MYOCARDIAL EXTRACELLULAR VOLUME FRACTION ADDS PROGNOSTIC INFORMATION BEYOND MYOCARDIAL REPLACEMENT FIBROSIS

Abstract

Background

Cardiac magnetic resonance (CMR) techniques permit quantification of the myocardial extracellular volume fraction (ECV), representing a surrogate marker of reactive interstitial fibrosis, and late gadolinium enhancement (LGE), representing replacement fibrosis or scar. ECV and LGE have been independently linked with heart failure (HF) events. In deriving ECV, coronary artery disease (CAD) type LGE, but not non-CAD type LGE, has been consistently excluded. We examined the associations between LGE, global ECV derived from myocardial tissue segments free of any detectable scar, and subsequent HF events.

Methods

Mid short-axis T1 maps were divided into 6 cardiac segments, each classified as LGE absent or present. Global ECV was derived from only segments without LGE. ECV was considered elevated if >30%, the upper 95% bounds of a reference group without known cardiac disease (n=28). Patients were divided into 4 groups by presence of elevated ECV and of any LGE. Subsequent HF hospitalization and any death were ascertained. Their relationship with ECV was examined separately and as a composite with Cox proportional hazard models.

Results

Of 1,604 serial patients with T1 maps, 1,255 were eligible after exclusions and followed over a median 26.3 (interquartile range 15.9–37.5) months. Patients with elevated ECV had increased risk for death (hazard ratio [HR] 2.45, 95% confidence interval [CI] 1.76–3.41), HF hospitalization (HR 2.45, 95% CI 1.77–3.40), and a combined endpoint of both outcomes (HR 2.46, 95% CI 1.94–3.14). After adjustments for covariates including LGE, the relationship persisted for death (HR 1.82, 95% CI 1.28–2.59), hospitalization (HR 1.60, 95% CI 1.12–2.27), and combined endpoints (HR 1.73, 95% CI 1.34–2.24).

Conclusions

ECV measures of diffuse myocardial fibrosis were associated with HF outcomes, despite exclusion of replacement fibrosis segments from their derivation and even among patients without any scar. ECV may have a synergistic role with LGE in HF risk assessment.

Clinical Summary

Elevations in myocardial extracellular matrix volume have been previously shown to be associated with a variety of cardiac pathologies. Extracellular volume fraction (ECV) quantification through non-invasive T1 mapping techniques by cardiovascular magnetic resonance imaging (CMR) has enabled the noninvasive measure of a surrogate marker of myocardial interstitial fibrosis. A few studies have found an association between ECV measures and future heart failure events. However, these studies used ECV measures without excluding permanent myocardial replacement fibrosis or scarring, another known marker of adverse heart failure outcomes. In our study, we show that elevations of such measures are associated with an increased risk for adverse heart failure outcomes in a large CMR referral base, independent of myocardial scarring, and may have use as a prognostic marker for increased risk of heart failure events.

Introduction

The overt clinical manifestation of heart failure often results from pathologic cardiac remodeling of the left ventricle due to various cardiac diseases. Fundamentally, such remodeling is recognized as myocardial changes on the genomic, molecular, cellular, and interstitial level in response to these diseases.¹ Ongoing clinical validation of methods for detecting pathologic cardiac remodeling hold promise for improving risk assessment in heart failure progression and for monitoring individual response to heart failure therapies.

In clinical practice, cardiovascular imaging is frequently employed to assess cardiac remodeling through morphologic and functional parameters.² Of the various, available imaging modalities, only cardiac magnetic resonance (CMR) permits the non-invasive, direct visualization of permanent replacement fibrosis or scar, an irreversible consequence of myocardial infarction.^{3, 4} Recent advances in cardiac magnetic resonance now permit the non-invasive assessment of myocardial extracellular matrix volume fraction (ECV), a measure that captures reactive interstitial fibrosis. Thus, ECV measures have been shown to be a surrogate marker for a key cardiac tissue component affected in earlier stages of cardiac remodeling by various cardiac diseases (e.g., diabetes, hypertension, and valvular heart diseases).^5–13

Both replacement and interstitial forms of cardiac fibrosis have become increasingly recognized as key components of pathologic cardiac remodeling.^{14, 15} Assessment of replacement fibrosis has been accepted as an established clinical method for determining myocardial viability and characterizing specific cardiomyopathies. Detection of this form of myocardial fibrosis has also become firmly established as a prognostic marker for adverse outcomes in a variety of cardiomyopathies.¹⁶ In contrast, only a few large, population-based studies have linked measures of myocardial interstitial fibrosis, assessed by ECV, with heart failure outcomes in a general CMR referral base.^{13, 17–20}

Questions remain over the utility of ECV measures for heart failure risk prediction. Prior outcome-based investigations employed ECV measures inclusive of visually detectable replacement fibrosis not associated with myocardial infarction. The predictive value of ECV, independent from that of detectable replacement fibrosis, has not been fully established. In addition, the development of incident heart failure risk prediction models is still at an early stage, and thus data on the incremental benefit of ECV measure to any such risk model have been limited.²¹

To address these issues, we examined ECV measures, derived independently of any visually detectable replacement fibrosis, and their association with heart failure outcomes in a large cohort of patients referred for clinical CMR. We hypothesized that such scar-free ECV measures 1) would have a significant association with heart failure outcomes and 2) would have an incremental improvement to a heart failure risk model for a clinical CMR referral base.

Methods

Because of confidentiality issues, data sets and study materials are safeguarded by the Houston Methodist Research Institute and cannot be made available to outside parties.

Study Population

The study was approved by the local institutional review board and written informed consent was obtained from all participants prior to enrollment. Subjects were serially recruited from patients referred to the cardiovascular magnetic resonance (CMR) laboratories (indications in Supplemental Table 1) of a single tertiary center for CMR imaging from June 2011 through January 2015, as part of an ongoing prospective cohort study designed to examine the relationship of myocardial extracellular volume fraction (ECV) with heart failure outcomes. Patients were included if they were aged ≥18 years, able to receive intravenous gadolinium-based contrast agents (i.e., no prior allergy to gadolinium agents, estimated glomerular filtration rate ≥30 mL/min/1.73 m²), had no contraindications to CMR, had completed the T1 mapping procedure (i.e., pre- and post-contrast imaging), had hematocrit levels available (because of its use in the derivation of ECV). Exclusion criteria included any missing relevant demographic data, technical issues with imaging, and any infiltrative cardiomyopathy or any cardiac tumors by clinical history or detected on CMR. Out of an initial group of 1,607 individuals recruited, 1,255 participants were selected for analyses after application of these inclusion and exclusion criteria (Figure 1). The patients were further divided into subcohorts based on the presence of significant valvular heart disease (VHD) (i.e., moderate or more VHD on other cardiac imaging), coronary artery disease (CAD) (i.e., documented acute coronary syndrome, angiographically confirmed coronary stenosis >50%, or prior coronary revascularization), and nonischemic cardiomyopathy (i.e., cardiomyopathy in the absence of VHD or CAD).²² An additional 28 healthy volunteers (n=13, aged 35.5 [SD 5.2] years, 69% male on the 1.5 Tesla scanner; n=15, aged 39.5 [SD 7.8] years, 80% male on the 3.0 Tesla scanner) without any known cardiovascular diseases or risk factors were recruited to determine the distribution of extracellular volume fractions in a reference control group.

Figure 1.

Study flow diagram showing application of inclusion and exclusion criteria to the cohort.

Outcomes

Prospective follow-up of serially recruited subjects was conducted through structured phone interviews with the participants, review of the available electronic health records (EHR), or contact with the referring clinic through December 31, 2016. The primary outcome of a first major heart failure event was defined as the earliest occurrence of either a heart failure hospitalization or death from any cause following the baseline CMR scan. Individual outcome types were examined separately as secondary outcomes. As established in the 2014 American College of Cardiology / American Heart Association (AHA) Cardiovascular Endpoints Data Standards, heart failure hospitalization events were defined as any hospitalization with a primary diagnosis of heart failure where the patient has a length of stay of at least 24 hours, symptoms or objective evidence of new or worsening heart failure, and initiation or intensification of heart failure therapies as documented in the hospital course or discharge summary by the admitting physician.²³ Deaths were ascertained from EHR review and/or Social Security Death Index queries. Because death certificates and medical records were unavailable in a significant proportion of patients who died, we elected not to include a sub-category of cardiovascular deaths. Events were adjudicated by a committee consisting of three board-certified cardiologists (D.J.S., F.N., E.Y.Y.).

CMR Imaging

Patients underwent CMR scans on either a 1.5-Tesla or a 3.0-Tesla clinical scanner (Siemens Avanto or Verio, respectively; Siemens, Erlangen, Germany) with phased-array receiver coil systems. Imaging protocols consisted at a minimum of an electrocardiography (ECG)-gated cine section and late gadolinium enhancement (LGE) imaging as previously described, and each of these protocol sections were followed with modified Look-Locker inversion recovery (MOLLI) sequences for T1 mapping just prior to and ~15 minutes after contrast administration.^{4, 6, 24–27}

Volumes & Function

Briefly, cardiac cines were acquired with a steady-state free precession sequence (SSFP) over 25–30 cardiac phases with in-plane spatial resolutions of 1.7 to 2.0 mm by 1.4 to 1.6 mm using sequential short-axis stacks (i.e., ventricular base to apex) with 10-mm increments (6-mm thickness, 4-mm gap) and standard cardiac long-axis views (i.e., left ventricular [LV] based 3-, 4-, and 2-chamber views). Cardiac chamber parameters were measured by level III trained CMR readers and indexed for body surface area.²⁸

Late Gadolinium Enhancement

LGE images were acquired over slice positions matched to cines about 10–15 minutes following intravenous gadolinium-based contrast administration (gadopentetate dimeglumine, gadoterate meglumine; 0.15 mmol/kg) with in-plane spatial resolutions of 1.8 mm by 1.3 mm and slice thicknesses of 6–7 mm with 3–4 mm gap. LGE images were obtained using inversion-recovery gradient echo sequences with inversion times set to null myocardial tissue signal (inversion time [TI] 250–350 msec).^{3, 29}

The presence and extent of myocardial LGE, considered synonymous with replacement fibrosis or scar tissue, was assessed using the AHA 17-segment model by level III trained CMR readers independently and included in official clinical reports.^{28, 30} For each myocardial segment, the extent of regional LGE was scored according to the spatial extent of LGE within each segment (0 = no LGE; 1 = 1%–25% LGE; 2 = 26%–50% LGE; 3 = 51%–75% LGE; and 4 = 76%–100% LGE).³¹ The total scar burden for the entire left ventricle was expressed as a percentage of LV myocardial volume and derived by averaging the score for all 17 segments. Data on total scar burden were abstracted from official clinical reports for use in this study.

T1 Mapping & ECV

An ECG-gated MOLLI sequence with SSFP image readout with motion correction was performed at a representative mid short-axis view of the left ventricle at two distinct time points within a CMR scan: once following cine imaging but prior to contrast administration (pre-contrast), and once ~15 minutes following LGE imaging (post-contrast). The mid short-axis was chosen due to concerns with partial volume due to through-plane cardiac motion near the base and near the apex. Technical details of the MOLLI sequence setup can be found in the Supplemental Methods.

A single reader – blinded to the clinical history, other CMR images, and clinical outcomes – post-processed all cases in randomized order and abstracted the T1 values for ECV calculations. Intra- and inter-reader reproducibility of ECV measures were separately assessed in a selected sub-group, which was re-randomized for each reader (Supplemental Figure 1). ECV was assessed for the 6 myocardial segments of the mid short-axis using post-processing image analysis tools (cvi⁴² software, Circle Cardiovascular Imaging, Calgary, Canada; supplemental material). The LGE image corresponding to the T1 maps was reviewed for LGE presence. Myocardial segments with LGE were classified as coronary artery disease (CAD) or non-CAD type LGE for subsequent exclusion from ECV calculations (Figure 2).

Figure 2.

Native (A, D) and post-contrast (B, E) T1 parametric maps and late gadolinium enhancement images (C, F) of corresponding mid cardiac short-axis locations for two patients are shown. Epicardial (green), endocardial (red), and blood pool (yellow) contours are shown on each T1 map. The white borders indicate divisions of the left ventricular short-axis into six segments combined with a 20% thickness offset from the epicardial and endocardial borders. Except for white blocks indicating exclusion of segments from ECV calculations because of the presence of non-infarct scar, all other segments were used to derive area-weighted averages of T1 times and ECV. The images for the top row came from a 75 year-old man with a hematocrit level of 39.7%, ECV of 34%, and 6% non-CAD scar in the septum, who was subsequently hospitalized 23.6 months later. The images for the bottom row came from a 73 year-old man with a hematocrit level of 44.6%, ECV of 26%, and 5% non-CAD scar in the anteroseptal and anterior walls, and who did not have a heart failure event over 42.7 months of follow-up

For each of the 6 myocardial segments, myocardial ECV was derived using pre- and post-contrast T1 values of myocardium and blood pool as previously described in other works.^{32, 33} Briefly, ECV was calculated as (1 – Hct) × (ΔR1_myocardium / ΔR1_blood), where Hct is the hematocrit level and ΔR1 represents the change in T1 relaxivity (R1 = 1/T1) before and after gadolinium-based contrast administration. Myocardial segments with any artifacts or LGE were excluded from the ECV calculation used for the main analyses. A mean global scar-free ECV value was then calculated, average-weighted for the size of each remaining usable myocardial segment. This parameter was not reported to the referring physician because it has not been established for use in clinical management, and imaging analyses for T1 mapping and ECV were conducted retrospectively.

Statistical Analyses

All data variables were assembled and statistical analyses were conducted using Stata 14.2 (College Station, TX), with key authors (E.Y.Y., D.T.N., E.A.G., D.J.S.) having full access to a copy of the raw and derived data sets. From the 28 healthy volunteers, the upper 95% cutoff or mean + 2 standard deviations (SD) of ECV was determined to be 30.2% on the 1.5-Tesla scanner and 31.1% on the 3.0-Tesla scanner. Receiver operator characteristic (ROC) analysis with Youden’s index was also used on the entire cohort to identify an optimal ECV cutoff for discriminating a first composite event over a fixed 4-year follow-up period.³⁴ The main cohort was dichotomized into low (≤ mean+2SD) vs. high (> mean+2SD) ECV categories based on ECV values and scanner type, and also dichotomized into scar absent (LV myocardial scar burden = 0%) vs. present (scar burden >0%).

Baseline characteristics and CMR parameters were described. All statistical tests were two-sided with significance set at p<0.05. Nominal variables were compared among ECV and scar groups using Chi-squared or Fisher’s exact testing as appropriate. Normality of continuous variables was assessed using Shapiro-Wilk testing. Because continuous variables were predominantly non-normally distributed, non-parametric comparison among ECV and scar groups were conducted for all continuous variables using Wilcoxon rank-sum testing.

Multivariable Cox proportional hazard risk models for all-cause mortality, heart failure hospitalization, or first composite event were developed using ECV both as a continuous variable (per every 5% increase) and as a binary variable (normal vs. elevated, based on cutoffs derived from the healthy volunteers). Patient survival and event-free survival for the first heart failure hospitalization, or first composite event were also depicted by the adjusted survival curves using the Stata’s stcurve function, which were adjusted for all covariates included in the final Cox proportional hazard model. Competing risk analysis for heart failure hospitalization was performed using the method of Fine and Gray.³⁵ Cause-specific cumulative incidence for heart failure hospitalization was estimated and plotted. Subdistribution hazard ratio (SHR) between subgroups (elevated ECV versus normal ECV) was also reported. Based on a previously validated heart failure prediction model from the Atherosclerosis Risk in Communities Study (ARIC),³⁶ we employed multivariable models with stepwise addition of covariate clusters to assess whether scar-free ECV categories remained significant predictors of incident heart failure events. The ARIC Heart Failure Model was chosen because of the limited availability of blood biomarker data for this cohort, required by other heart failure risk models, and applicability to a more general population. We further adjusted for other relevant CMR parameters. We also examined a parsimonious model, developed using Bayesian model averaging to select out only significant predictors of a first composite outcome unique for this cohort.^{37, 38}

Both Harrell’s c-statistic (survival model equivalent to area under the receiver operator curve or AUC) and likelihood ratio test of the models were also examined within the cohort and compared among multivariate models to determine the incremental benefit of scar-free ECV categories. The incremental benefit of scar-free ECV categories to the models’ discrimination powers was evaluated by comparing the c-statistic using Stata’s somersd package.³⁹ The model fitness was determined by the likelihood ratio test.

Multivariable models were constructed as follows: Model 1 – age, gender, black; Model 2 – Model 1 + body mass index, systolic blood pressure, heart rate, current smoker, former smoker, diabetes, prior myocardial infarction, and antihypertensive medication use (ARIC equivalent); Model 3 – Model 2 + left ventricular (LV) ejection fraction (EF), LV indexed myocardial mass; Model 4 – Model 3 + LV myocardial scar burden.

Results

Overall, the cohort was middle-aged with a slight male predominance (Table 1). Individuals with higher ECV or myocardial scar were likely to be diabetic, to have a prior history of myocardial infarction (MI), and higher heart rates. They were also more likely to be on renin-angiotensin-aldosterone system inhibitors, diuretics, and digoxin at the time of enrollment. Such individuals were also likely to have larger biventricular and left atrial volumes and left ventricular myocardial mass, and smaller biventricular stroke volumes and lower ejection fractions (Table 2). Furthermore, subjects with higher ECV tended to be female and non-white and to have lower blood pressures; whereas, subjects with myocardial scar tended to be male, hypertensive, hyperlipidemic, and current or former smokers, and to be taking aspirin, thienopyridines, statins, and beta-blockers.

Table 1.

^* Baseline characteristics and anthropometrics.

		Overall	Normal ECV	Elevated ECV		Scar (−)	Scar (+)
		(N=1,255)	(n=793)	(n=462)	p-value	(n=710)	(n=545)	p-value
Age (yrs)		60.8 (49.3, 69.1)	59.6 (47.4, 68.6)	61.9 (51.8, 70.3)	0.002	58.2 (44.1, 67.5)	63.7 (54.9, 71.2)	<0.001
Male (%)		57.1	60.3	51.7	0.003	48.6	68.3	<0.001
Non-white (%)		27.2	23.2	34.0	<0.001	26.5	28.1	0.84
Scanner	1.5 Tesla (%)	51.6	41.1	69.5	<0.001	47.6	56.7	0.001
	3.0 Tesla (%)	48.4	58.9	30.5		52.4	43.3
History of HTN (%)		64.5	62.3	68.3	0.03	57.1	74.1	<0.001
History of HLD (%)		52.2	51.6	53.3	0.58	43.4	69.7	<0.001
History of MI (%)		15.3	12.5	20.2	<0.001	5.5	28.1	<0.001
Diabetes (%)		21.8	18.3	27.8	<0.001	15.8	29.6	<0.001
Smoking Status	Current (%)	9.7	8.6	11.5	0.09	8.0	11.9	<0.001
	Former >1 yr (%)	30.6	30.6	30.5		27.8	34.1
Systolic Blood Pressure (mm Hg)		125.0 (114.0, 138.0)	128.0 (117.0, 139.0)	121.0 (110.0, 134.0)	<0.001	126.0 (116.0, 138.0)	124.0 (112.0, 138.0)	0.10
Diastolic Blood Pressure (mm Hg)		73.0 (65.0, 82.0)	75.0 (66.0, 83.0)	70.0 (62.0, 80.0)	<0.001	73.0 (65.0, 83.0)	73.0 (64.0, 82.0)	0.19
Heart Rate (bpm)		72.0 (63.0, 82.0)	70.0 (62.0, 80.0)	76.0 (66.0, 87.0)	<0.001	71.0 (62.0, 82.0)	74.0 (65.0, 85.0)	0.004
Body Mass Index (kg/sq. m)		27.5 (24.3, 32.0)	27.5 (24.5, 31.9)	27.4 (23.3, 32.2)	0.13	27.3 (23.8, 31.8)	27.8 (24.7, 32.2)	0.07
Medication Usage	Aspirin (%)	43.8	42.7	45.8	0.30	34.5	56.1	<0.001
	Thienopyridine (%)	11.4	11.0	12.1	0.53	7.2	16.9	<0.001
	Statin (%)	46.0	46.5	45.3	0.70	37.4	57.4	<0.001
	ACE inhibitor (%)	31.5	28.6	36.7	0.003	23.0	42.7	<0.001
	ARB (%)	13.9	14.0	13.7	0.86	12.8	15.3	0.22
	Spironolactone (%)	8.9	5.9	14.1	<0.001	6.1	12.7	<0.001
	β-blocker (%)	57.2	56.0	59.2	0.27	48.1	69.1	<0.001
	Calcium channel blocker (%)	14.4	14.0	15.2	0.57	14.2	14.7	0.82
	Diuretic (%)	36.1	27.5	51.0	<0.001	27.5	47.4	<0.001
	Digoxin (%)	8.9	6.9	12.1	0.002	6.3	12.1	<0.001
Cardiac Pathology	Significant Coronary Disease (%)	21.3	18.3	26.4	0.001	6.3	40.7	<0.001
	Nonischemic Cardiomyopathy (%)	12.5	10.1	16.7	0.001	9.5	16.9	<0.001
	Aortic stenosis (%)	10.7	12.0	8.4	0.050	7.9	14.3	<0.001
	Aortic insufficiency (%)	4.2	4.4	3.9	0.66	5.5	2.6	0.01
	Mitral stenosis (%)	0.9	0.6	1.3	0.22	1.3	0.4	0.09
	Mitral insufficiency (%)	9.8	9.1	11.0	0.26	8.9	11.0	0.21
	Hypertrophic cardiomyopathy (%)	4.1	4.7	3.0	0.16	3.1	5.3	0.048

^*All values are median (interquartile ranges) or proportions.

HTN = hypertension, HLD = hyperlipidemia, MI = myocardial infarction

Table 2.

^* Cardiac magnetic resonance characteristics.

			Overall	Normal ECV	Elevated ECV		Scar (−)	Scar (+)
			(N=1,210)	(n=766)	(n=444)	p-value	(n=686)	(n=524)	p-value
Left Ventricle	EDVi (mL/sq. m)		71.1 (56.8, 96.8)	69.0 (56.3, 89.1)	77.5 (57.8, 109.1)	<0.001	67.3 (54.7, 85.0)	79.6 (60.9, 112.3)	<0.001
	ESVi (mL/sq. m)		25.9 (16.5, 49.8)	24.2 (16.2, 40.3)	33.5 (17.3, 66.4)	<0.001	21.7 (15.4, 33.0)	38.9 (21.2, 71.6)	<0.001
	SVi (mL/sq. m)		40.5 (32.9, 48.7)	42.2 (35.0, 49.4)	37.6 (29.9, 46.9)	<0.001	43.1 (35.3, 51.3)	37.1 (30.0, 44.9)	<0.001
	EF (%)		62.0 (44.0, 72.0)	64.0 (52.0, 72.0)	57.0 (34.0, 70.0)	<0.001	67.0 (59.0, 74.0)	50.0 (31.0, 67.0)	<0.001
Right Ventricle	EDVi (mL/sq. m)		66.1 (53.3, 81.2)	65.4 (53.2, 79.0)	67.7 (53.8, 86.4)	0.02	67.6 (55.6, 82.1)	64.1 (50.7, 80.0)	0.01
	ESVi (mL/sq. m)		29.7 (22.5, 40.6)	28.7 (21.7, 37.6)	32.2 (23.6, 49.9)	<0.001	29.2 (23.0, 38.6)	30.5 (21.7, 43.9)	0.20
	SVi (mL/sq. m)		34.7 (27.2, 42.7)	35.8 (28.5, 43.3)	32.8 (25.2, 40.6)	<0.001	37.2 (29.8, 44.8)	31.3 (24.7, 38.7)	<0.001
	EF (%)		55.0 (47.0, 61.0)	56.0 (49.0, 62.0)	51.0 (40.0, 60.0)	<0.001	56.0 (50.0, 61.0)	51.0 (41.0, 60.0)	<0.001
Left Atrium	indexed Volume (mL/sq. m)		49.2 (36.9, 63.7)	46.0 (35.8, 59.1)	55.6 (40.6, 72.8)	<0.001	46.7 (36.1, 60.9)	51.4 (38.4, 66.7)	<0.001
Left Ventricle	MMi (gm/sq. m)		70.8 (57.8, 91.2)	69.4 (57.5, 88.4)	74.2 (58.4, 95.5)	0.047	64.0 (52.4, 79.2)	83.3 (66.3, 103.4)	<0.001
	Scar Burden (% myocardium)		0 (0, 3)	0 (0, 2)	2 (0, 6)	<0.001	NA	NA
	Any Scar Presence (%)		43.4	37.2	54.1	<0.001	NA	NA
	CAD Scar Presence (%)		17.0	14.4	21.4	<0.001	NA	NA
	ECV without Scar (% myocardium)	1.5T	30.0 (27.3, 33.3)	27.3 (25.7, 28.8)	33.4 (31.7, 35.5)	NA	29.2 (26.6, 32.6)	31.1 (28.4, 34.3)	<0.001
		3.0T	28.3 (26.1, 30.7)	27.3 (25.7, 28.9)	33.1 (31.9, 35.5)	NA	27.8 (25.8, 30.1)	28.9 (26.8, 31.7)	<0.001
	Elevated ECV Prevalence (%)		462 (36.8%)			NA	212 (29.9%)	250 (45.9%)	<0.001

^*All values are median (interquartile ranges) or proportions.

EDVi = indexed end diastolic volume; ESVi = indexed end systolic volume; SVi = indexed stroke volume; EF = ejection fraction; MMi = indexed myocardial mass; CAD = coronary artery disease; ECV = extracellular volume fraction

Over a median follow-up of 26.3 (interquartile range [IQR] 15.9, 37.5) months, there were 265 first composite events (140 all-cause deaths, 149 heart failure hospitalizations). We identified an optimal ECV cutoff to be 31% for the studied population by ROC analysis. Because of the similarity of this cutoff with the original cutoffs derived from the population of healthy volunteers, we elected to keep the original cutoffs for subsequent analyses. On risk adjusted survival curves, patients with elevated scar-free ECV measures were at continuously increased risk for a first composite event, death, or heart failure hospitalization compared with patients with normal scar-free ECV measures (Supplemental Figure 2, all log-rank p <0.001). When stratified by scar present vs. absent, this relationship persisted between patients with elevated vs. normal ECV for first composite events (Figure 3, Supplemental Figure 3). Conversely, patients without any scar and normal ECV measures had the best first composite event-free survival compared with patients with scar present and/or elevated ECV (Supplemental Figure 3).

Figure 3.

Kaplan-Meier survival curves stratified by scar absent (−) vs. present (+) for first composite event (A, B), mortality (C, D), and heart failure hospitalizations (E, F) with patients dichotomized by normal vs. elevated extracellular volume fraction.

In a parsimonious model developed with Bayesian model averaging,^{37, 38} the following covariates were identified as potential cohort-specific predictors of heart failure events: age, non-white race, hypertension, diabetes, systolic blood pressure, heart rate, diuretic and insulin usage, aortic stenosis, aortic insufficiency, LVEF, and scar burden. Even after addition of these covariates to the model, elevated scar-free ECV remained a significant predictor of a first composite event (hazard ratio [HR] 1.70 [95% confidence interval (CI) 1.31, 2.20]; p <0.001), mortality (HR 1.84 [95% CI 1.29, 2.62]; p=0.001), and heart failure hospitalization (HR 1.54 [95% CI 1.09, 2.19]; p=0.02) (Supplemental Table 2).

Individuals having higher ECV values or elevated vs. normal ECV categories had significantly increased risk for a subsequent first composite event, mortality, or heart failure hospitalization (Table 3). Except for increasing continuous ECV values for heart failure hospitalization, this relationship persisted even after multivariable model adjustments (Table 3). These associations with heart failure outcomes persisted regardless of whether myocardial segments with non-CAD scar only or any scar (CAD or non-CAD) were included in the ECV derivation (Supplemental Tables 3). In patients without any detectable scar, the competing risk of heart failure hospitalization was significantly higher in patients with elevated ECV than in patients with normal ECV [cause-specific cumulative incidence at 5 years: 19.6% versus 8.2%, respectively; SHR: 2.53 (95% CI 1.45, 4.43), p=0.001]. Although associated with a significant hazard risk in a fully adjusted model (Supplemental Table 4), Similar analysis in patients with any detectable scar found no significant difference in the competing risk of heart failure hospitalization between patients with elevated ECV versus those with normal ECV (Figure 4).When the population was further stratified between those with preserved (≥50%) vs. reduced (<50%) LVEF, a significant association between increasing ECV values or having elevated ECV was maintained only for those with preserved LVEF, but not for those with reduced LVEF (Supplemental Table 5). No significant difference was found for the interactions between major disease subtypes (such as valvular heart disease, coronary artery disease and hypertrophic cardiomyopathy), ECV and outcomes.

Table 3.

Full multivariable Cox proportional hazard models for first composite event, mortality, and heart failure hospitalization.

0	First composite event				Mortality				Heart failure hospitalization
	Model without ECV		Model + ECV		Model without ECV		Model + ECV		Model without ECV		Model + ECV
	Adj. HR		Adj. HR	p-value	Adj. HR		Adj. HR		Adj. HR		Adj. HR
	(95% CI)	p-value	(95% CI)		(95% CI)	p-value	(95% CI)	p-value	(95% CI)		(95% CI)	p-value
Modeling with binary ECV Category (Elevated vs. Normal)
Elevated vs. Normal	--	--	1.73 (1.34, 2.24)	<0.001	--	--	1.82 (1.28, 2.59)	0.001	--	--	1.60 (1.12, 2.27)	0.01
Age (per yr)	1.03 (1.02, 1.04)	<0.001	1.03 (1.02, 1.04)	<0.001	1.03 (1.02, 1.05)	<0.001	1.03 (1.02, 1.05)	<0.001	1.02 (1.01, 1.04)	<0.001	1.02 (1.01, 1.04)	<0.001
Male	0.95 (0.72, 1.25)	0.71	0.98 (0.74, 1.29)	0.88	1.02 (0.71, 1.48)	0.91	1.04 (0.72, 1.50)	0.85	0.81 (0.56, 1.16)	0.25	0.83 (0.57, 1.20)	0.31
Black	1.48 (1.08, 2.04)	0.02	1.37 (0.99, 1.88)	0.054	1.27 (0.81, 2.00)	0.30	1.16 (0.74, 1.84)	0.51	1.55 (1.03, 2.34)	0.04	1.43 (0.95, 2.17)	0.09
Body mass index (per kg/m2)	1.00 (0.98, 1.02)	0.78	1.00 (0.98, 1.02)	0.83	0.98 (0.96, 1.01)	0.27	0.98 (0.96, 1.01)	0.29	1.01 (0.99, 1.04)	0.27	1.01 (0.99, 1.04)	0.31
Systolic blood pressure (per mm Hg)	0.99 (0.98, 1.00)	0.003	0.99 (0.98, 0.99)	0.02	0.98 (0.97, 1.00)	0.004	0.99 (0.98, 1.00)	0.02	0.99 (0.98, 1.00)	0.24	1.00 (0.99, 1.01)	0.42
Heart rate (per bpm)	1.02 (1.01, 1.03)	<0.001	1.02 (1.01, 1.03)	<0.001	1.02 (1.01, 1.03)	0.001	1.02 (1.01, 1.03)	0.003	1.01 (1.00, 1.03)	0.01	1.01 (1.00, 1.03)	0.01
History of current smoking	1.13 (0.75, 1.71)	0.57	1.11 (0.73, 1.68)	0.62	1.70 (0.99, 2.91)	0.06	1.65 (0.96, 2.84)	0.07	0.60 (0.31, 1.14)	0.12	0.59 (0.31, 1.13)	0.11
History of previous smoking	0.97 (0.74, 1.27)	0.81	1.00 (0.76, 1.31)	0.98	0.98 (0.67, 1.43)	0.92	1.02 (0.70, 1.48)	0.92	0.89 (0.62, 1.28)	0.52	0.91 (0.63, 1.31)	0.60
History of diabetes	1.80 (1.38, 2.37)	<0.001	1.72 (1.31, 2.26)	<0.001	1.91 (1.32, 2.75)	<0.001	1.81 (1.25, 2.61)	0.002	1.61 (1.12, 2.32)	0.01	1.54 (1.07, 2.23)	0.02
History of myocardial infarction	0.99 (0.70, 1.39)	0.94	0.94 (0.67, 1.33)	0.73	1.29 (0.81, 2.04)	0.28	1.23 (0.77, 1.94)	0.38	0.90 (0.56, 1.43)	0.65	0.86 (0.54, 1.37)	0.53
History of anti-HTN medication use	0.74 (0.36, 1.50)	0.40	0.72 (0.36, 1.48)	0.38	0.85 (0.31, 2.30)	0.74	0.81 (0.30, 2.20)	0.68	0.90 (0.33, 2.46)	0.84	0.90 (0.33, 2.44)	0.83
LV EF (per %)	0.99 (0.98, 1.00)	0.02	0.99 (0.98, 1.00)	0.08	1.00 (0.99, 1.01)	0.92	1.00 (0.99, 1.01)	0.82	0.98 (0.97, 0.99)	<0.001	0.98 (0.97, 0.99)	<0.001
LV MMi (per gm/m2)	1.01 (1.00, 1.01)	0.02	1.01 (1.00, 1.01)	0.03	1.00 (0.99, 1.01)	0.90	1.00 (0.99, 1.01)	0.91	1.01 (1.00, 1.01)	0.01	1.01 (1.00, 1.01)	0.01
Scar burden (per %)	1.01 (1.00, 1.03)	0.16	1.01 (1.00, 1.03)	0.18	1.00 (0.97, 1.02)	0.81	1.00 (0.97, 1.02)	0.79	1.01 (0.99, 1.03)	0.19	1.01 (0.99, 1.03)	0.20
Modeling with continuous ECV (per 5% increments)
ECV, per 5% increments	--	--	1.28 (1.11, 1.47)	0.001	--	--	1.40 (1.16, 1.70)	0.001	--	--	1.19 (0.98, 1.44)	0.08
Age (per yr)	1.03 (1.02, 1.04)	<0.001	1.03 (1.02, 1.04)	<0.001	1.03 (1.02, 1.05)	<0.001	1.03 (1.02, 1.05)	<0.001	1.02 (1.01, 1.04)	<0.001	1.02 (1.01, 1.04)	0.001
Male	0.95 (0.72, 1.25)	0.71	0.98 (0.74, 1.28)	0.86	1.02 (0.71, 1.48)	0.91	1.06 (0.73, 1.54)	0.75	0.81 (0.56, 1.16)	0.25	0.81 (0.56, 1.18)	0.27
Black	1.48 (1.08, 2.04)	0.02	1.34 (0.97, 1.85)	0.08	1.27 (0.81, 2.00)	0.30	1.19 (0.75, 1.88)	0.46	1.55 (1.03, 2.34)	0.04	1.37 (0.90, 2.09)	0.15
Body mass index (per kg/m2)	1.00 (0.98, 1.02)	0.78	1.00 (0.98, 1.02)	0.68	0.98 (0.96, 1.01)	0.27	0.98 (0.96, 1.01)	0.25	1.01 (0.99, 1.04)	0.27	1.02 (0.99, 1.04)	0.19
Systolic blood pressure (per mm Hg)	0.99 (0.98, 1.00)	0.003	0.99 (0.98, 1.00)	0.01	0.98 (0.97, 1.00)	0.004	0.99 (0.98, 1.00)	0.01	0.99 (0.98, 1.00)	0.24	0.99 (0.98, 1.00)	0.30
Heart rate (per bpm)	1.02 (1.01, 1.03)	<0.001	1.02 (1.01, 1.02)	<0.001	1.02 (1.01, 1.03)	0.001	1.01 (1.00, 1.03)	0.01	1.01 (1.00, 1.03)	0.01	1.01 (1.00, 1.02)	0.02
History of current smoking	1.13 (0.75, 1.71)	0.57	1.17 (0.77, 1.77)	0.47	1.70 (0.99, 2.91)	0.06	1.71 (0.99, 2.93)	0.053	0.60 (0.31, 1.14)	0.12	0.62 (0.33, 1.18)	0.15
History of previous smoking	0.97 (0.74, 1.27)	0.81	1.04 (0.79, 1.37)	0.80	0.98 (0.67, 1.43)	0.92	1.05 (0.72, 1.53)	0.80	0.89 (0.62, 1.28)	0.52	0.95 (0.65, 1.37)	0.78
History of diabetes	1.80 (1.38, 2.37)	<0.001	1.79 (1.36, 2.34)	<0.001	1.91 (1.32, 2.75)	0.001	1.83 (1.26, 2.64)	0.001	1.61 (1.12, 2.32)	0.01	1.64 (1.14, 2.36)	0.01
History of myocardial infarction	0.99 (0.70, 1.39)	0.94	0.90 (0.62, 1.29)	0.56	1.29 (0.81, 2.04)	0.28	1.19 (0.74, 1.90)	0.47	0.90 (0.56, 1.43)	0.65	0.82 (0.50, 1.35)	0.44
History of anti-HTN medication use	0.74 (0.36, 1.50)	0.40	0.82 (0.39, 1.76)	0.61	0.85 (0.31, 2.30)	0.74	0.82 (0.30, 2.24)	0.70	0.90 (0.33, 2.46)	0.84	1.18 (0.37, 3.73)	0.78
LV EF (per %)	0.99 (0.98, 1.00)	0.02	0.99 (0.98, 1.00)	0.03	1.00 (0.99, 1.01)	0.92	1.00 (0.99, 1.01)	0.93	0.98 (0.97, 0.99)	<0.001	0.98 (0.97, 0.99)	<0.001
LV MMi (per gm/m2)	1.01 (1.00, 1.01)	0.02	1.01 (1.00, 1.01)	0.02	1.00 (0.99, 1.01)	0.90	1.00 (0.99, 1.01)	0.89	1.01 (1.00, 1.01)	0.01	1.01 (1.00, 1.01)	0.01
Scar burden (per %)	1.01 (1.00, 1.03)	0.16	1.01 (0.99, 1.03)	0.17	1.00 (0.97, 1.02)	0.81	0.99 (0.97, 1.02)	0.62	1.01 (0.99, 1.03)	0.19	1.02 (0.99, 1.04)	0.15

ECV, extracellular volume fraction; Adj. HR = adjusted hazard ratio; LV EF, left ventricular ejection fraction; LVMMi, indexed myocardial mass; HR, hazard ratio; CI, confidence interval

Figure 4.

Competing risk analysis for heart failure hospitalization over a 5-year follow-up for the entire cohort (A), patients with scar absent (−) (B), and patients with scar present (+) (C). SHR = subdistribution hazard ratio.

The addition of scar-free ECV both as a continuous variable or as a binary category to a hazard risk model inclusive of heart failure risk variables and imaging markers, resulted in a marginal improvement in c-statistic for first composite events (p=0.01 for both, Table 4, Supplemental Tables 6 & 7), suggesting an incremental improvement to the HF model to discriminate for future composite events. The incremental benefit when scar-free ECV as a binary category was added to the full HF model was not statistically significant for mortality (p=0.07) and heart failure hospitalization (p=0.054) (Table 4). The addition of scar-free ECV as a continuous variable to the HF model resulted in incremental improvement in discrimination for mortality (p=0.04) but not heart failure hospitalization events (p=0.05) (Table 4). The loss of incremental benefit to the c-statistic for these two outcomes was seen with the addition of covariates for the ARIC heart failure equivalent model (Supplemental Table 7).

Table 4.

Discrimination power gain by adding ECV to the final Cox proportional hazard models.

	First composite event		Mortality		Heart failure hospitalization
	Model without ECV	Model + ECV	Model without ECV	Model + ECV	Model + ECV	Model without ECV
Modeling with binary ECV Category (Elevated vs. Normal)
C-statistic (95% CI)	0.72 (0.69, 0.75)	0.74 (0.71, 0.77)	0.72 (0.67, 0.76)	0.74 (0.70, 0.77)	0.76 (0.73, 0.80)	0.77 (0.74, 0.81)
Discrimination improvement (95% CI)	0.018 (0.006, 0.030), p=0.01		0.019 (−0.001, 0.040), p=0.07		0.011 (−0.0002, 0.023), p=0.054
NRI (95% CI)	0.519 (0.390, 0.660)		0.497 (0.313, 0.664)		0.463 (0.262, 0.633)
IDI (95% CI)	0.012 (0.001, 0.030)		0.006 (−0.002, 0.025)		0.005 (−0.001, 0.020)
Modeling with continuous ECV (per 5% increments)
C-statistic (95% CI)	0.72 (0.69, 0.76)	0.74 (0.71, 0.77)	0.72 (0.67, 0.76)	0.74 (0.70, 0.78)	0.77 (0.73, 0.80)	0.77 (0.74, 0.81)
Discrimination improvement (95% CI)	0.014 (0.005, 0.024), p=0.01		0.022 (0.001, 0.043), p=0.04		0.007 (0.0001, 0.015), p=0.048
NRI (95% CI)	0.519 (0.390, 0.653)		0.497 (0.336, 0.664)		0.463 (0.292, 0.627)
IDI (95% CI)	0.013 (0.002, 0.032)		0.007 (−0.001, 0.024)		0.005 (−0.001, 0.022)

ECV, extracellular volume fraction; Discrimination improvement = C-statistic difference between the model with ECV and model without ECV; NRI, Net Reclassification Improvement; IDI, Integrated Discrimination Improvement; CI, confidence interval

However, by model goodness-of-fit scores, the addition of scar-free ECV as a categorical variable to a model with all variables (i.e., ARIC HF risk variables, imaging markers) resulted in a significant increase in likelihood ratio X² for all outcomes (Figure 5), suggestive of an increased model fit. Similarly, the addition of scar-free ECV as a continuous variable also resulted in a significant increase in likelihood ratio X² for all outcomes (X²=165.0, p <0.001 for first composite event; X²=84.8, p <0.001 for all-cause mortality; p-values comparing models adjusted for heart failure risk variables and imaging markers with vs. without continuous ECV), except heart failure hospitalization (X²=119.4, p=0.08).

Figure 5.

Global likelihood ratio X2 goodness-of-fit of each multivariate model by outcomes, with asterisks indicating a significant (p <0.05) improvement from the preceding model, with ECV as a binary variable (normal vs. elevated) added in the last model.

Lastly, when the cohort was dichotomized into scanner-specific abnormal (i.e., outside the 95% confidence interval for native T1 times in the reference group of healthy volunteers) vs. normal myocardial native T1 times, binary categorization by abnormal vs. normal native T1 times was found not be associated with heart failure outcomes on substitution for binary ECV category or in a complete model including binary ECV category (Supplemental Tables 8–12).

Discussion

We derived ECV measures of diffuse myocardial fibrosis, excluding any form of focal replacement scar, and tested the association of these measures with heart failure outcomes in a cardiovascular MRI referral base. Our key findings were that: a) scar-free ECV was associated with heart failure outcomes in a general population independent of any form of replacement scar, b) scar-free ECV had incremental benefit when added to an existing prediction model for incident heart failure. In contrast, similar analyses with abnormal vs. normal native myocardial T1 times in our cohort resulted in no significant associations for native T1 times with heart failure outcomes, when substituted for or added to myocardial ECV. In our study, we excluded patients with any infiltrative cardiomyopathies and cardiac masses, groups known to portend a grim prognosis and have a heart failure risk profile less reflective of a general population. Also, our patients were roughly distributed equally between scanners of different magnet strengths, which was controlled for in the derivation of magnet-specific cutoffs for normal vs. elevated ECV values. Although we observed a significant association between ECV values and heart failure outcomes only in patients with preserved LV systolic function, we were likely underpowered to detect such an association in patients with reduced LV systolic function.

The myocardial extracellular matrix has become increasingly recognized as an active component in various cardiac diseases.¹⁶ Although myofibroblasts have been implicated as the main effector cell in the matrix, the environment includes many other cell actors and active biomolecular pathways such as chemokines, endothelin-1, the renin-angiotensin-aldosterone system, matricellular proteins and inhibitors, and growth factors, making it a target rich environment for assessing the effects of drug and procedural interventions.⁴⁰ With advances that now make noninvasive measures of the myocardial extracellular matrix possible, future studies of such interventions will likely rely on CMR techniques increasingly for such assessments. Indeed, results from a recent study examining valve replacement for aortic stenosis have lent support to the hypothesis of a dynamic extracellular matrix environment.⁴¹ However, results from large population-based studies have only recently begun to emerge, to support thresholds above which ECV measures may be clinically relevant and pathologic.

Consistent with findings in prior studies which included non-CAD scar in their ECV derivations, we found that elevated scar-free ECV measures were associated with an increased risk for mortality and heart failure hospitalizations but over a longer follow-up period with more events.^{13, 17–20} Scar, or replacement fibrosis, has been previously shown to be prognostic of cardiovascular outcomes in the ischemic and non-ischemic cardiomyopathy settings.¹⁶ Although the current consensus for ECV measures allows for inclusion of non-CAD scar in ECV derivations, the overlap of ECV and non-CAD scar in existing studies has left questions over whether ECV truly represents a novel prognostic marker independent of non-infarction scar in the myocardium.⁴² Thus, our study is among the first to demonstrate ECV measures as a prognostic risk marker for incident heart failure events among survivors, completely independent of myocardial scar tissue. Our findings support the utility of ECV assessment over and above conventional LGE assessment in a general CMR referral base.

As a potential risk marker for heart failure events, data are currently limited on the inclusion of ECV measures into existing heart failure models. Indeed, the selection of an appropriate heart failure risk model for clinical practice guidelines remains a challenge. Although a recent systematic review identified a plethora of such models in the literature, many were identified to have methodological limitations and lack of external validation.⁴³ We selected the Atherosclerosis Risk in Communities (ARIC) model because it made use of risk variables readily obtainable in most clinical settings.³⁶ A majority of the ARIC covariates were similarly identified in this systematic review as common to many other risk models for incident heart failure.⁴³ Even after accounting for the ARIC risk model, LVEF, LV myocardial mass, and scar burden, we found elevated scar-free ECV to maintain a significant association with heart failure events. When added back to these models, ECV categorization provided a significant incremental improvement to the models for predicting heart failure outcomes.

Criteria have been previously put forward by the American Heart Association as necessary for the establishment of a novel risk marker for risk evaluation.⁴⁴. Besides proper definitions, these steps include 1) establishing a statistically significant association between the marker and outcomes of interest and 2) demonstrating incremental prognostic information beyond established risk markers.^{13, 17–19} Prior large population-based studies of ECV on heart failure risk assessment have fulfilled the first criterion but not unequivocally the second. Our study is among the first to use an established risk model for incident heart failure events. Further studies, including multicenter collaborations, are still needed to more firmly establish ECV as a risk marker ready for use in the clinical setting.

Limitations

The study was not without limitations. As a prospective cohort study, we accounted for many variables associated with heart failure outcomes at baseline, but unanticipated variables associated with heart failure outcomes may not have been captured at initial enrollment. For example, natriuretic peptide assessment was not always clinically indicated in every patient and thus unavailable in a majority of patients. Furthermore, more objective measures of function and symptoms such as 6-minute walk tests or cardiomyopathy symptom questionnaires were not administered in the routine clinical CMR setting. A single mid ventricular short axis slice was used to derive ECV and may not fully represent global myocardial tissue characteristics, an issue that could introduce variability and diminish the link the between ECV measures and heart failure outcomes. Despite this limitation, we still demonstrated a significant association between ECV measures and subsequent heart failure outcomes even after adjustments for covariates. Further, the group of normal volunteers was admittedly younger than the cohort examined, and increasing ECV values has been previously been shown to be associated with older age. Despite that limitation, our ROC analysis identified a similar ECV cutoff as the upper 95% limit for the normal volunteers for discriminating a first composite event within our cohort. Every reasonable effort was made to determine the vital status of participants and to contact participants regarding interval clinical events. However, outcomes of interest may still be missed despite our best efforts. We also applied a model developed for a general population to a CMR referral base. Despite this issue, we still found that the ARIC heart failure risk model to have a reasonable discriminative ability for future heart failure outcomes in our population with c-statistics >0.70 for all studied outcomes, even before the addition of imaging parameters to the risk models. Lastly, we used a CMR referral base at a tertiary care center, which may have a selection bias not only for more prevalent cardiac pathologies but also for less severe renal disease, because of the requirement for gadolinium contrast use. Thus, our study population may be less generalizable to other populations.

Conclusions

We found that ECV derived independently of myocardial replacement fibrosis of any type was significantly associated with subsequent heart failure events. Even after stratifying for the presence of myocardial replacement fibrosis anywhere in the heart, this association remained. When ECV measures were tested as an imaging risk marker alongside an established risk model for incident heart failure and other imaging risk markers, it provided incremental improvement to these models. Further studies are needed to establish more generalizable criteria for clinically relevant elevations in ECV.

Abbreviation List

AHA

American Heart Association

ARIC

Atherosclerosis Risk in Communities Study

CAD

coronary artery disease

CMR

cardiovascular magnetic resonance

ECG

electrocardiography

ECV

extracellular volume fraction

EHR

electronic health record

LGE

late gadolinium enhancement

LV

left ventricular

MI

myocardial infarction

MOLLI

modified Look-Locker inversion recovery

SSFP

steady-state free precession

TI

inversion time

VHD

valvular heart disease