Association Between Endothelial Alterations, Cardiac Function, and Outcomes From Health to Heart Failure: Insight From the STANISLAS, MEDIA‐DHF, and BIOSTAT‐CHF Cohorts

Abstract

Background

Heart failure with preserved ejection fraction (HFpEF) is a multifaceted syndrome, likely stemming from comorbidity‐induced inflammation resulting in endothelial dysfunction. Endothelial glycocalyx degradation's role in the development and prognosis of HFpEF remains largely unexplored. Our study aimed at exploring the association between glycocalyx degradation and diastolic dysfunction and determining whether glycocalyx degradation can predict clinical outcomes in patients with HFpEF.

Methods

Perlecan and thrombomodulin concentrations were assessed in individuals deemed healthy (STANISLAS [Suivi Temporaire Annuel Non‐Invasif de la Santé des Lorrains Assurés Sociaux (Annual Noninvasive Temporary Monitoring of the Health of Insured Lorrainers)] cohort, n=1705) and patients with HFpEF (MEDIA‐DHF [Metabolic Road to Diastolic Heart Failure], n=460 and BIOSTAT‐CHF [Biology Study to Tailored Treatment in Chronic Heart Failure], n=556) to evaluate endothelial glycocalyx degradation.

Results

In patients with HFpEF, perlecan but not thrombomodulin was increased compared with controls (P<0.0001 versus P=0.73). In adjusted analysis, perlecan was associated with peak early mitral inflow velocity/peak early diastolic mitral annular velocity ratio and thrombomodulin with peak early diastolic mitral annular velocity in control individuals, whereas perlecan and thrombomodulin were associated with peak early mitral inflow velocity/peak early diastolic mitral annular velocity and left atrial volume index in patients with HFpEF (all P<0.03). Perlecan was significantly associated with cardiovascular hospitalization and death in the MEDIA‐DHF (adjusted hazard ratio [HR] for highest tertile versus first tertile, 2.44 [95% CI, 1.11–5.34]; P=0.026) and BIOSTAT‐CHF cohorts (adjusted HR, 2.12 [95% CI, 1.49–3.03]; P<0.0001). Thrombomodulin was associated with a worse outcome in BIOSTAT‐CHF (P=0.004) but not in MEDIA‐DHF.

Conclusions

Higher circulating levels of the endothelial glycocalyx degradation biomarkers like perlecan and, to a lesser extent, thrombomodulin are associated with features of diastolic dysfunction in population and HFpEF settings and predict poor outcome in patients with HFpEF. These results suggest that glycocalyx degradation may be an early step in the pathological processes leading to HFpEF and gain further prognostic value in later stages (ie, overt HFpEF).

Registration

URL: https://clinicaltrials.gov/; Unique identifiers: NCT01391442, https://clinicaltrials.gov/study/NCT01391442?cond=stanislas&rank=1; NCT02446327; URL: https://cordis.europa.eu; BIOSTAT‐CHF ID: 242209.

Nonstandard Abbreviations and Acronyms

Bio‐SHiFT

Serial Biomarker Measurements and New Echocardiographic Techniques in Chronic Heart Failure Patients Result in Tailored Prediction of Prognosis

BIOSTAT‐CHF

Biology Study to Tailored Treatment in Chronic Heart Failure

CVD

cardiovascular death

CVH

cardiovascular health

DIAST‐CHF

Diagnostic Study on Prevalence and Clinical Course of Diastolic Dysfunction and Diastolic Heart Failure

e′

peak early diastolic mitral annular velocity

E

peak early mitral inflow velocity

HFpEF

heart failure with preserved ejection fraction

HFrEF

heart failure with reduced ejection fraction

IPW

inverse probability weight

LAVi

left atrial volume index

LVMi

left ventricular mass index

MEDIA‐DHF

Metabolic Road to Diastolic Heart Failure

NPX

normalized protein expression

PASP

pulmonary artery systolic pressure

STANISLAS

Suivi Temporaire Annuel Non‐Invasif de la Santé des Lorrains Assurés Sociaux (Annual Noninvasive Temporary Monitoring of the Health of Insured Lorrainers)

Clinical Perspective.

What Is New

• Endothelial glycocalyx degradation, assessed by plasma perlecan levels, is higher in patients with heart failure with preserved ejection fraction compared with control individuals from a populational cohort, even after adjusting for age, sex, and body mass index.

• Features of diastolic dysfunction are increased in individuals with higher circulating levels of perlecan and thrombomodulin both in the populational cohort and in patients with heart failure with preserved ejection fraction.

What Are the Clinical Implications

• Endothelial glycocalyx degradation is part of endothelial dysfunction development and glycocalyx degradation biomarkers have been reported to be higher in heart failure with preserved ejection fraction versus heart failure with reduced ejection fraction.

• Higher levels of perlecan and thrombomodulin are correlated with a significantly increased risk of hospitalization and death in patients with heart failure with preserved ejection fraction even after adjusting for baseline differences.

Heart failure (HF) with preserved ejection fraction (HFpEF) is a clinical syndrome with multiple origins and whose pathophysiology is complex and without targeted neurohormonal signaling pathway therapy, unlike HF with reduced ejection fraction (HFrEF). ¹ The increasing prevalence of HFpEF and its poor prognosis underscores the importance of understanding the factors contributing to the heterogeneity of the syndrome and its underlying mechanisms. ² While several studies have attempted to identify distinct HFpEF subgroups with specific prognoses based on clinical, echocardiographic and hemodynamic characteristics and biomarkers, ³ , ⁴ , ⁵ no single echocardiographic parameter or plasma biomarker to date is sufficiently accurate and reproducible for stratifying risk in patients with HFpEF. ⁶

In the early 2010s, Paulus and Tschöpe proposed a paradigm shift in the pathophysiology of HFpEF in which HFpEF development arises from a comorbidities‐related proinflammatory state. Because of this continuous low‐grade chronic inflammation, nitric oxide bioavailability in coronary microvascular endothelium decreases leading to increased cardiomyocyte stiffness, interstitial fibrosis development and, ultimately, pathological HF. ⁷ HFrEF predominantly arises from oxidative stress following ischemic injury in cardiomyocytes. In contrast, the systemic inflammatory state driven by comorbidities has been identified as a predictive factor for HFpEF. ⁸

Endothelial cells are protected from aggression by circulating cells by a layer of proteoglycans called the “endothelial glycocalyx.” ⁹ The endothelial glycocalyx is a specialized extracellular matrix composed of proteoglycans, glycosaminoglycan, proteins anchored to the endothelial cell membrane and absorbed plasma proteins. The endothelial glycocalyx thickness ranges from 0.2 to 4.5 μm. ¹⁰ , ¹¹ Endothelial glycocalyx degradation occurs in settings of inflammatory or tissue injury and trauma and is an early step in the development of endothelial dysfunction. ⁹ , ¹² Most of the data on glycocalyx status in the field of HF has been gathered from studies on patients with HFrEF and compared with healthy individuals or patients with HFpEF. ¹³ , ¹⁴ While current research on the transition from health to HFpEF has focused on hemodynamics, inflammation, and oxidative stress, ¹⁴ , ¹⁵ there is a notable lack of evidence regarding the role and significance of glycocalyx degradation in this process.

Thrombomodulin is a membrane protein with anticoagulant properties expressed on the surface of endothelial cells and protected from cleavage by the glycocalyx layer. ¹⁶ Proteolysis cleavage of thrombomodulin leads to its shedding and release in the circulation. Over the past years, soluble thrombomodulin has frequently been used as a biomarker of endothelial glycocalyx degradation. Soluble thrombomodulin was reported to be increased in acute conditions such as sepsis, COVID‐19, immune‐mediated thrombotic thrombocytopenic purpura or coronary artery bypass grafting. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ Other glycocalyx components such as perlecan and decorin are currently used as new glycocalyx degradation markers. Decorin is a leucine‐rich proteoglycan embedded in the endothelial glycocalyx able to regulate collagen fibril assembly and cell function such as adhesion or angiogenesis. ²¹ Decorin was recently found to be increased in chronic kidney disease. ²² Perlecan is ubiquitously synthesized in basal membranes and exists as a large heparan sulfate proteoglycan or heparan sulfate–chondroitin sulfate complex. ²³ However, endothelial cells synthesize this proteoglycan as a mono‐substituted heparan sulfate. ²⁴ Despite its exposure and release occurring on the basal side of endothelial cells, perlecan can be released into the circulation. Circulating perlecan may then be captured by the endothelial glycocalyx and become part of this layer similar to other non–membrane‐bound glycosaminoglycans (eg, hyaluronan). When bound to the glycocalyx, perlecan acts as a flow sensor for endothelial cells, modulating membrane polarization and extracellular matrix remodeling. ²⁵ , ²⁶ Perlecan has been relatively understudied compared with other biomarkers of the glycocalyx regarding its role in the endothelial environment and its changes during diseases, suggesting its potential as a novel biomarker.

Understanding the role of glycocalyx degradation, as an early indicator of endothelial dysfunction, in the development of HFpEF could hold prognostic significance and support the proposal of early therapeutic interventions. In light of the above, the present study aimed to compare plasma concentrations of thrombomodulin and perlecan between a populational cohort (the STANISLAS [Suivi Temporaire Annuel Non‐Invasif de la Santé des Lorrains Assurés Sociaux (Annual Noninvasive Temporary Monitoring of the Health of Insured Lorrainers) cohort]: a familial longitudinal population‐based cohort from the Nancy region of France:) and patients with HFpEF (the MEDIA‐DHF (Metabolic Road to Diastolic Heart Failure study cohort) and their association with diastolic dysfunction. In addition, the association between perlecan, thrombomodulin, and outcomes was also investigated in patients with HFpEF from the MEDIA‐DHF and BIOSTAT‐CHF (Biology Study to Tailored Treatment in Chronic Heart Failure) cohorts.

METHODS

Data Availability

The data underlying this article will be shared upon reasonable request to the corresponding author.

Study Populations

The STANISLAS cohort design has been previously described in detail. ²⁷ Briefly, this family‐based longitudinal cohort included 4598 healthy individuals from 1006 French families in the Lorraine region in northeastern France. Participants were examined every 5 to 10 years over a period of 20 years. The fourth visit included a total of 1705 participants between 2011 and 2016 and consisted of a medical examination and an interview. The study protocols for all examinations were reviewed and approved by the local ethics committee of the Comité de Protection des Personnes Est 3, France. All participants provided written informed consent to participate in the study.

The MEDIA‐DHF cohort is a multicenter (10 centers), multinational, observational study that enrolled 626 patients between 2012 and 2014. ²⁸ Patients were included following a diagnosis of left ventricular diastolic dysfunction using 2007 European Society of Cardiology consensus recommendations. ²⁹ MEDIA‐DHF patients represented (1) patients with acute decompensated HF, (2) patients recently discharged after admission for an acute HF episode (<60 days), or (3) ambulatory patients with chronic disease. Key inclusion criteria were (1) signs and symptoms of HF with a preserved (ie, ≥50%) left ventricular (LV) ejection fraction and an LV end‐diastolic volume index <97 mL/m² and (2) elevated serum concentration of brain natriuretic peptide >100 pg/mL or NT‐proBNP (N‐terminal pro‐B‐type natriuretic peptide) >300 pg/mL. The primary outcome was time to the composite of cardiovascular death (CVD) or first HF hospitalization. Secondary outcomes were CVD and hospitalization for syncope. Follow‐up assessments were performed at 3, 6, and 12 months after inclusion. The study protocol complied with the Declaration of Helsinki and was approved by the respective ethics committees of the participating establishments. All patients provided written informed consent (ClinicalTrials.gov identifier: NCT02446327). In the present analysis, only patients with available biomarkers (N=460, no patients with acute decompensation) were included (Figure S1).

The BIOSTAT‐CHF cohorts (index, n=2516; and validation, n=1738) ³⁰ were from a multicenter study that enrolled patients from 11 European countries. The BIOSTAT‐CHF project included the index cohort to create a risk score for nonresponse to therapy defined as death or HF hospitalization. The score was then validated in the validation cohort. Participants were aged ≥18 years and had symptoms of new‐onset or worsening HF, combined with an LV ejection fraction ≤40% or BNP and, or NT‐proBNP plasma levels >400 or >2000 pg/mL, respectively. Participants in the validation cohort were aged ≥18 years, were diagnosed with HF, and had a previous admission for HF requiring diuretic treatment. The primary outcome was a composite of all‐cause death and rehospitalization for HF censored at 2‐year follow‐up, with each individual component constituting a secondary outcome. The study was performed according to the Declaration of Helsinki. The study protocol was approved by local and national ethics committees (EudraCT 2010‐020808‐29; R&D Reference No. 2008‐CA03; MREC No. 10/S1402/39), and all participants provided written informed consent. Analysis of our study were conducted in patients with HFpEF from the pooled index and validation cohorts who had perlecan, thrombomodulin, or decorin available (n=131 in index cohort, 425 in validation). Analyses were stratified for cohort (index, validation) in BIOSTAT‐CHF. Perlecan, thrombomodulin, and decorin were standardized separately according to the Z score within each cohort before pooling of the data.

Biomarker Measurements

Plasma samples obtained at inclusion in each study were analyzed for protein biomarkers (including perlecan, thrombomodulin, and decorin for the 3 studies) using different Olink Proseek Multiplex panels: cardiovascular disease II and cardiovascular disease III (Olink Proteomics, Uppsala, Sweden). Perlecan is included in the cardiovascular disease III panel, and thrombomodulin and decorin are included in the cardiovascular disease II panel. For this assay, a proximity extension assay technology is used in which 92 oligonucleotide‐labeled antibody probe pairs per panel may bind to their respective targets in a 1‐μL plasma sample. ³¹ Results are expressed in normalized protein expression (NPX), which is in log₂ scale arbitrary units where an elevated value corresponds to a higher protein expression. Olink measurements were performed in 2 batches for MEDIA‐DHF and STANISLAS but were subsequently harmonized using bridging samples from both cohorts. A normalization strategy (OlinkAnalyze package) was applied to ensure direct comparability between the cohorts.

Statistical Analysis

Continuous variables are expressed as mean±SD or median (interquartile range) according to their distributions; categorical variables are expressed as frequency (percentage), and β, odds ratio, and hazard ratio (HR) as estimate and 95% CI. Comparisons of patient characteristics were carried out using ANOVA or Kruskal–Wallis tests and χ² test or Fisher's exact test as appropriate. A propensity score analysis was conducted for comparison of biomarkers of endothelial glycocalyx degradation (perlecan/thrombomodulin) between the STANISLAS and MEDIA‐DHF cohorts. Inverse probability weights (IPWs) and stabilized weights were computed from a logistic regression model with the study (STANISLAS versus MEDIA‐DHF) as outcome and sex, age, and body mass index as explanatory variables. ³² In the propensity score analysis, patients from the STANISLAS cohort aged 60 to 80 years were included to match the age of MEDIA‐DHF cohort patients (Table S1). Relationships between cardiac and vascular echocardiography parameters as outcomes and biomarkers of glycocalyx degradation as explanatory variables were examined using linear regression. Univariable‐ and multivariable‐adjusted (sex, age, body mass index), diabetes, hypertension, and estimated glomerular filtration rate) models were considered. Linear model validity assumptions were verified. Univariable and multivariable logistic regression models were used to assess the association between endothelial glycocalyx degradation biomarkers and binary cardiac and vascular echography parameter outcomes. All of the above analyses were performed in both the STANILAS and MEDIA‐DHF cohorts. In addition, Cox regression analyses were conducted in the MEDIA‐DHF and BIOSTAT‐CHF cohorts to assess the association between glycocalyx degradation biomarkers and the composite end points of CVD or cardiovascular hospitalization (CVH) and death or CVH. The proportionality assumption was verified by testing the covariate–time interaction terms. Kaplan–Meier curves were plotted to illustrate the risk of cardiovascular end points (CVH or CVD, death or CVH) in participant subsets defined according to tertiles of biomarkers of glycocalyx degradation. Crude and adjusted HRs with their respective CIs are reported.

All analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC) and R software version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria). Two‐sided P values <0.05 were considered statistically significant.

RESULTS

Baseline Characteristics

The following cutoffs were implemented for thrombomodulin and perlecan tertiles on the basis of the distribution of biomarkers of glycocalyx degradation in the STANISLAS cohort and MEDIA‐DHF cohort: (1) perlecan <6.96 NPX (n=550), 6.96 to 7.20 (n=560), >7.20 (n=544), and thrombomodulin <10.18 NPX (n=544), 10.18 to 10.42 (n=561), >10.42 (n=549) in the STANISLAS cohort; and (2) perlecan <6.41 NPX (n=151), 6.41 to 6.82 (n=157), >6.82 (n=152) and thrombomodulin <8.49 NPX (n=150), 8.49 to 8.82 (n=155), >8.82 (n=151) in the MEDIA‐DHF cohort. Participants with elevated thrombomodulin levels in the STANISLAS cohort (Table 1) were older and more prone to be men and former smokers. They were more likely to present metabolic abnormalities (overweight, elevated low‐density lipoprotein, plasma glucose) as well as altered kidney function (lower estimated glomerular filtration rate). Cohort participants with higher thrombomodulin had more lipid‐lowering agents and angiotensin receptor blockers. Cardiovascular function was also impaired: elevated systolic blood pressure (P<0.0001), left ventricular mass index (LVMi) (P<0.0001), peak early mitral inflow velocity/e' (peak early diastolic mitral annular velocity)ratio (E/e′) (P=0.006), and E' (P<0.0001). Similar changes were observed for perlecan with the addition of dyslipidemia and elevated triglycerides for the metabolic parameters. Cohort participants with higher perlecan had more treatments such as lipid‐lowering agents, angiotensin‐converting enzyme inhibitors, angiotensin receptor blockers, β blocker diuretics, calcium channel blockers, antiplatelets, anticoagulants, antidiabetic agents, and statins. Impairment of cardiovascular function also included significant treated hypertension history, left atrial volume index (LAVi), and left ventricular hypertrophy (Table 1).

Table 1.

Characteristics of Control Participants (STANISLAS Populational Cohort) According to Perlecan and Thrombomodulin (Olink) Tertiles

	Perlecan (NPX)				Thrombomodulin (NPX)
	Tertile 1	Tertile 2	Tertile 3	P value	Tertile 1	Tertile 2	Tertile 3	P value
	(5.81–6.96)	(6.96–7.20)	(7.20–8.53)		(7.48–10.18)	(10.18–10.42)	(10.42–11.43)
	(n=550)	(n=560)	(n=544)		(n=544	(n=561)	(n=549)
Clinical data
Age, y	40 (33–58)	55 (34–60)	58 (43–62)	<0.0001	41 (33–58)	56 (35–60)	57 (38–62)	<0.0001
Female sex	320 (58.2)	273 (48.8)	264 (48.5)	0.001	367 (67.5)	277 (49.4)	213 (38.8)	<0.0001
Body mass index, kg/m2	23.7 (21.5–26.5)	25.1 (22.9–28.2)	26.7 (23.9–29.8)	<0.0001	24.5 (22.0–28.1)	25.2 (22.5–28.8)	25.5 (23.1–28.2)	0.004
Smoking status
Current smoker	151 (27.5)	118 (21.1)	86 (15.9)		140 (25.7)	117 (21.0)	98 (17.9)
Former smoker	150 (27.3)	189 (33.9)	191 (35.2)	<0.0001	153 (28.1)	188 (33.7)	189 (34.5)	0.015
Never smoked	249 (45.3)	251 (45.0)	265 (48.9)		251 (46.1)	253 (45.3)	261 (47.6)
HR, bpm	64.0±9.1	63.6±9.5	63.7±9.5	0.72	64.3±9.0	63.8±9.8	63.1±9.2	0.096
Office SBP, mm Hg	122.3±14.3	125.5±14.9	128.8±16.1	<0.0001	122.2±14.2	126.1±15.5	128.3±15.5	<0.0001
Medical history
Treated hypertension	66 (12.1)	93 (16.7)	144 (27.0)	<0.0001	84 (15.6)	108 (19.5)	111 (20.4)	0.091
Stroke or TIA	5 (0.9)	5 (0.9)	12 (2.2)	0.13	7 (1.3)	6 (1.1)	9 (1.6)	0.69
Peripheral artery disease	3 (0.5)	3 (0.5)	2 (0.4)	1.00	3 (0.6)	2 (0.4)	3 (0.5)	0.83
Dyslipidemia	107 (19.5)	120 (21.4)	160 (29.4)	0.0002	118 (21.7)	128 (22.8)	141 (25.7)	0.28
Biology
Plasma glucose, g/L	0.87 (0.81–0.94)	0.88 (0.82–0.95)	0.89 (0.84–0.97)	0.0004	0.87 (0.81–0.94)	0.89 (0.83–0.95)	0.89 (0.83–0.96)	0.001
Hemoglobin A1c, %	5.5 (5.3–5.7)	5.5 (5.3–5.8)	5.6 (5.4–5.9)	<0.0001	5.5 (5.3–5.8)	5.6 (5.4–5.8)	5.6 (5.4–5.8)	0.002
Triglycerides, g/L	0.85 (0.62–1.15)	0.87 (0.67–1.22)	1.04 (0.75–1.43)	<0.0001	0.90 (0.65–1.24)	0.92 (0.67–1.28)	0.92 (0.70–1.26)	0.64
HDL, g/L	0.58±0.14	0.58±0.14	0.57±0.14	0.35	0.58±0.13	0.58±0.14	0.57±0.14	0.22
LDL, g/L	1.28±0.33	1.35±0.33	1.39±0.35	<0.0001	1.31±0.33	1.34±0.35	1.37±0.34	0.008
Creatinine, mg/L	7.65±1.38	8.13±1.36	8.71±1.74	<0.0001	7.59±1.32	8.21±1.46	8.67±1.69	<0.0001
eGFR, mL/min per 1.73 m2	102.7v13.4	97.7±13.8	88.9±15.2	<0.0001	101.3±13.9	95.8±15.1	92.4±15.3	<0.0001
Echocardiographic variables
LVEF, %	64.9±6.1	64.9±6.2	65.3±6.8	0.40	65.3±6.2	64.7±6.5	65.1±6.5	0.35
LVMi, g/m2	73.1±17.5	75.7±18.3	79.4±21.0	<0.0001	72.9±18.5	76.9±19.8	78.3±18.9	<0.0001
LAVi, mL/m2	22.4±6.5	22.3±7.0	23.5±8.2	0.011	22.4±6.8	23.0±7.6	22.7±7.5	0.42
E′, cm/s	12.31±3.28	11.45±3.16	10.43±2.91	<0.0001	12.13±3.29	11.29±3.29	10.80±2.91	<0.0001
E/e′	6.10±1.62	6.33±1.78	6.95±1.95	<0.0001	6.26±1.75	6.51±1.96	6.60±1.73	0.006
LVH	56 (10.3)	76 (13.7)	102 (19.2)	0.0002	62 (11.5)	88 (15.9)	84 (15.7)	0.059
Treatment
Lipid‐lowering agents	65 (11.8)	69 (12.3)	116 (21.3)	<0.0001	59 (10.8)	87 (15.5)	104 (18.9)	0.0008
AvCE inhibitors	16 (2.9)	16 (2.9)	39 (7.2)	0.0005	22 (4.0)	23 (4.1)	26 (4.7)	0.83
ARB	26 (4.7)	43 (7.7)	77 (14.2)	<0.0001	31 (5.7)	55 (9.8)	60 (10.9)	0.004
β Blockers	32 (5.8)	37 (6.6)	61 (11.2)	0.002	42 (7.7)	47 (8.4)	41 (7.5)	0.85
Diuretics	20 (3.6)	31 (5.5)	63 (11.6)	< 0.0001	29 (5.3)	41 (7.3)	44 (8.0)	0.19
Calcium channel blockers	22 (4.0)	24 (4.3)	48 (8.8)	0.0009	23 (4.2)	35 (6.2)	36 (6.6)	0.18
Antiplatelets	24 (4.4)	27 (4.8)	47 (8.6)	0.006	27 (5.0)	33 (5.9)	38 (6.9)	0.40
Anticoagulants	28 (5.1)	27 (4.8)	56 (10.3)	0.0004	31 (5.7)	36 (6.4)	44 (8.0)	0.30
Antidiabetic agents	20 (3.6)	11 (2.0)	30 (5.5)	0.007	18 (3.3)	21 (3.7)	22 (4.0)	0.83
Statin	52 (9.5)	64 (11.4)	88 (16.2)	0.003	53 (9.7)	76 (13.5)	75 (13.7)	0.074

Data are expressed as n (%), mean±SD, or median (interquartile range). P values are obtained from ANOVA or Kruskal–Wallis tests (continuous variables) and χ² or Fisher's exact tests (categorical variables). ACE indicates angiotensin‐converting enzyme; ARB, angiotensin receptor blocker; E/e′, peak early mitral inflow velocity/ peak early diastolic mitral annular velocity ratio; eGFR, estimated glomerular filtration rate; HDL, high‐density lipoprotein; HR, heart rate; LAVi, left atrial volume index; LDL, low‐density lipoprotein; LVEF, left ventricular ejection fraction; LVH, left ventricular hypertrophy; LVMi, left ventricular mass index; SBP, systolic blood pressure; STANISLAS, Suivi Temporaire Annuel Non‐Invasif de la Santé des Lorrains Assurés Sociaux (Annual Noninvasive Temporary Monitoring of the Health of Insured Lorrainers); and TIA, transient ischemic attack.

In patients with HFpEF (Table 2), higher levels of thrombomodulin were correlated with age, former tobacco use, metabolic abnormalities (increased triglycerides), and kidney dysfunction (lower estimated glomerular filtration rate). Patients with higher thrombomodulin concentrations had more diuretics and calcium channel blockers. For cardiovascular parameters, thrombomodulin levels were correlated with previous hospitalization for HF, LAVi, E′, and pulmonary artery systolic pressure (Table 2), while perlecan levels were correlated with increases in low‐density lipoprotein and chronic obstructive pulmonary disease in addition to the differences observed for thrombomodulin. Patients with higher perlecan concentrations had more β blockers, diuretics, and anticoagulants. Regarding cardiovascular function, higher systolic blood pressure, LVMi, LAVi, E/e′, pulmonary artery systolic pressure and left ventricular hypertrophy were all correlated with higher perlecan concentrations (Table 2). In addition to these associations, thrombomodulin and perlecan were further compared between the 2 cohorts to determine whether or not they were elevated in the presence of HFpEF compared with control populational participants (Figure 1). STANISLAS participants aged between 60 and 80 years were compared with MEDIA‐DHF patients and age, sex, and body mass index effects were accounted for via IPW leading to nonsignificant differences with respect to these variables (Table S1). Of note, the age of the IPW cohorts was similar (66±4.8 for STANISLAS versus 66±6.3 years for MEDIA‐DHF; P=0.44). Before adjustment, both thrombomodulin and perlecan were increased with HFpEF, with a more significant increase with perlecan (P<0.0001 versus P=0.037) (Figure 1A and 1B). In IPW analysis, perlecan remained increased in patients with HFpEF (7.21±0.31 NPX versus 7.33±0.49; P<0.0001) but not thrombomodulin (Figure 1C and 1D). Decorin levels were also investigated and found to be elevated in patients with HFpEF from the MEDIA‐DHF cohort compared with volunteers from the STANISLAS cohort following IPW analysis (Table S2). In the STANISLAS cohort, decorin was associated with LAVi after adjustment, while in the MEDIA‐DHF cohort, it was linked to LVMi but not to other indicators of diastolic dysfunction (Table S3). In the MEDIA‐DHF cohort, decorin was not significantly associated with outcomes; however, it was significantly associated with outcomes in the BIOSTAT‐CHF cohorts, even after adjustment (P=0.003; Tables S4 and S5).

Table 2.

Characteristics of Patients With HFpEF (MEDIA‐DHF Cohort) According to Perlecan and Thrombomodulin (Olink) Tertiles

	Perlecan (NPX)				Thrombomodulin (NPX)
	Tertile 1	Tertile 2	Tertile 3	P value	Tertile 1	Tertile 2	Tertile 3	P value
	(5.32–6.41)	(6.41–6.82)	(6.82–8.30)		(7.34–8.49)	(8.49–8.82)	(8.82–9.97)
	(n=151)	(n=157)	(n=152)		(n=150)	(n=155)	(n=151)
Clinical data
Age, y	71 (65–77)	75 (69–80)	77 (70–82)	<0.0001	73 (67–79)	74 (68–80)	76 (69–82)	0.009
Female sex	89 (58.9)	110 (70.1)	89 (58.6)	0.057	98 (65.3)	104 (67.1)	83 (55.0)	0.063
Body mass index, kg/m2	29.7±5.1	30.5±6.2	31.3±6.6	0.075	30.5±6.1	30.9±5.9	30.1±6.1	0.54
Smoking status
Current smoker	19 (12.7)	6 (3.8)	12 (8.0)		14 (9.4)	9 (5.8)	14 (9.4)
Former smoker	47 (31.3)	55 (35.3)	67 (44.7)	0.008	51 (34.2)	48 (31.2)	67 (45.0)	0.039
Never smoked	84 (56.0)	95 (60.9)	71 (47.3)		84 (56.4)	97 (63.0)	68 (45.6)
NYHA class
Class I	13 (8.7)	0 (0.0)	9 (6.0)	<0.0001	7 (4.7)	4 (2.6)	10 (6.7)	0.0104
Class II	124 (83.2)	130 (82.8)	97 (64.7)		125 (83.3)	124 (81.0)	99 (66.4)
Class III	12 (8.1)	27 (17.2)	43 (28.7)		18 (12.0)	25 (16.3)	39 (26.2)
Class IV	0 (0.0)	0 (0.0)	1 (0.7)		0 (0.0)	0 (0.0)	1 (0.7)
HR, bpm	65 (60–73)	68 (60–77)	68 (59–78)	0.24	65 (58–77)	68 (60–75)	68 (60–77)	0.87
SBP, mm Hg	138.4±19.6	140.6±23.3	133.9±23.6	0.029	137.6±21.0	137.1±21.9	138.7±24.2	0.83
Medical history
History of hypertension	132 (88.0)	135 (86.5)	137 (91.3)	0.39	124 (83.2)	140 (90.9)	136 (91.3)	0.059
Diabetes	54 (36.2)	52 (33.1)	69 (45.7)	0.064	51 (34.2)	54 (34.8)	68 (45.6)	0.080
Dyslipidemia	90 (60.8)	88 (56.1)	88 (59.9)	0.68	81 (54.7)	93 (60.8)	88 (59.9)	0.52
Previous hospitalization for HF	44 (29.3)	37 (24.3)	71 (47.3)	<0.0001	41 (27.9)	45 (29.6)	65 (43.6)	0.008
Coronary artery disease	41 (28.3)	46 (30.5)	56 (37.6)	0.21	46 (31.9)	40 (26.8)	54 (36.5)	0.21
Stroke or TIA	17 (11.3)	21 (13.5)	14 (9.4)	0.55	23 (15.5)	16 (10.3)	12 (8.1)	0.12
Peripheral artery disease	8 (5.4)	10 (6.5)	18 (12.4)	0.068	7 (4.8)	11 (7.2)	18 (12.4)	0.056
Asthma	17 (11.4)	17 (10.8)	8 (5.4)	0.13	17 (11.5)	13 (8.4)	12 (8.2)	0.56
Chronic obstructive pulmonary disease	16 (10.8)	34 (21.7)	25 (17.0)	0.036	24 (16.3)	27 (17.4)	23 (15.8)	0.94
Sleep disorder (sleep apnea)	9 (6.1)	9 (5.7)	13 (8.8)	0.55	8 (5.4)	11 (7.1)	11 (7.5)	0.77
Biology
Plasma glucose, g/L	1.05 (0.95–1.26)	1.01 (0.94–1.26)	1.03 (0.90–1.27)	0.52	1.01 (0.94–1.26)	1.03 (0.94–1.26)	1.05 (0.90–1.27)	1.00
Hemoglobin A1c, %	6.00 (5.60–6.70)	6.25 (5.90–7.00)	6.20 (5.80–7.00)	0.18	6.00 (5.70–6.60)	6.20 (5.80–6.80)	6.20 (5.80–7.00)	0.42
Triglycerides, g/L	1.17 (0.86–1.67)	1.17 (0.88–1.54)	1.06 (0.82–1.55)	0.53	1.14 (0.80–1.33)	1.24 (0.90–1.77)	1.06 (0.84–1.60)	0.049
HDL, g/L	0.53 (0.44–0.63)	0.52 (0.43–0.62)	0.49 (0.39–0.60)	0.088	0.54 (0.44–0.63)	0.50 (0.43–0.61)	0.49 (0.39–0.62)	0.13
LDL, g/L	1.02±0.35	1.08±0.36	0.89±0.39	0.0001	1.00±0.33	1.03±0.40	0.98±0.39	0.56
Creatinine, mg/L	8.00 (7.00–9.62)	9.30 (8.14–10.97)	13.57 (11.09–17.60)	<0.0001	8.37 (7.24–9.95)	9.50 (8.03–11.20)	13.01 (10.18–16.80)	<0.0001
eGFR, mL/min per 1.73 m2	79.3±15.4	66.2±16.3	43.3±18.0	<0.0001	74.7±16.7	65.8±19.3	48.7±22.3	<0.0001
Echocardiographic variables
LVEF, %	61.9±7.7	60.8±6.4	60.4±6.6	0.13	61.1±6.9	61.2±7.2	60.7±6.7	0.79
LVMi, g/m2	121.3±33.2	109.8±35.4	117.8±35.3	0.020	112.9±34.6	114.1±29.0	121.9±40.1	0.080
LAVi, mL/m2	41.6±14.3	41.7±15.3	46.6±17.9	0.009	45.4±15.7	40.2v13.6	44.0±18.3	0.013
E′, cm/s	7.22±2.12	7.30±2.04	7.33±1.95	0.91	7.66±2.15	7.07±2.08	7.08±1.82	0.021
E/e′	12.0 (9.56–15.00)	11.3 (9.40–14.54)	12.7 (10.22–16.94)	0.018	11.6 (9.48–14.51)	12.2 (9.67–15.43)	12.6 (9.80–16.71)	0.14
PASP, mm Hg	30.08±10.85	33.39±11.88	39.11±12.37	<0.0001	33.21±11.22	32.50±12.31	37.20±13.03	0.009
LVH	104 (83.9)	99 (69.2)	93 (70.5)	0.010	100 (73.5)	102 (76.7)	92 (73.0)	0.77
Treatments
ACE inhibitors	78 (52.3)	75 (47.8)	68 (45.3)	0.47	79 (53.0)	71 (46.1)	70 (47.0)	0.44
ARB	48 (32.2)	63 (40.4)	47 (31.1)	0.18	51 (34.5)	54 (35.1)	51 (34.0)	0.99
β Blockers	100 (67.1)	109 (69.4)	120 (79.5)	0.036	113 (75.8)	108 (70.1)	104 (69.3)	0.39
Diuretics	99 (66.9)	120 (76.4)	135 (89.4)	<0.0001	113 (76.4)	110 (71.4)	127 (84.7)	0.020
Calcium channel blockers	64 (43.0)	65 (41.4)	68 (45.0)	0.81	51 (34.2)	69 (44.8)	77 (51.3)	0.011
Antiplatelets	83 (55.7)	76 (48.4)	79 (52.3)	0.44	81 (54.4)	71 (46.1)	84 (56.0)	0.18
Anticoagulants	48 (32.4)	60 (38.2)	76 (50.3)	0.006	64 (43.0)	57 (37.3)	60 (40.0)	0.60
Antidiabetic agents	45 (30.2)	46 (29.5)	57 (37.7)	0.24	47 (31.5)	42 (27.5)	57 (38.0)	0.15
Statin	90 (60.4)	89 (56.7)	100 (66.2)	0.22	93 (62.4)	94 (61.0)	88 (58.7)	0.80

Data are expressed as n (%), mean±SD, or median (interquartile range). P values are obtained from ANOVA or Kruskal–Wallis tests (continuous variables) and χ² or Fisher's exact tests (categorical variables). ACE indicates angiotensin‐converting enzyme; ARB, angiotensin receptor blocker; eGFR, estimated glomerular filtration rate; HDL, high‐density lipoprotein; HF, heart failure; HFpEF, heart failure with preserved ejection fraction; HR, heart rate; LAVi, left atrial volume index; LDL, low density‐lipoprotein; LVEF, left ventricular ejection fraction; LVH, left ventricular hypertrophy; LVMi, left ventricular mass index; MEDIA‐DHF, Metabolic Road to Diastolic Heart Failure; NYHA, New York Heart Association; PASP, pulmonary artery systolic pressure; SBP, systolic blood pressure; and TIA, transient ischemic attack.

Figure 1. Comparison of perlecan and thrombomodulin (Olink) distributions between control populational participants and patients with HFpEF (STANISLAS versus MEDIA‐DHF cohorts).

A and B, Comparison between populational participants and HFpEF of plasma concentrations of perlecan and thrombomodulin before adjustment. C and D, Comparison between populational participants and HFpEF of plasma concentrations of perlecan and thrombomodulin after IPW adjustment. P values were obtained from logistic regression models with difference‐in‐difference as outcome and Olink biomarkers as explanatory continuous variables. Populational participants, n=466; patients with HFpEF, n=312. Mean±SD values are presented above each panel. HFpEF, heart failure with preserved ejection fraction; IPW, inverse probability weight; MEDIA‐DHF, Metabolic Road to Diastolic Heart Failure; NPX, normalized protein expression; and STANISLAS, Suivi Temporaire Annuel Non‐Invasif de la Santé des Lorrains Assurés Sociaux (Annual Noninvasive Temporary Monitoring of the Health of Insured Lorrainers).

Associations Between Biomarkers of Glycocalyx Degradation and Echocardiographic Variables

In STANISLAS participants, perlecan linear regression was positively associated in multivariable models with higher LV filling pressure (β for E/e'=0.41 [95% CI, 0.13–0.69]; P=0.005). Echocardiographic parameters were also analyzed as binary outcomes, with left atrial dilatation positively associated after adjustment (β for LAVi =2.33 [95% CI, 1.08–5.01]; P=0.031; Table 3). After adjustment, thrombomodulin linear regression was negatively associated for E′ (β=−0.49 [95% CI, −0.86 to −0.11]; P=0.015). In patients with HFpEF, perlecan was positively associated after adjustment for E/e′, LAVi, and pulmonary artery systolic pressure, respectively: β=2.65 [95% CI, 1.20–4.10]; P=0.0004; β=5.68 [95% CI, 1.08–10.27]; P=0.016; and β=10.8 [95% CI, 6.91–14.68]; P<0.0001. Thrombomodulin was positively associated after adjustment with LV filling pressure and LV mass E/e' and LV mass LVMi (β for E/e′=2.08 [95% CI, 0.70–3.45]; P=0.003; β for LVMi, 17.27 [95% CI, 7.40–27.40]; P=0.0006), and negatively with E′ and LAVi (β=−1.17, [95% CI, −1.72 to −0.63]; P<0.0001; and β=−5.06 [95% CI, −9.45 to −0.67]; P=0.024), respectively. To further examine the association between echocardiographic variables and changes in thrombomodulin or perlecan levels with HFpEF, the data were also analyzed according to biomarker concentration tertiles (Table S6). After full adjustment, E/e′ was increased in the third tertile of thrombomodulin, while LAVi was decreased in the second tertile. After full adjustment, perlecan was increased in the third tertile for E/e′, LAVi, and pulmonary artery systolic pressure (P=0.009, P=0.027, and P<0.0001, respectively).

Table 3.

Linear Relationship Between Perlecan, Thrombomodulin, and Echocardiographic Variables (E/e′, e′, LAVi, LVMi) Used as Continuous Variables in Control Participants (STANISLAS Populational Cohort) and Patients With HFpEF (MEDIA‐DHF Cohort)

Perlecan	Continuous or binary outcome		Populational participants			P value	Patients with HFpEF	P value	Adjusted‡	P value
			Univariable	P value	Adjusted‡		Univariable
E/e′	Continuous	β (95% CI)	1.27 (0.99 to 1.55)	<0.0001	0.41 (0.13 to 0.69)	0.005	1.42 (0.45 to 2.38)	0.004	2.65 (1.20 to 4.10)	0.0004
	Binary, ≥8/15 *	OR (95%CI)	3.54 (2.23 to 5.60)	<0.0001	1.29 (0.74 to 2.27)	0.37	1.715 (1.11 to 2.65)	0.015	5.15 (2.38 to 11.15)	<0.0001
E′	Continuous	β (95% CI)	−2.63 (−3.12 to −2.13)	<0.0001	−0.04 (−0.43 to 0.35)	0.83	0.08 (−0.32 to 0.48)	0.68	−0.33 (−0.92 to 0.26)	0.27
	Binary, <9 cm/s	OR (95%CI)	3.877 (2.60 to 5.77)	<0.0001	1.04 (0.62 to 1.73)	0.88	1.02 (0.63 to 1.67)	0.92	1.28 (0.59 to 2.79)	0.53
LAVi	Continuous	β (95% CI)	1.69 (0.55 to 2.83)	0.004	1.05 (−0.22 to 2.32)	0.11	4.32 (1.29 to 7.34)	0.005	5.68 (1.08 to 10.27)	0.016
	Binary, >34 mL/m2	OR (95%CI)	3.32 (1.74 to 6.31)	0.0003	2.33 (1.08 to 5.01)	0.031	1.46 (0.94 to 2.27)	0.092	1.51 (0.77 to 2.95)	0.23
LVMi	Continuous	β (95% CI)	9.77 (6.83 to 12.70)	<0.0001	0.59 (−2.31 to 3.49)	0.69	−1.35 (−8.59 to 5.88)	0.71	2.56 (−8.30 to 13.425)	0.64
	LVH	OR (95%CI)	2.88 (1.80 to 4.60)	<0.0001	1.37 (0.79 to 2.38)	0.27	0.56 (0.35 to 0.89)	0.014	0.51 (0.24 to 1.09)	0.083
PASP†	Continuous	β (95% CI)	/	/	/	/	7.83 (5.29 to 10.36)	<0.0001	10.8 (6.91 to 14.68)	<0.0001
	Binary, ≥40 mm Hg	OR (95%CI)	/	/	/	/	3.84 (2.29 to 6.43)	<0.0001	9.51 (3.85 to 23.50)	<0.0001
Thrombomodulin
E/e′	Continuous	β (95% CI)	0.51 (0.21 to 0.80)	0.0008	0.12 (−0.15 to 0.39)	0.33	1.81 (0.69 to 2.94)	0.002	2.08 (0.70 to 3.45)	0.003
	Binary, ≥8/1515 *	OR (95%CI)	1.66 (1.06 to 2.60)	0.026	1.13 (0.66 to 1.93)	0.66	1.86 (1.11 to 3.12)	0.018	2.730 (1.37 to 5.46)	0.004
E′	Continuous	β (95% CI)	−2.09 (−2.60 to −1.58)	<0.0001	−0.49 (−0.86 to −0.11)	0.015	−0.80 (−1.26 to −0.34)	0.0007	−1.17 (−1.72 to −0.63)	<0.0001
	Binary, <9 cm/s	OR (95%CI)	3.04 (2.05 to 4.49)	<0.0001	1.79 (1.09 to 2.93)	0.021	2.87 (1.57 to 5.24)	0.0006	3.46 (1.655 to 7.251)	0.0010
LAVi	Continuous	β (95% CI)	0.91 (−0.28 to 2.10)	0.13	−0.08 (−1.35 to 1.18)	0.87	−2.67 (−6.21 to 0.87)	0.14	−5.06 (−9.45 to −0.67)	0.024
	Binary, >34 mL/m2	OR (95%CI)	1.96 (1.03 to 3.73)	0.040	1.37 (0.66 to 2.84)	0.40	0.65 (0.40 to 1.07)	0.093	0.51 (0.26 to 1.0)	0.049
LVMi	Continuous	β (95% CI)	9.57 (6.51 to 12.64)	<0.0001	0.90 (−1.98 to 3.78)	0.61	13.04 (5.03 to 21.04)	0.001	17.27 (7.40 to 27.14)	0.0006
	LVH	OR (95%CI)	2.12 (1.33 to 3.39)	0.002	1.63 (0.96 to 2.78)	0.072	1.16 (0.68 to 1.98)	0.58	1.32 (0.66 to 2.61)	0.43
PASP	Continuous	β (95% CI)	/	/	/	/	2.69 (−0.34 to 5.7)	0.082	0.89 (−2.80 to 4.59)	0.63
	Binary, ≥40 mm Hg	OR (95%CI)	/	/	/	/	1.88 (1.08 to 3.28)	0.027	1.53 (0.75 to 3.11)	0.25

Results are expressed as β (95% CI) or OR (95% CI); values obtained from linear regression models between echocardiography parameters (outcome) and thrombomodulin/perlecan (explanatory variables). E/e′, peak early mitral inflow velocity/ peak early diastolic mitral annular velocity ratio; HFpEF, heart failure with preserved ejection fraction; LAVi, left atrial volume index; LVH, left ventricular hypertrophy; LVMi, left ventricular mass index; MEDIA‐DHF, Metabolic Road to Diastolic Heart Failure; PASP, pulmonary artery systolic pressure; and STANISLAS, Suivi Temporaire Annuel Non‐Invasif de la Santé des Lorrains Assurés Sociaux (Annual Noninvasive Temporary Monitoring of the Health of Insured Lorrainers).

^*E/e′>8 (in healthy volunteers since only few cases of E/e′ ≥15), and E/e′ ≥15 in patients with HFpEF.

^† Thrombomodulin categorized according to the median for patients with HFpEF: relationship (to meet linearity assumption).

^‡ Adjusted for sex, age, body mass index, diabetes, treated hypertension, and estimated glomerular filtration rate.

Association Between Biomarkers of Glycocalyx Degradation and Clinical Outcomes in Patients With HFpEF (MEDIA‐DHF)

After a mean follow‐up of 12 months, Kaplan–Meier curves highlighted an increased risk of CVH and CVD in the tertile of patients with elevated perlecan concentrations (log‐rank test, P=0.0006; Figure 2A). Perlecan was positively correlated with CVH and CVD in the entire population but also in the third tertile of patients presenting elevated plasma concentrations (HR, 2.13 [95% CI, 1.36–3.34]; P=0.0009; and HR, 2.34 [95% CI, 1.30–4.22; P=0.004; Figure 2B). This observation was furthermore maintained after full adjustment (HR, 1.99 [95% CI, 1.11–3.56;, P=0.021); and HR, 2.44 [95% CI, 1.11–5.34]; P=0.026). On the other hand, thrombomodulin was not significantly associated with CVH and CVD (Figure 2A and 2B).

Figure 2. Association between perlecan, thrombomodulin, and CVH/CVD outcomes in patients with HFpEF (MEDIA‐DHF cohort).

A, Kaplan–Meier curves to illustrate the risk of CVD/CVH outcome in participant subsets defined according to tertiles of perlecan or thrombomodulin. B, Tertile values of perlecan or thrombomodulin without and with adjustment for sex, age, body mass index, diabetes, hypertension, and estimated glomerular filtration rate in patients with HFpEF (MEDIA‐DHF cohort). Univariable and adjusted hazard ratios were obtained from Cox regression models. CVD indicates cardiovascular death; CVH, cardiovascular health; HFpEF, heart failure with preserved ejection fraction; and MEDIA‐DHF, Metabolic Road to Diastolic Heart Failure.

Validation Analysis in the BIOSTAT‐CHF Cohorts

Mean follow‐up in this group was 3 years. Characteristics of patients with HFpEF from BIOSTAT index and validation cohorts are displayed in Table S7, and characteristics of patients from the pooled BIOSTAT cohorts, namely, BIOSTAT‐CHF according to perlecan and thrombomodulin, are presented in Tables S8 and S9, respectively. Patients with elevated perlecan were older, with a worsened kidney function, increased NT‐proBNP levels, hospitalization for HF, and death. Similar results were present for thrombomodulin except that HF hospitalization was not increased with higher thrombomodulin levels. Perlecan (log‐rank test, P<0.0001) was positively correlated with the risk of rehospitalization for heart failure and all‐cause death (Figure 3A). In addition, patients from the BIOSTAT‐CHF cohorts displayed an increased risk for rehospitalization for heart failure and all‐cause death with elevated circulating concentrations of perlecan even after adjustment (HR, 1.54 [95% CI, 1.32–1.80]; P<0.0001; Figure 3B). The highest risk was found in the tertile with elevated perlecan concentrations (HR, 2.12 [95% CI, 1.49–3.03]; P<0.0001 after adjustment). Thrombomodulin changes were less significant compared with perlecan (log‐rank test, P=0.0003). The highest risk was also present in the third tertile (HR, 1.70 [95% CI, 1.9–2.44]; P=0.004 after adjustment).

Figure 3. Association between perlecan and composite outcome of rehospitalization for heart failure hospitalization/all‐cause death in patients with HFpEF (BIOSTAT‐CHF cohorts).

A, Kaplan–Meier curves illustrating the risk of rehospitalization for heart failure hospitalization/all‐cause death outcome in participant subsets defined according to tertiles of perlecan or thrombomodulin in patients with HFpEF (BIOSTAT‐CHF cohorts). B, Tertile values of perlecan or thrombomodulin without and with adjustment for sex, age, body mass index, diabetes, hypertension and estimated glomerular filtration rate in HFpEF patients (BIOSTAT‐CHF cohort). Univariable and adjusted HRs were obtained from Cox regression models stratified for cohort (index, validation). Adjusted for age, sex, body mass index, diabetes, hypertension and estimated glomerular filtration rate <60. BIOSTAT‐CH indicates Biology Study to Tailored Treatment in Chronic Heart Failure; HF, heart failure; HFpEF, heart failure with preserved ejection fraction; and HR, hazard ratio.

DISCUSSION

The endothelial glycocalyx is a mesh formed by membrane‐bound proteoglycans and glycoproteins that covers the intraluminal side of endothelial cells and is part of the protective endothelial barrier. The present study provides innovative findings regarding an association between endothelial glycocalyx degradation and impaired diastolic function in a population‐based cohort as well as in patients with HFpEF. In addition, increased glycocalyx degradation in patients with HFpEF was found to be positively correlated with hospitalization and death. Finally, perlecan was identified as a new robust biomarker strongly correlated with diastolic dysfunction and outcomes.

Association Between Diastolic Function and Glycocalyx Degradation in Patients With HFpEF as Well as in Individuals Without HF

In HFpEF, arterial and myocardial stiffness increases with a concomitant decrease in LV relaxation and LV end‐diastolic pressure leading to impaired LV filling. Diastolic dysfunction is associated with other vascular dysfunctions such as abnormal ventricular–arterial coupling, pulmonary hypertension, and coronary microvascular dysfunction. ³³ Hypertension is a risk factor for HFpEF, and both hypertension and HFpEF are linked to endothelial dysfunction. As endothelial glycocalyx is a fragile layer, it was hypothesized that hypertension could trigger its degradation. ³⁴ However, hypertension not only includes hemodynamic modifications but also the presence of a low‐grade chronic inflammation, and both were described as being able to alter the endothelial glycocalyx. ³⁵ Still, to what extent each of these 2 aspects is predominant in glycocalyx degradation and may participate in the development of HFpEF is still questioned.

In the present study, plasma concentrations of perlecan, but not thrombomodulin, were increased with HFpEF compared with a deemed healthy population even while accounting for confounders, including age with IPW analysis. We also observed that in populational participants and patients with HFpEF, diuretics were increased with higher plasma concentrations of perlecan. This observation may be linked to higher levels of congestion associated with higher levels of diastolic dysfunction, thus requiring more diuretics. This relation between glycocalyx damage and HFpEF progression through the renin–angiotensin–aldosterone system, and increased congestion was recently reviewed and is in line with our observations. ³⁶ Regarding diastolic features, after full adjustment, perlecan was correlated with diastolic dysfunction (E/e') in control subjects while both perlecan and thrombomodulin were significantly correlated after adjustment with diastolic dysfunction (E/e′) in HFpEF. These results indicate that biomarkers of endothelial glycocalyx degradation are associated with diastolic dysfunction even before the occurrence of HFpEF. These observations are consistent with the current pathophysiological paradigm of HFpEF pointing to a continuous degradation of endothelial cell function due to comorbidity‐associated chronic inflammation. ⁷ However, the role of the endothelial glycocalyx was not yet integrated in the paradigm set forth by Paulus and Tschöpe. ⁹ In a study from 2020, patients with HFpEF showing endothelium‐independent microvascular dysfunction had worse diastolic dysfunction (E/e′) and a higher overall mortality rate. ³⁷ This observation is a cornerstone in understanding the pathophysiology of HFpEF and is in keeping with our results of the preponderant association of perlecan compared with thrombomodulin with diastolic dysfunction and outcomes. Of note, no single clinical characteristic investigated herein was able to identify the presence or absence of endothelium‐independent coronary dysfunction. Thus, this study also highlights the importance of finding new biomarkers reflecting the pathophysiology of HFpEF (eg, endothelial glycocalyx degradation over endothelial dysfunction). Our results indicate that endothelial glycocalyx degradation is of central importance in the shift toward diastolic dysfunction and subsequent HFpEF development. To our knowledge, this is the first instance in which glycocalyx degradation was found to be relevant in the early development of HFpEF and not only in outcomes. Altogether, the distinction between HFpEF and HFrEF highlighted by Paulus and Tschöpe aligns with the potential predominant pathophysiological role of endothelial glycocalyx dysfunction in HFpEF as highlighted by these results.

Association of Glycocalyx Degradation Markers With Outcomes

Endothelial glycocalyx degradation is not only correlated with the early cascade of events leading to HFpEF but was also found in our study to be strongly positively correlated with hospitalization and death. In the MEDIA‐DHF cohort, outcomes were strongly correlated with perlecan (P=0.0006) but not with thrombomodulin (P=0.22). In a second cohort of patients with HFpEF (the BIOSTAT‐CHF cohorts), a similar association with perlecan was observed (P<0.0001), while the association with thrombomodulin also reached significance (P=0.0003). Decorin may be a new marker for glycocalyx degradation. It was elevated with HFpEF but was associated with LVMi only in the MEDIA‐DHF cohort and with LAVi in the STANISLAS cohort. While it was not significantly associated with outcome in the MEDAI‐DHF cohort, it reached significance in the BIOSTAT‐CHF cohorts. A recent study comparing decorin to common biomarkers of glycocalyx degradation, including hyaluronic acid and heparin sulfate, in women with preeclampsia found no significant increase in plasma decorin levels in the presence of complications (such as acute kidney injury, eclampsia, abruption, pulmonary edema, or HELLP syndrome [hemolysis, elevated liver enzyme levels, and a low platelet count]). ³⁸ In contrast, the other biomarkers frequently showed elevated levels in these conditions, suggesting that decorin may be a less reliable marker for glycocalyx degradation than other biomarkers.

Certain findings regarding endothelial glycocalyx degradation in HFpEF have previously been reported, using other glycocalyx biomarkers such as syndecan‐1. Syndecan‐1 was found to be increased in patients with chronic HFpEF and correlated with an increased risk of death during a 3‐year follow‐up but not in patients with chronic HFrEF. ³⁹ In another study focusing on patients with HFrEF in which syndecan‐1, hyaluronan, and heparan sulfate were measured, ⁴⁰ all of these endothelial glycocalyx biomarkers were unchanged in patients with HFrEF compared with controls, thus highlighting a limited implication of endothelial glycocalyx degradation in patients with chronic HFrEF. These results are also in line with the differential development of HFpEF compared with HFrEF resulting from cardiomyocyte death. ⁷ It should be emphasized that in some HFpEF and HFrEF populations, syndecan‐1 was increased in HFrEF compared with patients with HFpEF. ⁴¹ However, in this latter study, patients were not matched for New York Heart Association stage and most cases of HFpEF were of class I and II, while most cases of HFrEF were of class III and IV. This underscores a pressing challenge when aiming to study endothelial glycocalyx degradation in patients with HF since most studies intermix the type and stage of HF (or fail to provide this information), thus limiting the possibility of further unraveling the pathophysiological mechanisms underpinning endothelial glycocalyx degradation.

Perlecan as a Novel Biomarker in HFpEF

New biomarkers may represent better options for assessing endothelial status than those currently identified. To our knowledge, only 3 studies in the field of HF or closely related to HF reported findings pertaining to perlecan, all of which were recently published. ⁴² , ⁴³ , ⁴⁴ In the DIAST‐CHF (Diagnostic Study on Prevalence and Clinical Course of Diastolic Dysfunction and Diastolic Heart Failure) study, ⁴² perlecan was found to be increased in patients with HFrEF (n=38) as well as being a key protein in a network analysis of exclusive protein–protein correlation. ⁴² In patients with HFpEF, despite being present in the biomarkers list that stood out in the network analysis, perlecan was not significantly modified. As highlighted above, the absence of disease gradation precludes drawing conclusions from these results. In the Bio‐SHiFT (Serial Biomarker Measurements and New Echocardiographic Techniques in Chronic Heart Failure Patients Result in Tailored Prediction of Prognosis) study ⁴³ consisting of a prospective cohort study of patients with chronic HF (majority HFrEF), perlecan was an independent predictor (with matrix metalloproteinase‐2, matrix metalloproteinase‐9, and tissue inhibitor of metalloproteinase‐4) of the primary end point (composite of cardiac death, heart transplantation, left ventricular assist device implantation, hospitalization for acute or worsened HF). In addition, this study further highlighted an increase in perlecan concentrations upon approaching the primary end point event. The third study explored patients with pulmonary artery hypertension since this condition is closely related to the development of right HF. The elevation of perlecan was also correlated with poor survival in this population. ⁴⁴

Owing to the present results, perlecan can be considered as one of these novel biomarkers. In a cohort of patients with HFrEF, plasma syndecan‐1 was not elevated compared with controls, while hyaluronan, a component of the glycocalyx not bound to the endothelial membrane similarly to perlecan, was significantly increased. ¹³ A nonmembrane bound biomarker such as perlecan may thus be more prone to respond to small glycocalyx regulations and to be released in the circulation in the case of endothelial impairment, whereas membrane‐bound proteins such as syndecan‐1 or thrombomodulin are more likely released only in instances of more advanced endothelial damage. Indeed, most studies that assessed soluble thrombomodulin concentrations as a marker of glycocalyx degradation marker and found raised concentration compared with controls were performed in severe diseases such as sepsis or COVID‐19. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ Accordingly, in a less severe condition such as HFpEF, we observed that perlecan but not thrombomulin was increased compared with a control population.

Additionally, glycocalyx degradation can reflect extracellular matrix remodeling and the shift toward profibrotic signaling, which are also of central importance in the establishment of HFpEF. ⁴⁵ , ⁴⁶ Perlecan also participates in endothelial barrier function and vascular smooth muscle cell proliferation inhibition. ⁴⁷ , ⁴⁸ Thus, its release in the blood may reflect a deeper alteration of vascular homeostasis in addition to damaged endothelial glycocalyx. Given the above, perlecan could be used as an early risk biomarker in individuals at risk of developing HFpEF.

Clinical Perspectives

The present results are of clinical significance for 2 aspects, namely, (1) highlighting the relevance of endothelial glycocalyx degradation in the course of events leading to HFpEF, and (2) identifying a new useful biomarker of glycocalyx degradation for HFpEF prognosis.

Maintaining endothelial glycocalyx integrity is crucial for endothelial cell function since its degradation contributes to endothelial dysfunction. Strategies to preserve the glycocalyx include injections of glycocalyx components or inhibition of enzymes that lead to its degradation. Large molecules, such as albumin, can counteract glycocalyx shedding. ⁴⁹ In diabetic cardiomyopathy, restoring glycocalyx with angiopoietin‐1 has been shown to improve heart function in mouse models. ⁵⁰

Recent efforts have aimed to pharmacologically enhance the endothelial glycocalyx using supplements, mixing antioxidants, hyaluronan, glucosamine sulfate, and fucoidan (a heparan sulfate mimetic able to inhibit heparinase). ⁵¹ Clinical trials have primarily used glycocalyx improvement as a marker rather than a therapeutic target. A study investigating the use of supplements, which began in the spring of 2023, is currently undergoing a phase 2 clinical trial in patients with HF and is expected to be concluded in 2026 (NCT05966415). Regarding the use of perlecan as a biomarker, increased plasma perlecan concentrations can be viewed as both an endothelial glycocalyx degradation biomarker and an endothelial cell barrier disruption biomarker. The use of such a dual‐faceted biomarker could represent a highly valuable informative tool compared with other endothelial or glycocalyx biomarkers since it is yielding 2 sets of information at once. In addition, perlecan could be used as an early biomarker for heart function impairment as well as a biomarker for worsening HFpEF toward hospitalization and death.

Strengths and Limitations

A strength of our study is that 2 endothelial biomarkers were explored in 2 distinct populations: 1 consisting of populational participants (n=1705), and 1 composed of patients with HFpEF (n=460) using the same biomarker measurement protocol (Olink Proseek Multiplex panels: CVD II, CVD III). Additionally, our findings regarding perlecan elevation in HFpEF correlated with hospitalization and death were further confirmed in a second population of patients with HFpEF (n=556) in which the same multiplex panels were used.

As with any observational study, causality cannot be established. Additionally, residual bias may persist despite our analysis strategy designed to minimize it. To assess the association between glycocalyx markers and HFpEF presence, we used an IPW approach aimed at reducing bias; however, this method may not fully account for unmeasured confounding. For the association between glycocalyx markers and outcomes, models were adjusted for key confounders, including age, sex, body mass index, diabetes, hypertension, and estimated glomerular filtration rate. While additional variables may be relevant, sample size and power limitations prevented their inclusion. Nonetheless, from a clinical perspective, our models demonstrate that even after adjusting for the main drivers of confounding and outcome, glycocalyx markers remain significantly associated with outcome.

A limitation of our study is that the biomarker assay used herein does not provide standard concentration units of the measured proteins. Thus, proper plasma concentration measurements will be necessary to create cutoff thresholds for use by clinicians. In addition, glycocalyx degradation was assessed using only circulating biomarkers. Being able to assess the local or systemic degradation of endothelial glycocalyx with heart or peripheral vessel biopsies would be helpful but biopsies were not available in our study.

The 2 HFpEF populations are observational studies, and few cardiovascular events were detected; thus, we could not assess whether treatments may limit CVD and be correlated with decreased perlecan concentrations.

The MEDIA‐DHF cohort consisted of chronic and recently discharged patients with HFpEF; thus, observations may not be extrapolated to sicker HFpEF populations. The BIOSTAT‐CHF cohorts recruited an array of patients with HF, including a minority of HFpEF; however, merging the index and validation data sets of the latter study enabled retrieving a sizeable number of patients and ensured a larger representativeness of geographic area, albeit both MEDIA‐DHF and BIOSTAT‐CHF are European cohorts.

CONCLUSIONS

The present study highlights for the first time the association between biomarkers of endothelial glycocalyx degradation and diastolic dysfunction in both population‐based and HFpEF cohorts and its association with clinical outcomes in patients with HFpEF. These findings indicate that the degradation of the endothelial glycocalyx could represent an initial stage in the pathological pathways leading to HFpEF, potentially gaining further prognostic significance in more advanced stages.

Sources of Funding

This project was supported by the French National Research Agency: ANR‐24‐CE17‐0492‐01. The fourth examination of the STANISLAS cohort was sponsored by the Centre Hospitalier Régional Universitaire of Nancy and the French Ministry of Health (Program Hospitalier de Recherche Clinique Inter‐régional 2013), by the Contrat de Plan Etat‐Lorraine and Fonds Européen de Développement Régional (FEDER Lorraine), and by a public grant overseen by the French National Research Agency as part of the second Investissements d'Avenir program FIGHT‐HF (reference: ANR‐15‐RHU‐0004) and by the French Projet investissement d'avenir project Lorraine Université d'Excellence (reference: ANR‐15‐IDEX‐04‐LUE). The MEDIA‐DHF study was supported by a grant from the European Union (FP7‐HEALTH‐2010‐MEDIA), by the French Programme Hospitalier de Recherche Clinique and by the RHU Fight‐HF, a public grant overseen by the French National Research Agency as part of the second Investissements d'Avenir program (reference: ANR‐15‐RHU‐0004), the GEENAGE (ANR‐15‐IDEX‐04‐LUE) program, by the Contrat de Plan Etat Région Lorraine, and FEDER IT2MP. The BIOSTAT‐CHF project was funded by a grant from the European Commission (FP7‐242209‐BIOSTAT‐CHF; EudraCT 2010‐020808‐29).

Disclosures

G.B. reports receiving personal fees from Abbott, AstraZeneca, and Boehringer. L.M. reports receiving personal fees from AstraZeneca and Vifor. J.M.t.M. declares speaker fees to her institution from Boehringer Ingelheim and Novartis. F.Z. has received personal fees from Boehringer Ingelheim, Janssen, Novartis, Boston Scientific, Amgen, CVRx, AstraZeneca, Vifor Fresenius, Cardior, Cereno Pharmaceutical, Applied Therapeutics, Merck, Bayer, CellProthera, CVCT, and Cardiorenal. A.A.V. received consultancy fees or research grants from Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Cytokinetics, Merck, Myokardia, Novartis, Novo Nordisk, and Roche Diagnostics. N.G. reports receiving personal fees from Novartis, Bayer, AstraZeneca, Lilly, Boehringer, and Vifor. All other authors have nothing to disclose.

Supporting information

Tables S1–S9

Figure S1