Association of Elevated NT-proBNP With Myocardial Fibrosis in the Multi-Ethnic Study of Atherosclerosis (MESA)

Abstract

BACKGROUND

Serum N-terminal pro–B-type natriuretic peptide (NT-proBNP) is considered a marker that is expressed in response to myocardial strain and possibly fibrosis. However, the relationship to myocardial fibrosis in a community-based population is unknown.

OBJECTIVES

The authors evaluated the relationship between cardiac magnetic resonance (CMR) measures of fibrosis and NT-proBNP levels in the MESA (Multi-Ethnic Study of Atherosclerosis) study.

METHODS

A total of 1,334 participants (52% white, 23% black, 11% Chinese, 14% Hispanic, and 52% men with a mean age of 67.6 years) at 6 sites had both serum NT-proBNP measurements and CMR with T1 mapping of indices of fibrosis at 1.5 T. Univariate and multivariable regression analyses adjusting for demographics, cardiovascular risk factors, and left ventricular (LV) mass were performed to examine the association of log NT-proBNP with CMR T1 mapping indices.

RESULTS

In the fully adjusted model, each 1-SD increment (0.44 pg/ml) of log NT-proBNP was associated with a0.62% increment in extracellular volume fraction (p < 0.001), 0.011 increment in partition coefficient (p < 0.001), and 4.7-ms increment in native T1 (p = 0.001). Results remained unchanged after excluding individuals with clinical cardiovascular disease or late gadolinium enhancement (n = 167), and after replacing LV mass by LV end-diastolic volume in the regression models.

CONCLUSIONS

Elevated NT-proBNP is related to subclinical fibrosis in a community-based setting. (Multi-Ethnic Study of Atherosclerosis [MESA]; NCT00005487) (J Am Coll Cardiol 2017;70:3102–9) Published by Elsevier on behalf of the American College of Cardiology Foundation.

Myocardial fibrosis is a common histological feature of the failing heart. Myocardial fibrosis occurs either as a reparative process after acute cardiac injury or as a nonspecific response to chronic conditions, including elevated myocardial strain and pressure overload. The identification of myocardial fibrosis by cardiac magnetic resonance imaging (CMR) provides a highly sensitive method to evaluate myocardial scar with prognostic significance (1,2). In the setting of heart failure, fibrosis may also be diffuse throughout the myocardium (3). Diffuse fibrosis may be identified in vivo noninvasively using CMR T1 mapping methods (3–5).

B-type natriuretic peptide (BNP) is synthesized in the myocardium as a pre-hormone (pro-BNP) and enzymatically cleaved into biologically active (BNP) and inactive (N-terminal pro-BNP [NT-proBNP]) fragments upon release into the circulation. Both BNP and NT-proBNP are well-described markers of congestive heart failure (6–8). NT-proBNP also predicts new cardiac events (9) as well as subclinical heart and vessel disease (10). NT-proBNP and BNP have been correlated with left ventricular (LV) wall stress (11) and myocardial scar size (12). A murine study showed that fibrosis determines cardiac BNP expression (13). Diffuse fibrosis assessed by CMR T1 mapping techniques were associated with elevated serum BNP in childhood hypertrophic cardiomyopathy (14) and systemic hypertension (15). However, the relationship between NT-proBNP and diffuse fibrosis in a community-based population is unknown.

The purpose of the present study was to examine the relationship between serum NT-proBNP levels and diffuse myocardial fibrosis as detected by CMR in a community-based study. These findings may help in understanding the association of increased neuro-hormonal levels on diffuse fibrosis and bridge the gap between the known prognostic implications of certain serum biomarkers and specific clinical phenotypes.

METHODS

STUDY DESIGN.

In 2000, the MESA (Multi-Ethnic-Study of Atherosclerosis) study began with the enrollment of 6,814 individuals, including individuals of white, African-American (black), Chinese, and Hispanic ethnicity. The baseline cohort included men and women with no clinical history of cardiovascular disease and between 45 to 84 years of age recruited from 6 field centers (Baltimore City and Baltimore County, Maryland; Chicago, Illinois; Forsyth County, North Carolina; Los Angeles County, California; Northern Manhattan and the Bronx, New York; and St. Paul, Minnesota). In 2010, MESA participants were examined for the 10-year follow-up (fifth examination). A total of 3,012 participants had CMR examinations, among which 1,334 participants (44%) consented to receive a CMR contrast agent and had T1 mapping. All participants underwent plasma NT-proBNP measurements in the fifth examination. Only participants with both CMR T1 mapping and NT-proBNP measurements were included in the final analyses. By the time of the follow-up exam, 80 participants (6%) had experienced an incident cardiovascular event (myocardial infarction, resuscitated cardiac arrest, angina, congestive heart failure, and stroke). One hundred seven participants (8%) had findings of late gadolinium enhancement (LGE) in the CMR studies of the fifth examination. Altogether, 167 participants had either a cardiovascular event or LGE. Institutional review boards at each of the 6 field centers approved the study protocols, and all attendees at the examinations provided written informed consent. Details of the MESA study design, NT-proBNP, and risk factor measurements have been described elsewhere (16,17).

CMR MEASURES.

1.5-T whole-body imagers were used in all CMR studies. LV cine imaging was performed using steady-state free precession sequences. LV structural and functional measures included LV mass, end-diastolic volume (EDV), end-systolic volume (ESV), mass-to-volume ratio, stroke volume, and ejection fraction. Myocardial T1 maps and CMR methods were described previously (18). Briefly, T1 mapping indices including pre-(native) and post-contrast T1 times at 12 and 25 min, partition coefficient (l), and extracellular volume fraction (ECV) (ECV = 100 × λ × [1 – hematocrit]) were assessed using a single–breath-hold modified Look-Locker inversion recovery sequence.

In the current study, hematocrit was measured in only 608 participants (45.5%) with CMR T1 mapping. ECV was calculated in these participants. Using the same methodology described in Treibel et al. (19), we performed an errors-in-variables linear regression between hematocrit as the dependent variable and pre-contrast T1 of the blood pool as the independent variable to model the synthetic hematocrit values as HCTsyn = 726.19 × (1/T1 _{precontrast blood}) – 0.07. The synthetic ECV and real ECV were highly correlated (intraclass correlation coefficient = 0.89) and the difference between the two was not significant (mean difference: 0.017 ± 1.435%; p = 0.75). These synthetic hematocrit values were then used for ECV calculations for those without hematocrit in the study.

STATISTICAL ANALYSIS.

Continuous variables were expressed as mean ± SD, and categorical variables as percentages. Linear regression was used to determine the relationship between log₁₀ NT-proBNP and each T1 index. Regression models were examined as follows: unadjusted models were fit, followed by models minimally adjusted for age, sex, and race, and finally fully adjusted models including minimal + smoking status, body mass index, heart rate, systolic blood pressure, diastolic blood pressure, low-density lipo-protein cholesterol, high-density lipoprotein cholesterol, triglycerides, hypertension (including those with antihypertensive treatment), diabetes, serum creatinine, statin usage, cardiovascular events or LGE, and LV mass. In secondary analyses, the fully adjusted model was repeated by replacing LV mass with LV EDV. We also repeated the fully adjusted model after excluding individuals with prior cardiovascular events or LGE. All analyses were performed using SPSS (version 23.0, IBM, Chicago, Illinois), and significance was declared as p < 0.05.

RESULTS

A total of 1,334 participants (52% men, age range 55 to 94 years, mean 67.6 years) had both serum NT-proBNP measurements and CMR T1 mapping quantifications. The study population was 52% white, 11% Chinese, 23% black, and 14% Hispanic ethnicity. One hundred sixty-seven participants (12.5%) who had incident cardiovascular events or LGE were included in the primary analyses. Plasma NT-proBNP levels ranged between 5 to 2,754 pg/ml (interquartile range: 33 to 124 pg/ml) with a median value of 64.9 pg/ml. The mean ECV was within normal limits (25.9 ± 3.0%) as was the native T1 time (977 ± 42.3 ms) (20). Other characteristics of the study cohort are shown in Table 1. Figure 1 demonstrates the box plot of log NT-proBNP across quartiles of ECV. When compared with the rest of participants without CMR T1 mapping acquisitions (unpaired Student’s t-test), the participants with T1 mapping were younger (67.6 ± 8.8 years vs. 70.1 ± 9.4 years; p < 0.001), with less hypertension (54% vs. 60%; p < 0.001), and lower log NT-proBNP levels (1.83 ± 0.44 pg/ml vs. 1.93 ± 0.47 pg/ml; p < 0.001).

TABLE 1.

Characteristics of Study Participants in the MESA’s Fifth Examination (N = 1,334)

Age, yrs	67.6 ± 8.8
Male	52
Race  White  Chinese  Black  Hispanic	52 11 23 14
Log NT-proBNP, pg/ml	1.83 ± 0.44
NT-proBNP, pg/ml	64.9 (33–124)
Body mass index, kg/m2	28.4 ± 5.3
Systolic blood pressure, mm Hg	121 ± 19
Diastolic blood pressure, mm Hg	68 ± 10
Heart rate, beats/min	64.1 ± 10.3
Current smoker	7
LDL cholesterol, mg/dl	105.1 ± 31.0
HDL cholesterol, mg/dl	54.3 ± 16.0
Triglyceride, mg/dl	112.0 ± 64.9
Hypertension	54
Diabetes	16
Serum creatinine, mg/dl	0.86 ± 0.19
Statins use	39
Presence of CVD* or LGE	167 (12.5)
LV EDV index, ml/m2	65.5 ± 13.7
LV ESV index, ml/m2	25.3 ± 8.4
LV mass index, g/m2	65.9 ± 13.0
LV M/V ratio, g/ml	1.03 ± 0.23
LV stroke volume index, ml/m2	40.2 ± 8.4
LV ejection fraction, %	61.7 ± 7.1
Extracellular volume fraction, %	25.9 ± 3.0
Partition coefficient	0.45 ± 0.04
Native T1, ms	977.0 ± 42.3
Post-12-min T1, ms	456 ± 40
Post-25-min T1, ms	519 ± 41

Values are mean ± SD, %, median (interquartile range), or n (%). LV structure index is by body surface area.

^*Including myocardial infarction, resuscitated cardiac arrest, angina, congestive heart failure, and stroke.

CVD = cardiovascular disease; EDV = end-diastolic volume; ESV = end-systolic volume; HDL = high-density lipoprotein; LDL = low-density lipoprotein; LGE = late gadolinium enhancement; LV = left ventricular; M/V = mass to end diastolic volume; NT-proBNP = N-terminal pro–B-type natriuretic peptide.

FIGURE 1. Box Plot of Log NT-proBNP Across Quartiles of CMR ECV.

Box plot showing the relationship between NT-proBNP and quartile of ECV. CMR = cardiac magnetic resonance; ECV = extracellular volume fraction; NT-proBNP = N-terminal pro–B-type natriuretic protein.

Table 2 shows the relationship of CMR markers of myocardial fibrosis and log NT-proBNP. In the regression analyses, log NT-proBNP was positively associated with all markers in the unadjusted analyses (all p < 0.001). However, the associations only remained significant in ECV (β = 1.4, p < 0.001), partition coefficient (β = 0.024, p < 0.001), and native T1 (β = 10.66, p = 0.001) in the fully adjusted model. Post-T1 times were not associated with NT-proBNP levels after adjustments. The unadjusted associations of ECV, native T1, and post-contrast T1 times at 12 min with log NT-proBNP are presented in Figure 2.

TABLE 2.

The Relationship of LV T1 Indices With Plasma Log NT-proBNP

Model		ECV (%)	Partition Coefficient	Native T1 (ms)	Post-12-min T1 (ms)	Post-25-min T1 (ms)
Unadjusted	β (SE)	2.2 (0.177)	0.025 (0.003)	14.39 (2.65)	−8.87 (2.5)	−13.38 (2.53)
	p Value	<0.001	<0.001	<0.001	<0.001	<0.001
Minimally adjusted*	β (SE)	1.55 (0.195)	0.023 (0.003)	10.85 (2.98)	−0.05 (2.64)	−2.81 (2.66)
	p Value	<0.001	<0.001	<0.001	0.98	0.29
Fully adjusted†	β (SE)	1.4 (0.2)	0.024 (0.003)	10.66 (3.11)	−0.38 (2.59)	−3.004 (2.57)
	p Value	<0.001	<0.001	0.001	0.88	0.24

Bold indicates p < 0.05.

^*Minimally adjusted for age, sex, and race.

^†Fully adjusted = minimally adjusted þ smoking status, body mass index, heart rate, systolic blood pressure, diastolic blood pressure, LDL cholesterol, HDL cholesterol, triglycerides, and hypertension (including those with antihypertensive treatment), diabetes, serum creatinine, statins usage, cardiovascular events or LGE, and LV mass.

ECV = extracellular volume fraction; SE = standard error; other abbreviations as in Table 1.

FIGURE 2. The Unadjusted Relationships Between Log NT-proBNP and T1 Indices.

The relationships between log NT-proBNP and myocardial fibrosis assessed by CMR ECV (A), native T1 (B), and post-contrast T1 (C) at 12 min. Unadjusted linear regression demonstrated significant positive correlation between log NT-proBNP and ECV as well as native T1 (p < 0.001). The correlation remained significant after adjusting for cardiovascular risk factors. NT-proBNP was negatively related to post contrast T1 in the unadjusted model (p < 0.001) but not after adjustment for cardiovascular risk factors. Abbreviations as in Figure 1.

Table 3 further displays standardized results for the association of ECV, partition coefficient, and native T1 with log NT-proBNP in the fully adjusted model, in which a 1-SD increment of log NT-proBNP(0.44 pg/ml) was associated with increments of0.62% in ECV, 0.011 in partition coefficient, and 4.72 ms in native T1. In the fully adjusted model, results were not substantially changed after excluding individuals with clinical cardiovascular disease or LGE (12.5% of the study population) or replacing LV mass with LV EDV (results not shown). Adjusting further for confounders, including educational level, physical activity, or coronary calcium score, also did not substantially alter the results. ECV appeared strongly associated with log NT-proBNP regardless of the covariates. Using only the ECV with laboratory hematocrit (n = 608, 46%) did not change the results, either. We also performed the fully adjusted analyses stratified by race/ethnicity. Whites and blacks demonstrated similar trends and patterns in the associations between log NT-proBNP and T1 indices as shown in Table 2, but these associations were not statistically significant for individuals of Chinese and Hispanic ethnicity.

TABLE 3.

Cross-Sectional Association Between CMR Markers of Fibrosis and Log NT-proBNP in the Fully Adjusted Model

	Extracellular Volume Fraction, %	Partition Coefficient	Native T1, ms
				β (95% CI*)	0.21 (0.15−0.27)	0.26 (0.19−0.32)	0.11 (0.01−0.20)
Increment in CMR marker per 1-SD change in log NT-proBNP (95% Cl)†	0.62 (0.4−0.72)	0.011 (0.008−0.013)	4.7 (0.43−8.5)

Markers of fibrosis include ECV, partition coefficient, and native T1. Bold indicates p < 0.05.

^*Standardized coefficient adjusted for age, sex, race, smoking status, body mass index, heart rate, systolic blood pressure, diastolic blood pressure, LDL cholesterol, HDL cholesterol, triglycerides, hypertension (including those with antihypertensive treatment), diabetes, serum creatinine, statin usage, cardiovascular events or LGE, and LV mass.

^†Increment in ECV (%), partition coefficient, or native T1 (ms) per 1-SD increment in log NT-proBNP (1-SD increment = 0.44 pg/ml).

CI = confidence interval; other abbreviations as in Tables 1 and 2.

DISCUSSION

CMR T1 mapping may be used to derive markers that reflect diffuse myocardial fibrosis. CMR is unique in this regard, as there is no other widely accepted method that can noninvasively assess myocardial tissue composition and diffuse fibrosis. NT-proBNP levels are expressed in response to increased myocardial wall stress and have also been linked to myocardial fibrosis in an animal study (13). Prior studies of patients with known cardiovascular disease have shown the utility of both T1 mapping indices and NT-proBNP to determine disease severity (7,8,21–24). However, our results show that even in a community setting of largely asymptomatic individuals, NT-proBNP is associated with myocardial fibrosis detected by CMR. These findings may be relevant to our understanding of the transition from subclinical disease to a clinical disease pheno-type that is ultimately characterized by myocardial stiffening and fibrosis, (e.g., as in diastolic heart failure) (Central Illustration).

CENTRAL ILLUSTRATION. Association of NT-proBNP With CMR-Measured ECV.

(Left) Cardiovascular risk factors ultimately result in myocardial remodeling and increased myocardial wall stress. In response, cardiac natriuretic peptides are secreted and are detected in the serum. Myocardial remodeling and wall stress will also result in expansion of the extracellular matrix and collagen deposition. (Right) In the MESA (Multi-Ethnic Study of Atherosclerosis) population, greater levels of NT-proBNP are associated with greater ECV. Regression estimates based on the fully adjusted models are shown. CMR = cardiac magnetic resonance; ECV = extracellular volume fraction; LV = left ventricular; NT-proBNP = N-terminal pro–B-type natriuretic protein.

Natriuretic peptides are produced primarily within the heart and released into the circulation in response to increased wall tension (25). Prior MESA studies have shown NT-proBNP as an independent predictor of heart failure (6,22). Reduced myocardial perfusion was also associated with increased NT-proBNP in another MESA study (23). Although the prognostic value of BNP or NT-proBNP in heart failure has been demonstrated, there are only limited data to link higher NT-proBNP levels to reactive collagen deposition before reparative fibrosis (15). Given that ECV, NT-proBNP, and BNP each has predictive power for cardiovascular disease outcomes in patients, there is an unmet need to elucidate the relationship between the imaging phenotype and serum biomarkers. Our results extend the prior observations of NT-proBNP in relationship to cardiovascular events or death in several community studies (21,24) to a subclinical marker of diffuse fibrosis. In the current community-based study, an individual with NT-proBNP of 209 pg/ml would be predicted to have a 2% greater ECV compared with an individual with NT-proBNP of 10 pg/ml. Rommel et al. (26) showed a mean difference of 4% in ECV in patients with heart failure with preserved ejection fraction versus control subjects. This suggests a preclinical phase of diffuse fibrosis may exist in asymptomatic subjects, and supports the concept that even relatively low elevations of NT-proBNP are an indicator of diffuse fibrosis. Furthermore, these data provide an implication to studies describing the role of myocardial fibrosis in the development of heart failure and support biomarkers as predictors of heart failure or adverse cardiovascular events.

Excess deposition of collagen in the extracellular matrix can lead to increased myocardial stiffness and, subsequently, to cardiac hypertrophy and LV dysfunction (27). Increased myocardial scar size was associated with higher BNP in chronic ischemic cardiomyopathy, and myocardial fibrosis appeared to be related to BNP through the process of ventricular remodeling (12). Understanding the pathogenesis of interstitial fibrosis in the stressed heart and dissecting the mechanisms responsible for adverse cardiac outcomes are crucial to design new strategies for prevention of cardiac failure and myopathy. In an animal model, the change of ECV occurred earlier than myocardial dysfunction in diabetic rabbits (28). It is generally accepted that in the process of the pathological remodeling of cardiomyopathy, progressive accumulation of extracellular matrix components increases cardiac stiffness (29,30). Cardiac natriuretic peptides are secreted in response to elevated pressure load. Structural rearrangement of cardiac chambers follows, and the expansion of the extracellular matrix further augments wall tension, which in turn triggers the production of cardiac natriuretic peptides. Reciprocal interaction of BNP secretion, LV remodeling, and interstitial fibrosis plays a crucial role in heart failure as an irreversible clinical endpoint.

Racial/ethnic differences in the relationship with incident heart failure or coronary heart disease have been reported (31,32). Blacks are at a higher risk for incident heart failure than other ethnic groups, which is mainly attributed to their higher prevalence of hypertension and diabetes mellitus (31). However, Akwo et al. (33) in the Southern Community Cohort Study found no difference in the heart failure incidence between whites and blacks with similar socioeconomic status. While examining NT-proBNP and T1 indices relationships, we recognized previously that multiple factors influenced CMR-measured T1 indices, including age and sex (18). The MESA study includes participants of 4 races/ethnicities. The aim of our study was to investigate the relationship between T1 mapping indices and NT-proBNP; hence, race/ethnicity was treated as a covariable to increase the statistical power. While examining such relationships independently in each race, the associations between NT-proBNP and T1 indices remained significant in whites and blacks, but most of these associations were not statistically significant in Chinese and Hispanics when adjusted for comorbidities. The weaker associations may be due to smaller sample sizes in these 2 races versus lack of association. Further understanding is needed regarding the interaction of race/ethnicity with diffuse fibrosis and NT-proBNP.

STUDY LIMITATIONS.

Because our analysis was cross-sectional, a causal relationship between NT-proBNP and CMR T1 mapping indices cannot be determined. It is not clear whether the elevated ECV is a result of NT-proBNP expression, or the elevated NT-proBNP expression is a result of ECV. CMR T1 mapping indices are not specific for myocardial fibrosis, although fibrosis is by far the most common determinant of altered CMR T1 indices in this community population. Synthetic ECV was used for 54% of the studies without a laboratory hematocrit. LV wall stress was not assessed. Multivariable adjustment was based on traditional cardiovascular disease risk factors that were generally available. Temporal bias and selection bias between baseline (year 2000) and follow-up (year 2010) might have occurred, and those willing to receive a CMR contrast agent may have been systematically healthier than those who were not eligible or declined. A total of 167 subjects had interim clinical cardiovascular disease between the baseline and follow-up year 2010, but the analysis produced similar results with or without those subjects.

CONCLUSIONS

In a large community-based population, we demonstrated a positive relationship between an imaging biomarker for diffuse fibrosis (CMR ECV) and an established serum biomarker (NT-proBNP) for heart failure. Our results suggest that NT-proBNP identifies a subclinical profibrotic condition in a community population. Further clarification will importantly assist in understanding the significance and utility of these biomarkers in studying the pathogenesis of fibrosis in heart failure.

PERSPECTIVES.

COMPETENCY IN MEDICAL KNOWLEDGE:

Blood levels of NT-proBNP are associated with myocardial fibrosis detected by CMR even in a cohort of mainly asymptomatic individuals.

TRANSLATIONAL OUTLOOK:

The link between NT-proBNP and extracellular volume may enhance understanding of the evolution of myocardial fibrosis in the transition from subclinical to clinical disease.

ABBREVIATIONS AND ACRONYMS

BNP

B-type natriuretic peptide

CMR

cardiac magnetic resonance

ECV

extracellular volume fraction

EDV

end-diastolic volume

ESV

end-systolic volume

LGE

late gadolinium enhancement

LV

left ventricular

NT-proBNP

N-terminal pro–B-type natriuretic peptide