blood pressure, body mass index (BMI), waist circumference, fasting plasma glucose and insulin, haemoglobin A1c (HbA1c), and urinary albumin excretion rate (UAER). After adjustment for age, BMI, diabetes, and statin medication, only adiponectin [hazard ratio (HR) 1.18 (1.10–1.26, P = 4.1E‐08)], pyruvate [HR 1.38 (1.28–1.50, P = 8.2E‐05)], and UAER [HR 1.15 (1.11—1.18, P = 7.8E‐06)] remained statistically significant. In principal component analysis of biomarkers associated with HF in univariate Cox regression analysis, we identified six components, explaining 61.7% of total variance. Four principal components, one with significant loadings on waist, BMI, fasting plasma insulin, interleukin 1 receptor antagonist, and hs‐CRP; another on pyruvate, glycoprotein acetyls, alanine, glycerol and HbA1c; third on age and glomerular filtration rate; and fourth on systolic blood pressure, UAER, and adiponectin, significantly associated with incident HF.
Conclusions
Several novel metabolic and inflammatory biomarkers were associated with incident HF, suggesting early activation of respective pathways in the pathogenesis of HF.
Keywords: Heart failure, Hypertension, Biomarkers, Metabolomics, Inflammation, Principal component analysis
Introduction
Heart failure (HF), caused by abnormal cardiac structure and function, represents the end stage of several cardiac diseases, and significantly worsens both quality of life and life expectancy with a 5 year survival rate of approximately 50%. Lifetime risk of the development of HF is 20% and overall prevalence 2%. There are over 30 million cases of HF worldwide resulting in an overall annual economic cost of over $100 billion. 1 , 2 , 3
Natriuretic peptides and troponins have been used in clinical diagnosis and treatment of HF, and they also predict incident HF. 4 , 5 , 6 , 7 There are, however, only few studies on other biomarkers of incident HF, 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 and very few of them have investigated large panels of biomarkers. 10 , 13 , 14 , 15 Several pathogenic mechanisms are suggested to be operative in HF, but they are only partially understood. 1 , 7 Transition from current practice of treating HF to biomarker‐based precise medicine targeted to prevent HF requires investigation of multiple biomarkers reflective of pathophysiological pathways resulting in HF. 1 , 7
Inflammatory biomarkers are elevated in HF, 1 , 7 but there are only few prospective studies on the subject showing that high‐sensitivity C‐reactive protein (hs‐CRP) and interleukin‐6 predict HF. 8 , 14 , 15 , 17 Metabolomics, or metabolic profiling, permits a large‐scale analysis of many metabolic alterations occurring in HF. 1 , 10 , 13 There are only few studies on metabolic biomarkers associated with incident HF. 10 , 13 , 16 Elevated phenylalanine concentrations and four other metabolites of amino acid metabolism have been associated with incident HF. 10 , 13 Recently, we have used systems genetics approach to identify a new biomarker glycoprotein NMB, playing a role in tissue regeneration by regulation of inflammatory response and suppressing fibrosis, which was negatively associated with HF. 16
The aim of the present study was to investigate the association of a large panel of inflammatory biomarkers and metabolites, in addition to anthropometric measures and conventional risk factors, with incident HF in 10 106 Finnish men.
Methods
Subjects and clinical measurements at baseline
The baseline study of the Metabolic Syndrome in Men (METSIM) included 10 197 men, aged from 45 to 73 years, randomly selected from the population register of Kuopio, Eastern Finland (population 95 000). At baseline, every study subject participating in the study gave written consent, and had a 1 day outpatient visit to the Clinical Research Unit at the University of Kuopio, 18 where blood samples for proton nuclear magnetic resonance (NMR) analysis, metabolic, inflammatory, and other biomarkers were taken. The design and methods of the METSIM study have been previously described in detail. 19 In the present study, all participants with prevalent HF before the baseline study (n = 91) were excluded from statistical analyses, giving the final study population of 10 106 men. The baseline measurements are available practically in all 10 106 subjects (Table 1 ).
Table 1.
Cox regression analysis of demographic and anthropometric variables and biomarkers as predictors for heart failure in 10 106 participants of an 8.8 year follow‐up study of the METSIM cohort
Variable | N, non‐case/case | HR | 95% CI | P | P* | P** | P*** | P**** | |
---|---|---|---|---|---|---|---|---|---|
Lower | Upper | ||||||||
Age (years) | 9934/172 | 1.12 | 1.09 | 1.14 | 1.4E‐19 | ||||
Type 2 diabetes (yes/no) | 9910/171 | 3.33 | 2.43 | 4.57 | 9.2E‐14 | 6.8E‐09 | 1.5E‐04 | ||
Systolic blood pressure (mmHg) | 9933/172 | 1.36 | 1.19 | 1.56 | 5.1E‐06 | 0.014 | 0.162 | 0.391 | 0.347 |
Body mass index (kg/m2) | 9931/172 | 1.50 | 1.34 | 1.67 | 1.8E‐12 | 1.9E‐13 | |||
Waist circumference (cm) | 9929/172 | 1.63 | 1.45 | 1.83 | 5.9E‐16 | 3.6E‐15 | 0.002 | 0.011 | 0.010 |
Fasting plasma glucose (mmol/L) | 9934/172 | 1.22 | 1.13 | 1.33 | 1.1E‐06 | 3.5E‐05 | 0.044 | 0.569 | 0.602 |
Fasting plasma insulin (mU/L) | 9930/172 | 1.13 | 1.07 | 1.19 | 9.0E‐06 | 1.8E‐04 | 0.449 | 0.815 | 0.681 |
HbA1c (%) | 9911/172 | 1.33 | 1.25 | 1.43 | 8.3E‐17 | 3.9E‐11 | 1.1E‐05 | 0.038 | 0.056 |
Glomerular filtration rate | 9934/172 | 0.73 | 0.63 | 0.86 | 8.6E‐05 | 0.256 | 0.297 | 0.198 | 0.217 |
Urinary albumin excretion rate (μg/min) | 9798/167 | 1.15 | 1.11 | 1.18 | 8.7E‐17 | 5.8E‐13 | 1.3E‐09 | 3.8E‐06 | 7.8E‐06 |
Plasma adiponectin (μg/mL) | 9932/172 | 1.18 | 1.10 | 1.26 | 3.9E‐06 | 5.0E‐04 | 1.7E‐06 | 2.3E‐07 | 4.1E‐08 |
High sensitivity CRP (mg/L) | 9933/172 | 1.09 | 1.04 | 1.15 | 4.5E‐04 | 0.002 | 0.038 | 0.047 | 0.032 |
IL1 receptor antagonist (pg/L) | 9933/172 | 1.15 | 1.05 | 1.26 | 0.003 | 0.003 | 0.931 | 0.707 | 0.683 |
Glycoprotein acetyls (mmol/L) | 9878/170 | 1.25 | 1.11 | 1.41 | 3.1E‐04 | 7.2E‐05 | 0.036 | 0.186 | 0.223 |
Alanine (mmol/L) | 9885/170 | 1.31 | 1.15 | 1.49 | 5.5E‐05 | 0.001 | 0.081 | 0.448 | 0.537 |
Histidine (mmol/L) | 9891/170 | 0.75 | 0.64 | 0.88 | 0.001 | 0.017 | 0.058 | 0.122 | 0.163 |
Phenylalanine (mmol/L) | 9878/168 | 1.21 | 1.09 | 1.33 | 1.6E‐04 | 0.006 | 0.357 | 0.396 | 0.393 |
Tyrosine (mmol/L) | 9890/170 | 1.22 | 1.07 | 1.39 | 0.003 | 0.007 | 0.419 | 0.458 | 0.438 |
Acetoacetate (mmol/L) | 9891/170 | 1.14 | 1.05 | 1.24 | 0.002 | 0.008 | 0.006 | 0.035 | 0.027 |
Glycerol (mmol/L) | 9657/166 | 1.36 | 1.21 | 1.53 | 4.1E‐07 | 5.0E‐06 | 0.012 | 0.187 | 0.188 |
Pyruvate (mmol/L) | 9891/170 | 1.38 | 1.28 | 1.50 | 9.5E‐15 | 1.0E‐13 | 9.4E‐08 | 5.5E‐05 | 8.2E‐05 |
CI, confidence interval; CRP, C‐reactive protein; HbA1c, haemoglobin A1c.
Analysis includes 10 106 participants without heart failure (HF) at baseline, of whom 172 participants developed incident HF during the follow‐up period. Hazard ratios (HR) are standardized. P < 0.002 (0.05/21) is considered as statistically significant (bold). P < 0.05 is considered as nominally significant (underlined). P unadjusted. P* adjusted for age. P** adjusted for age and body mass index. P*** adjusted for age, body mass index and diabetes. P**** adjusted for age, body mass index, diabetes and statin medication.
Anthropometric measurements at baseline
Height and weight were measured to the nearest 0.5 cm and .1 kg, respectively. Body mass index (BMI) was calculated as weight (kg) divided by height (m) squared. Waist circumference was measured at the midpoint between the lateral iliac crest and lowest rib. Smoking status was defined as current smoking (yes vs. no). Diagnosis of hypertension was based on the use of antihypertensive medication (information obtained from the National Drug Reimbursement registry).
Oral glucose tolerance test at baseline
A 2 h oral glucose tolerance test (75 g of glucose) was performed, and samples for plasma glucose and insulin were drawn at 0, 30, and 120 min. Glucose tolerance based on an oral glucose tolerance test was classified according to ADA 2003 criteria, including type 2 diabetes (fasting glucose tolerance test ≥ 7.0 mmol/L and/or 2 h plasma glucose ≥ 11.1 mmol/L). 20
Laboratory measurements
Levels of plasma glucose, insulin, haemoglobin A1c (HbA1c), glomerular filtration rate and urinary albumin excretions rate were measured as previously described. 18 , 21
Inflammatory biomarkers
Serum concentrations of hs‐CRP were assayed using kinetic immunoturbidimetry (near‐infrared particle immunoassay; IMMAGE Immunochemistry System; Beckman Coulter, Fullerton, CA) and plasma interleukin 1 receptor antagonist (IL‐1RA) using a photometric immunoassay [enzyme‐linked immunosorbent assay (ELISA)] method (Quantikine DRA00 Human IL‐1RA; R&D Systems, Inc., Minneapolis, MN). Plasma adiponectin was measured using ELISA (human adiponectin ELISA kit; Linco Research, St. Charles, MI). Concentrations of glycoprotein acetyls (including mainly alpha‐1‐acid glycoprotein) were quantified from serum samples using a high‐throughput proton nuclear magnetic resonance metabolomics platform, as previously described. 21
Proton nuclear magnetic resonance analysis
For all METSIM study participants, proton NMR analysis was performed at the same time from the baseline serum samples. Altogether, 27 serum metabolic measures, including amino acids, fatty acids, ketone bodies, and lipoprotein lipids (Supporting Information, Table S3 ), were determined using a high‐throughput serum NMR platform operating at 500 MHz. Fasting serum samples collected at baseline were stored at −80°C and thawed overnight in a refrigerator before sample preparation. Aliquots of each sample (300 μL) were mixed with sodium phosphate buffer (300 μL). The NMR data were measured using a Bruker AVANCE III spectrometer operating at 500.36 MHz (1H observation frequency; 11.74 T) and equipped with an inverse selective SEI probehead, including an automatic tuning and matching unit and a z‐axis gradient coil for automated shimming. A BTO‐2000 thermocouple serves to temperature stabilization, the sample at the level of approximately 0.01°C. The NMR technology is widely used, standardized, and accredited (https://nightingalehealth.com/about/technology). Details of the NMR experimentation have been described previously. 22 , 23
Diagnosis of cardiovascular disease
Diagnoses of cardiovascular disease were obtained from the hospital registry of Kuopio University Hospital, which includes all diagnoses of inpatients and outpatients. Kuopio University Hospital is the only hospital and cardiology outpatient clinic treating patients with HF in the Kuopio region. All patients with HF from the Kuopio region are referred to the Kuopio University Hospital for echocardiography and diagnostic cardiac work‐up.
Diagnosis of heart failure
Medical records of all participants with hospital discharge or outpatient clinic diagnoses suggestive of HF (International Classification of Diseases, 10th Revision [ICD‐10]: I11.0, I11.9, and I50) before or after the baseline examination were reviewed in detail by a medical student (M.J.) and a cardiologist (J.K.). Diagnosis and aetiology of HF was based on clinical findings, N‐terminal pro brain natriuretic peptide (NT‐proBNP) measurements, electrocardiogram, chest X‐ray, and echocardiography. 4 The type of HF was classified either as HF with preserved ejection fraction (left ventricular ejection fraction ≥ 40% on echocardiography) or HF with reduced ejection fraction (left ventricular ejection fraction < 40%). Hypertensive HF was defined as HF in the presence of uncontrolled hypertension (ICD‐10: I10), and/or hypertensive heart disease (ICD‐10: I11.9) characterized by posterior wall (PW) > 12 mm, and/or left ventricle end‐diastolic diameter > 60 mm and/or left ventricle end‐systolic diameter > 35 mm on echocardiography, and no other heart disease severe enough to cause HF. Other aetiologies of HF were obtained from the medical records derived clinical diagnoses made by a cardiologist or internist and were verified by a cardiologist with a 20 year experience in clinical cardiology (J.K.). Medical records of all subjects with a diagnosis of acute coronary syndrome were reviewed by a cardiology fellow (R.J.) and a cardiologist (J.K.).
Study follow‐up and time to the diagnosis of incident heart failure
The study participants were followed‐up to a mean of 8.8 years (median 6.9 years, range 0.1–9.6 years), and incident cases of HF were recorded until 31 May 2017. Mean time from the baseline examination to the diagnosis of incident HF was 4.4 years (median 4.5 years, range 0.1–9.2 years).
Mortality data
Of the 172 study participants who developed HF, 27 died during the follow‐up. Causes of death were confirmed from medical records, and death certificates were obtained from the National Causes‐of‐Death Register of Statistics Finland, which records and reviews all deaths in Finland.
Statistical analysis
Statistical analyses were conducted using the SPSS version 21 (SPSS, Chicago, IL). Data are presented as mean ± SD. Independent samples t‐test and Pearson χ 2 test were used to assess statistical significance of the differences in clinical characteristics, inflammatory biomarkers, and proton NMR metabolic measures between the participants with and without incident HF. All variables except for age were log‐transformed to correct for their skewed distributions. Cox regression analysis was applied to investigate the association of the baseline variables with incident HF. The anthropometric parameters and biomarkers, which differed significantly in independent samples t‐test between the subjects with and without incident HF, were included in Cox regression analyses. P value 0.05, divided by the number of comparisons, was considered as statistically significant. Principal component (PC) analysis was applied to analyse the baseline variables associated with incident HF in unadjusted Cox regression analysis. Associations between the PCs and incident HF was investigated with Cox regression analysis.
Ethics statement
Study protocol was approved by the Ethics Committee of University of Eastern Finland and the Kuopio University Hospital. The study was conducted in accordance with the Helsinki Declaration. All study participants gave written informed consent.
Results
Altogether 172 (1.7%) men of the METSIM study developed incident HF during the follow‐up. Table S1 gives the baseline anthropometric, metabolic, and inflammatory marker measurements of the METSIM study participants who did and did not develop incident HF. Participants with incident HF were older, had more often type 2 diabetes, reimbursement for hypertension, and a history of myocardial infarction, compared with those without incident HF. Systolic blood pressure, BMI, waist circumference, levels of fasting plasma glucose and insulin, HbA1c, and urinary albumin excretion rate (UAER) were higher in subjects with incident HF compared with those without incident HF. Glomerular filtration rate was lower in subjects with incident HF compared with those without incident HF. Subjects with incident HF had higher concentrations of adiponectin, hs‐CRP, IL‐1RA, and glycoprotein acetyls. We investigated the number of cases of incident HF across the quintiles of plasma adiponectin concentration (Figure 1 ). The number of incident cases was lowest in the third quintile and highest in the fifth quintile. There was a statistically significant difference between the third quintile and the fifth quintile in incident HF cases. In the linearity test, there was a significant deviation from linearity (P = 2.2E‐17).
Figure 1.
Quintiles of adiponectin level contain the following adiponectin levels: I, 1.00–4.70; II, 4.80–6.10; III, 6.20–7.70; IV, 7.80–10.10; V, 10.20–112.60. P, Pearson χ 2 test over all quintiles. p, Pearson χ 2 test with the lowest quintile as the reference category.
Table S2 shows the levels of baseline serum metabolic measures in proton NMR analysis in participants with and without incident HF. The concentrations of alanine, phenylalanine, tyrosine, acetoacetate, glycerol, and pyruvate were significantly higher, and the concentration of histidine lower among the participants with incident HF compared with participants without HF.
Table S3 shows hospital medical records derived aetiology and echocardiographic, NT‐proBNP and electrocardiogram characteristics of the subjects with incident HF at the time of the diagnosis of HF, and outcomes during the follow‐up. Of subjects with HF, 68 had hypertension as the aetiology of HF. In 62, the aetiology of HF was coronary artery disease. Of them, 36 had suffered from a myocardial infarction before the baseline examination, 19 patients had HF due to an acute myocardial infarction during the follow‐up, and 7 subjects had ischemic HF without myocardial infarction. Cardiomyopathy was diagnosed in 31 (18.0%) of the subjects with incident HF. Altogether 13 (12.5%) patients had HF due to valvular heart disease. On echocardiography performed at the time of HF diagnosis, about half of the patients had preserved ejection fraction. The mean left ventricular ejection fraction was 42.9 ± 15.0%, left ventricular end‐diastolic diameter 57.3 ± 9.4 mm, and left ventricular end‐systolic diameter 44.1 ± 11.3 mm. Mean NT‐proBNP level in patients with incident HF was 4815.0 ± 8115.7 ng/L. Altogether, 82 (47.7%) participants were in sinus rhythm, and 73 (42.4%) had atrial fibrillation. During the follow‐up, implantable cardioverter‐defibrillator or cardiac resynchronization therapy device was implanted in 15 (8.7%) and 16 (9.3%) of subjects with incident HF, respectively. Altogether, 27 (15.7%) of those who developed incident HF were deceased by the end of the follow‐up period.
The clinical characteristics of incident HF by hypertensive vs. other aetiology are shown in Table S4 . Participants with hypertensive HF had more often preserved left ventricular ejection fraction and atrial fibrillation, and less often diagnosis of coronary artery disease including previous myocardial infarction, compared with those with non‐hypertensive HF.
Table 1 presents univariate Cox regression analyses of the associations of baseline anthropometric factors and biomarkers with incident HF. In unadjusted analyses, all inflammatory and metabolomic biomarkers were associated with incident HF with statistically significant hazard ratios ranging from 1.09 to 3.33, except for IL‐1RA, tyrosine, and acetoacetate, which had nominally significant associations with HF. When adjusted for age, the following biomarkers were associated with incident HF: type 2 diabetes, BMI, waist, fasting plasma glucose, fasting plasma insulin, HbA1c, UAER, adiponectin, glycoprotein acetyls, alanine, glycerol, and pyruvate. When adjusted for both age and BMI, the following parameters were significantly associated with incident HF: type 2 diabetes, blood HbA1c, UAER, plasma adiponectin, and pyruvate. When adjusted for age, BMI, diabetes and statin medication, only UAER, plasma adiponectin, and pyruvate remained statistically significant.
As many of the variables associated with incident HF in our study are intercorrelated, we used PC analysis to further study the baseline variables associated with incident HF. The components, their loadings, and percentages explained of the total variance are shown in Table 2 . Six components, accounting for 61.7% of the total variance and reflecting (1) obesity, hyperinsulinemia, and low‐grade inflammation; (2) small molecule metabolites, low‐grade inflammation, and hyperglycemia; (3) amino acids; (4) aging and declining kidney function; (5) high systolic blood pressure, albuminuria, and adiponectin levels; and (6) metabolites associated with myocardial energy metabolism, were identified (Table 2 and Figure 2 ).
Table 2.
Principal component analysis of baseline variables, which were associated with incident heart failure in unadjusted Cox regression analyses
Component | Variable | Loading | Percent variance explained | Component | Variable | Loading | Percent variance explained |
---|---|---|---|---|---|---|---|
PC 1 | Waist | 0.876 | 25.77 | PC 2 | Pyruvate | 0.798 | 9.46 |
Body mass index | 0.876 | Glycoprotein acetyls | 0.660 | ||||
Fasting plasma insulin | 0.680 | Alanine | 0.649 | ||||
IL1 receptor antagonist | 0.600 | Glycerol | 0.493 | ||||
High sensitivity CRP | 0.536 | Blood GHbA1C | 0.426 | ||||
Tyrosine | 0.315 | Fasting plasma insulin | 0.363 | ||||
Plasma adiponectin | −0.335 | Phenylalanine | 0.276 | ||||
Phenylalanine | 0.301 | Plasma adiponectin | −0.214 | ||||
Glycoprotein acetyls | 0.284 | Urinary albumin excretion | 0.212 | ||||
Blood GHbA1C | 0.293 | IL1 receptor antagonist | 0.189 | ||||
Glycerol | 0.249 | High sensitivity CRP | 0.170 | ||||
Histidine | −0.178 | Tyrosine | 0.145 | ||||
Urinary albumin excretion | 0.146 | Systolic blood pressure | 0.110 | ||||
Pyruvate | 0.146 | Waist | 0.092 | ||||
Systolic blood pressure | 0.132 | Histidine | −0.070 | ||||
Alanine | 0.086 | Body mass index | 0.061 | ||||
Age | −0.021 | Acetoacetate | −0.052 | ||||
Glomerular filtration rate | 0.009 | Glomerular filtration rate | −0.026 | ||||
Acetoacetate | 0.000 | Age | 0.011 | ||||
PC 3 | Histidine | 0.798 | 8.26 | PC 4 | Age | 0.769 | 6.70 |
Tyrosine | 0.750 | Glomerular filtration rate | −0.768 | ||||
Phenylalanine | 0.649 | Blood GHbA1C | 0.296 | ||||
Alanine | 0.423 | Phenylalanine | 0.257 | ||||
Glycoprotein acetyls | 0.208 | High sensitivity CRP | 0.177 | ||||
Glycerol | 0.164 | Histidine | −0.158 | ||||
Blood GHbA1C | −0.157 | Alanine | 0.134 | ||||
Fasting plasma insulin | 0.149 | Systolic blood pressure | 0.104 | ||||
Body mass index | 0.139 | Tyrosine | 0.078 | ||||
Plasma adiponectin | −0.137 | Acetoacetate | 0.067 | ||||
Waist | 0.122 | IL1 receptor antagonist | −0.061 | ||||
Systolic blood pressure | 0.109 | Pyruvate | −0.053 | ||||
Urinary albumin excretion | −0.105 | Plasma adiponectin | 0.050 | ||||
Glomerular filtration rate | −0.064 | Glycerol | −0.036 | ||||
Acetoacetate | −0.037 | Fasting plasma insulin | 0.035 | ||||
IL1 receptor antagonist | −0.033 | Glycoprotein acetyls | 0.027 | ||||
Pyruvate | −0.027 | Body mass index | −0.02 | ||||
High sensitivity CRP | −0.025 | Urinary albumin ecxretion | 0.023 | ||||
Age | −0.011 | Waist | 0.008 | ||||
PC 5 | Systolic blood pressure | 0.741 | 6.21 | PC 6 | Acetoacetate | 0.834 | |
Urinary albumin excretion | 0.531 | Glycerol | 0.503 | 5.29 | |||
Plasma adiponectin | 0.410 | Alanine | −0.330 | ||||
Age | 0.318 | High sensitivity CRP | 0.300 | ||||
Glycerol | 0.220 | Glycoprotein acetyls | 0.210 | ||||
Body mass index | 0.211 | IL1 receptor antagonist | 0.128 | ||||
Waist | 0.209 | Plasma adiponectin | 0.171 | ||||
High sensitivity CRP | −0.143 | Fasting plasma insulin | −0.171 | ||||
Fasting plasma insulin | 0.128 | Tyrosine | −0.156 | ||||
Pyruvate | 0.123 | Phenylalanine | 0.137 | ||||
Blood GHbA1C | 0.123 | Glomerular filtration rate | −0.089 | ||||
Alanine | 0.092 | Blood GHbA1C | −0.076 | ||||
Glomerular filtration rate | 0.090 | Histidine | 0.055 | ||||
Phenylalanine | −0.066 | Systolic blood pressure | 0.033 | ||||
Histidine | −0.063 | Body mass index | −0.030 | ||||
IL1 receptor antagonist | −0.060 | Age | −0.022 | ||||
Acetoacetate | 0.048 | Urinary albumin excretion | −0.005 | ||||
Tyrosine | 0.045 | Waist | −0.003 | ||||
Glycoprotein acetyls | 0.014 | Pyruvate | −0.001 |
CRP, C‐reactive protein; HbA1c, haemoglobin A1c; PC, principal component.
Analysis includes 10 106 participants without heart failure (HF) at baseline, of whom 172 participants developed incident HF during the follow‐up period. All variables except for age were log‐transformed for statistical analysis. Bold indicates variables with significant (>0.400) loadings.
Figure 2.
In a non‐selected population of 10 106 Finnish men without heart failure (HF) at baseline, biomarkers including novel small molecule energy metabolites and inflammatory markers, clustered in six principal components, which were associated with incident heart failure with high or borderline statistical significance, suggesting early activation of multiple pathogenic pathways in heart failure. HR, hazard ratio; PC, principal component.
Table 3 and Figure 2 show Cox regression analysis of PCs and incident HF. PCs 1, 2, 4, and 5 were significantly associated with HF with hazard ratios ranging from 1.44 to 1.87. PC 3 had nominal negative association with HF, and PC 6 had a borderline significant association with HF.
Table 3.
Principal components (Table 2 ) associated with incident heart failure in the METSIM cohort in adjusted Cox regression analysis
PC number | HR | 95% CI | P |
---|---|---|---|
PC 1 | 1.49 | 1.30, 1.71 | 2.0E‐08 |
PC 2 | 1.44 | 1.25, 1.66 | 3.9E‐07 |
PC 3 | 0.83 | 0.70, 0.97 | 0.019 |
PC 4 | 1.73 | 1.50, 1.99 | 2.3E‐14 |
PC 5 | 1.87 | 1.64, 2.15 | 1.0E‐19 |
PC 6 | 1.15 | 0.99, 1.34 | 0.061 |
CI, confidence interval; HR, hazard ratio; PC, principal component.
Cox regression includes 10 106 participants without heart failure (HF) at baseline, of whom 172 participants developed incident HF during the follow‐up period. P < 0.008 (0.05/6) is considered statistically significant (bold), and P < 0.05 is considered nominally significant (underlined).
Variables associating with hypertensive HF in unadjusted Cox models were very similar to those associating with incident HF caused by other aetiologies (Table S5 ). Systolic blood pressure, waist, adiponectin, and acetoacetate remained significant after adjustment for age, BMI, diabetes, and statin medication. Of six PCs (Table 2 ), all except for PC 3, which reflects amino acid levels, were associated with hypertensive HF in univariate Cox regression analysis (Table S6 ). We also compared biomarkers in subjects with myocardial infarction before baseline, who did (n = 36) or did not develop HF during the follow‐up (n = 424). Subjects with previous myocardial infarction who developed HF during the follow‐up had, in addition to more advanced age, higher prevalence of type 2 diabetes, higher GHbA1c and lower glomerular filtration rate, also higher levels of adiponectin than those who did not develop HF (data not shown). However, these differences were only nominally significant, preventing more detailed subgroup analysis.
Discussion
In the context of current literature
Aetiology of HF
Hypertension, coronary artery disease, particularly myocardial infarction, cardiomyopathies, and valvular heart disease were the most common causal etiologies of HF in our study, in agreement with current literature. 1 Age, type 2 diabetes, obesity, hyperinsulinemia, declining kidney function, and microalbuminuria, which are well‐established comorbidities predisposing to the development of HF, were associated with incident HF.
Novel biomarkers for incident HF
In the present study, several novel inflammatory and metabolic biomarkers—adiponectin, glycoprotein acetyls, glycerol, pyruvate, acetoacetate, alanine, and phenylalanine—were associated with incident HF in unadjusted models. Adiponectin is an anti‐inflammatory, insulin‐sensitizing, and antiatherogenic adipocytokine that enhances endothelial function. In few previous studies, adiponectin concentrations have been implicated in HF. 11 , 12 A J‐shaped association between adiponectin and HF has been suggested in one previous study. 12 Glycoprotein acetyls, biomarkers of low‐grade inflammation, have been associated with HF in one metabolomics study in unadjusted but not in adjusted models. 13 In our previous study, glycoprotein acetyls increased the risk of type 2 diabetes and cardiovascular events and correlated strongly with several cardiovascular risk factors. 21 Pyruvate is an intermediary metabolite, which enhances free energy of ATP hydrolysis in myocardium and has anti‐inflammatory properties. 24 There is only one previous publication on pyruvate in HF, 25 and there are no previous studies showing that pyruvate increases the risk of HF. Acetoacetate, a ketone body, and glycerol may be used as energy substrate by cardiac myocytes, particularly in the failing heart. Increased ketone levels have been reported in HF patients, 26 but there are no previous studies showing that acetoacetate or glycerol levels are associated with incident HF. Thus, our study is the first to show that pyruvate, acetoacetate, and glycerol are associated with incident HF. Of amino acids, elevated phenylalanine concentrations were associated with incident HF hospitalizations in the PROSPER and FINRISK 1997 studies. 13 In our previous study, alanine, leucine, isoleucine, tyrosine, and glutamine increased the risk of incident type 2 diabetes. 27 A catabolic defect in the metabolism of proteins and impaired uptake and utilization of amino acids are possible mechanisms.
Novel biomarkers of hypertensive heart failure and heart failure due to previous myocardial infarction
In hypertensive HF, adiponectin, glycerol, pyruvate, and acetoacetate were associated with incident HF in unadjusted models. In HF due to previous myocardial infarction, of novel biomarkers, only adiponectin was associated with incident HF with nominal significance. Biomarker profiles of incident HF may differ between diversing etiologies of HF, but larger studies are required to investigate this question.
Principal component analysis
Principal component analysis is a statistical procedure that converts a set of observations of correlated variables into a set of values of linearly uncorrelated variables called PCs, which can be investigated with respect to cardiovascular endpoints, and may suggest pathophysiological pathways. In the present study, PC analysis identified several components including inflammatory and metabolic biomarkers, which were associated with incident HF in addition to obesity, age, and blood pressure‐related components, suggesting that novel biomarkers have incremental effect in the pathogenesis of HF. To the best of our knowledge, PC analysis has not been used to investigate biomarkers in incident HF. In our previous study on elderly Finnish subjects, however, we applied factor analysis, which is a statistical method very similar to PC analysis, and identified insulin resistance factor, which predicted coronary events in elderly nondiabetic men. 28
Possible mechanisms
Chronic inflammation
Inflammation is implicated in HF pathophysiology by promoting cardiac injury, fibrosis, and dysfunction. 29 In the present study, adiponectin and pyruvate, both related to inflammation, were associated with incident HF in adjusted Cox regression analyses. Furthermore, in PC analysis, multiple inflammatory markers loaded on three different PCs associated with the risk of HF. Hs‐CRP and IL‐1RA loaded on the obesity‐hyperinsulinemia component, pyruvate and glycoprotein acetyls on the metabolite component, and adiponectin on the vascular component, suggesting several and diverse inflammatory cascades in HF, activated well before clinical HF.
Impairment in metabolic pathways
HF is associated with impairment in glucose, amino acid, and ketone metabolic pathways. 30 Ketone body oxidation is increased, and amino acid metabolism is deranged in HF. 31 In animal models, metabolic remodelling precedes, initiates, and sustains functional and structural heart remodelling. 31 In the present study, small molecular metabolites pyruvate, glycerol, and acetoacetate were associated with incident HF, and loaded on three PCs, one of which was significantly associated with incident HF, suggesting that metabolic dysfunction precedes clinical HF also in humans.
Strengths and limitations
The strengths of our study include a large prospective, population‐based study with carefully investigated phenotype at baseline, a comprehensive and detailed diagnostics of HF and its aetiology, and a long follow‐up period. Furthermore, the present study is among the first studies analysing a large panel of novel biomarkers, including proton NMR‐derived metabolic measures.
There are also limitations in our study. First, conventional biomarkers for HF natriuretic peptides and troponin T were not measured. In a recent study, several novel circulating biomarkers, including biomarkers of inflammation and intermediary metabolism, added the predictive value of HF‐related hospital admisissions on the top of NT‐proBNP and TnT in patients with HF and preserved ejection fraction, 32 suggesting that measuring large panels of biomarkers may give additional insight in the pathogenesis and progression of HF. Second, the present study included only middle‐aged and elderly Finnish men, and the number of participants with incident HF was rather low. Therefore, further studies in other populations are warranted to confirm our results and to assess their applicability to women and other populations and age groups.
Clinical implications
We identified several novel biomarkers for incident HF, including markers of inflammation and energy metabolism. Our findings give new insights into the pathophysiology of HF. Measuring inflammatory and metabolic biomarkers may help to predict HF and facilitate early diagnosis. Finally, targeting multiple pathophysiological pathways might be useful in the prevention and treatment of HF.
Conclusions
Plasma adiponectin and pyruvate were independently associated with HF in the 8.8 year follow‐up of the METSIM cohort. In addition, we identified six PCs of biomarkers, including novel small molecule energy metabolites and inflammatory biomarkers. Four PCs were significantly associated with indicent HF, suggesting an early activation of multiple pathophysiological pathways, including inflammation and metabolic dearrangement, in the pathogenesis of HF.
Conflict of interest
The authors declare no conflict of interest.
Funding
This work was supported by the Kuopio University Hospital and the Finnish Heart Research Foundation (grants to Johanna Kuusisto) and Academy of Finland (grants to Markku Laakso). Mika Ala‐Korpela is supported by a research grant from the Sigrid Juselius Foundation, Finland.
Supporting information
Table S1. Baseline anthropometric, metabolic, and inflammatory marker measurements of the study participants without and with incident heart failure
Table S2. Serum metabolic measures in proton NMR analysis at baseline in the participants with and without incident heart failure during the METSIM study follow‐up
Table S3. Clinical characteristics and outcomes of the METSIM study participants with incident heart failure.
Table S4. Clinical characteristics and outcomes by the etiology of heart failure in the METSIM study participants with incident heart failure
Table S5. Cox regression analysis of demographic and anthropometric variables and biomarkers as predictors of hypertensive heart failure in 10,002 study participants in an 8.8‐year follow‐up study of the METSIM cohort
Table S6. Cox regression analysis of the principal components shown in Table 2 as predictors of hypertensive HF in 10,002 study participants in an 8.8‐year follow‐up study of the METSIM cohort.
Jauhiainen, R. , Jauhiainen, M. , Vangipurapu, J. , Kuulasmaa, T. , Ala‐Korpela, M. , Laakso, M. , and Kuusisto, J. (2021) Novel biomarkers associated with incident heart failure in 10 106 Finnish men. ESC Heart Failure, 8: 605–614. 10.1002/ehf2.13132.
Raimo Jauhiainen and Matti Jauhiainen contributed equally to this work.
References
- 1. Braunwald E. Heart failure. JACC Heart Fail 2013; 1: 1–20. [DOI] [PubMed] [Google Scholar]
- 2. Bui AL, Horwich TB, Fonarow GC. Epidemiology and risk profile of heart failure. Nat Rev Cardiol 2011; 8: 30–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Braunwald E. The war against heart failure: the Lancet lecture. Lancet 2015; 385: 812–824. [DOI] [PubMed] [Google Scholar]
- 4. McMurray JJ, Adamopoulos S, Anker SD, Auricchio A, Böhm M, Dickstein K, Falk V, Filippatos G, Fonseca C, Gomez‐Sanchez MA, Jaarsma T, Køber L, Lip GY, Maggioni AP, Parkhomenko A, Pieske BM, Popescu BA, Rønnevik PK, Rutten FH, Schwitter J, Seferovic P, Stepinska J, Trindade PT, Voors AA, Zannad F, Zeiher A, Task Force for the Diagnosis and Treatment of Acute and Chronic Heart Failure 2012 of the European Society of Cardiology , Bax JJ, Baumgartner H, Ceconi C, Dean V, Deaton C, Fagard R, Funck‐Brentano C, Hasdai D, Hoes A, Kirchhof P, Knuuti J, Kolh P, McDonagh T, Moulin C, Popescu BA, Reiner Z, Sechtem U, Sirnes PA, Tendera M, Torbicki A, Vahanian A, Windecker S, McDonagh T, Sechtem U, Bonet LA, Avraamides P, Ben Lamin HA, Brignole M, Coca A, Cowburn P, Dargie H, Elliott P, Flachskampf FA, Guida GF, Hardman S, Iung B, Merkely B, Mueller C, Nanas JN, Nielsen OW, Orn S, Parissis JT, Ponikowski P, ESC Committee for Practice Guidelines . ESC guidelines for the diagnosis and treatment of acute and chronic heart failure 2012: the task force for the diagnosis and treatment of acute and chronic heart failure 2012 of the European Society of Cardiology. Developed in collaboration with the Heart Failure Association (HFA) of the ESC [published correction appears in Eur Heart J. 2013 Jan;34(2):158]. Eur Heart J 2012; 33: 1787–1847. [DOI] [PubMed] [Google Scholar]
- 5. Glick D, de Filippi CR, Christenson R, Gottdiener JS, Seliger SL. Long‐term trajectory of two unique cardiac biomarkers and subsequent left ventricular structural pathology and risk of incident heart failure in community‐dwelling older adults at low baseline risk. JACC Heart Fail 2013; 1: 353–360. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Natriuretic Peptides Studies Collaboration . Natriuretic peptides and integrated risk assessment for cardiovascular disease: an individual‐participant‐data meta‐analysis. Lancet Diabetes Endocrinol 2016; 4: 840–849. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Bayes‐Genis A, Voors AA, Zannad F, Januzzi JL, Mark Richards A, Díez J. Transitioning from usual care to biomarker‐based personalized and precision medicine in heart failure: call for action. Eur Heart J 2018; 39: 2793–2799. [DOI] [PubMed] [Google Scholar]
- 8. Cesari M, Penninx BW, Newman AB, Kritchevsky SB, Nicklas BJ, Sutton‐Tyrrell K, Rubin SM, Ding J, Simonsick EM, Harris TB, Pahor M. Inflammatory markers and onset of cardiovascular events: results from the Health ABC study. Circulation 2003; 108: 2317–2322. [DOI] [PubMed] [Google Scholar]
- 9. Frankel DS, Vasan RS, D'Agostino RB Sr, Benjamin EJ, Levy D, Wang TJ, Meigs JB. Resistin, adiponectin, and risk of heart failure the Framingham offspring study. J Am Coll Cardiol 2009; 53: 754–762. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Zheng Y, Yu B, Alexander D, Manolio TA, Aguilar D, Coresh J, Heiss G, Boerwinkle E, Nettleton JA. Associations between metabolomic compounds and incident heart failure among African Americans: the ARIC Study. Am J Epidemiol 2013; 178: 534–542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Lindberg S, Jensen JS, Bjerre M, Pedersen SH, Frystyk J, Flyvbjerg A, Mogelvang R. Cardio‐adipose tissue cross‐talk: relationship between adiponectin, plasma pro brain natriuretic peptide and incident heart failure. Eur J Heart Fail 2014; 16: 633–638. [DOI] [PubMed] [Google Scholar]
- 12. Karas MG, Benkeser D, Arnold AM, Bartz TM, Djousse L, Mukamal KJ, Ix JH, Zieman SJ, Siscovick DS, Tracy RP, Mantzoros CS, Gottdiener JS, de Filippi CR, Kizer JR. Relations of plasma total and high‐molecular‐weight adiponectin to new‐onset heart failure in adults ≥65 years of age (from the Cardiovascular Health study). Am J Cardiol 2014; 113: 328–334. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Delles C, Rankin NJ, Boachie C, McConnachie A, Ford I, Kangas A, Soininen P, Trompet S, Mooijaart SP, Jukema JW, Zannad F, Ala‐Korpela M, Salomaa V, Havulinna AS, Welsh P, Würtz P, Sattar N. Nuclear magnetic resonance‐based metabolomics identifies phenylalanine as a novel predictor of incident heart failure hospitalisation: results from PROSPER and FINRISK 1997. Eur J Heart Fail 2018; 20: 663–673. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. de Boer RA, Nayor M, de Filippi CR, Enserro D, Bhambhani V, Kizer JR, Blaha MJ, Brouwers FP, Cushman M, Lima J, Bahrami H, van der Harst P, Wang TJ, Gansevoort RT, Fox CS, Gaggin HK, Kop WJ, Liu K, Vasan RS, Psaty BM, Ho JE. Association of cardiovascular biomarkers with incident heart failure with preserved and reduced ejection fraction. JAMA Cardiol 2018; 3: 215–224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. AlBadri A, Lai K, Wei J, Landes S, Mehta PK, Li Q, Johnson D, Reis SE, Kelsey SF, Bittner V, Sopko G, Shaw LJ, Pepine CJ, Bairey Merz CN. Inflammatory biomarkers as predictors of heart failure in women without obstructive coronary artery disease: a report from the NHLBI‐sponsored Women's Ischemia Syndrome Evaluation (WISE). PLoS ONE 2017; 12: e0177684 Published 2017 May 19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Lin LY, Chun Chang S, O'Hearn J, Hui ST, Seldin M, Gupta P, Bondar G, Deng M, Jauhiainen R, Kuusisto J, Laakso M, Sinsheimer JS, Deb A, Rau C, Ren S, Wang Y, Lusis AJ, Wang JJ, Huertas‐Vazquez A. Systems genetics approach to biomarker discovery: GPNMB and heart failure in mice and humans. G3 (Bethesda) 2018; 8: 3499–3506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Bahrami H, Bluemke DA, Kronmal R, Bertoni AG, Lloyd‐Jones DM, Shahar E, Szklo M, Lima JA. Novel metabolic risk factors for incident heart failure and their relationship with obesity: the MESA (Multi‐Ethnic Study of Atherosclerosis) study. J Am Coll Cardiol 2008; 51: 1775–1783. [DOI] [PubMed] [Google Scholar]
- 18. Stancáková A, Javorský M, Kuulasmaa T, Haffner SM, Kuusisto J, Laakso M. Changes in insulin sensitivity and insulin release in relation to glycemia and glucose tolerance in 6,414 Finnish men. Diabetes 2009; 58: 1212–1221. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Laakso M, Kuusisto J, Stančáková A, Kuulasmaa T, Pajukanta P, Lusis AJ, Collins FS, Mohlke KL, Boehnke M. The metabolic syndrome in men study: a resource for studies of metabolic and cardiovascular diseases. J Lipid Res 2017; 58: 481–493. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. American Diabetes Association . Diagnosis and classification of diabetes mellitus. Diabetes Care 2013; 36: S67–S74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Fizelova M, Jauhiainen R, Kangas AJ, Soininen P, Ala‐Korpela M, Kuusisto J, Laakso M, Stancáková A. Differential associations of inflammatory markers with insulin sensitivity and secretion: the prospective METSIM study. J Clin Endocrinol Metab 2017; 102: 3600–3609. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Soininen P, Kangas AJ, Würtz P, Suna T, Ala‐Korpela M. Quantitative serum nuclear magnetic resonance metabolomics in cardiovascular epidemiology and genetics. Circ Cardiovasc Genet 2015; 8: 192–206. [DOI] [PubMed] [Google Scholar]
- 23. Inouye M, Kettunen J, Soininen P, Silander K, Ripatti S, Kumpula LS, Hämäläinen E, Jousilahti P, Kangas AJ, Männistö S, Savolainen MJ, Jula A, Leiviskä J, Palotie A, Salomaa V, Perola M, Ala‐Korpela M, Peltonen L. Metabonomic, transcriptomic, and genomic variation of a population cohort. Mol Syst Biol 2010; 6: 441. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Das UN. Pyruvate is an endogenous anti‐inflammatory and anti‐oxidant molecule. Med Sci Monit 2006; 12: RA79–RA84. [PubMed] [Google Scholar]
- 25. Kleeberg J, Gitelson S. A study of the blood pyruvic acid level in patients with congestive heart failure. J Clin Pathol 1954; 7: 116–121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Lommi J, Kupari M, Koskinen P, Naveri H, Leinonen H, Pulkki K, Harkonen M. Blood ketone bodies in congestive heart failure. J Am Coll Cardiol 1996; 28: 665–672. [DOI] [PubMed] [Google Scholar]
- 27. Stancáková A, Civelek M, Saleem NK, Soininen P, Kangas AJ, Cederberg H, Paananen J, Pihlajamäki J, Bonnycastle LL, Morken MA, Boehnke M, Pajukanta P, Lusis AJ, Collins FS, Kuusisto J, Ala‐Korpela M, Laakso M. Hyperglycemia and a common variant of GCKR are associated with the levels of eight amino acids in 9,369 Finnish men. Diabetes 2012; 61: 1895–1902. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Lempiainen P, Mykkanen L, Pyorala K, Laakso M, Kuusisto J. Insulin resistance syndrome predicts coronary heart disease events in elderly nondiabetic men. Circulation 1999; 100: 123–128. [DOI] [PubMed] [Google Scholar]
- 29. Heymans S, Hirsch E, Anker SD, Aukrust P, Balligand JL, Cohen‐Tervaert JW, Drexler H, Filippatos G, Felix SB, Gullestad L, Hilfiker‐Kleiner D, Janssens S, Latini R, Neubauer G, Paulus WJ, Pieske B, Ponikowski P, Schroen B, Schultheiss HP, Tschöpe C, Shah AM. Inflammation as a therapeutic target in heart failure? A scientific statement from the Translational Research Committee of the Heart Failure Association of the European Society of Cardiology. Eur J Heart Fail 2009; 11: 119–129. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Doehner W, Frenneaux M, Anker SD. Metabolic impairment in heart failure: the myocardial and systemic perspective. J Am Coll Cardiol 2014; 64: 1388–1400. [DOI] [PubMed] [Google Scholar]
- 31. Ritterhoff J, Tian R. Metabolism in cardiomyopathy: every substrate matters. Cardiovasc Res 2017; 113: 411–421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Cohen JB, Schrauben SJ, Zhao L, Basso MD, Cvijic ME, Li Z, Yarde M, Wang Z, Bhattacharya PT, Chirinos DA, Prenner S, Zamani P, Seiffert DA, Car BD, Gordon DA, Margulies K, Cappola T, Chirinos JA. Clinical phenogroups in heart failure with preserved ejection fraction: detailed phenotypes, prognosis, and response to spironolactone. JACC Heart Fail 2020; 8: 172–184. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. Baseline anthropometric, metabolic, and inflammatory marker measurements of the study participants without and with incident heart failure
Table S2. Serum metabolic measures in proton NMR analysis at baseline in the participants with and without incident heart failure during the METSIM study follow‐up
Table S3. Clinical characteristics and outcomes of the METSIM study participants with incident heart failure.
Table S4. Clinical characteristics and outcomes by the etiology of heart failure in the METSIM study participants with incident heart failure
Table S5. Cox regression analysis of demographic and anthropometric variables and biomarkers as predictors of hypertensive heart failure in 10,002 study participants in an 8.8‐year follow‐up study of the METSIM cohort
Table S6. Cox regression analysis of the principal components shown in Table 2 as predictors of hypertensive HF in 10,002 study participants in an 8.8‐year follow‐up study of the METSIM cohort.