Temporal Relation Between Myocardial Fibrosis and Heart Failure With Preserved Ejection Fraction: Association With Baseline Disease Severity and Subsequent Outcome

Key Points

Question  Does myocardial fibrosis occur during the evolution of heart failure with preserved ejection fraction (HFpEF) and is it associated with disease severity and outcome in those with HFpEF or at risk for HFpEF?

Findings  In this cohort study of 410 patients at risk for or with a diagnosis of HFpEF, myocardial fibrosis quantified by extracellular volume was associated with baseline brain-type natriuretic peptide level (disease severity surrogate) in linear regression models, and outcomes of heart failure hospitalization or death in Cox models.

Meaning  Among myriad changes in evolving HFpEF, myocardial fibrosis is prevalent and was associated with disease severity and adverse outcomes, so whether the cells and secretomes mediating myocardial fibrosis represent therapeutic targets in HFpEF warrants further evaluation.

Importance  Among myriad changes occurring during the evolution of heart failure with preserved ejection fraction (HFpEF), cardiomyocyte–extracellular matrix interactions from excess collagen may affect microvascular, mechanical, and electrical function.

Objective  To investigate whether myocardial fibrosis (MF) is similarly prevalent both in those with HFpEF and those at risk for HFpEF, similarly associating with disease severity and outcomes.

Design, Setting, and Participants  Observational cohort study from June 1, 2010, to September 17, 2015, with follow-up until December 14, 2015, at a cardiovascular magnetic resonance (CMR) center serving an integrated health system. Consecutive patients with preserved systolic function referred for CMR were eligible. Cardiovascular magnetic resonance was used to exclude patients with cardiac amyloidosis (n = 19).

Exposures  Myocardial fibrosis quantified by extracellular volume (ECV) CMR measures.

Main Outcome and Measures  Baseline BNP; subsequent hospitalization for heart failure or death.

Results  Of 1174 patients identified (537 [46%] female; median [interquartile range {IQR}] age, 56 [44-66] years), 250 were “at risk” for HFpEF given elevated brain-type natriuretic peptide (BNP) level; 160 had HFpEF by documented clinical diagnosis, and 745 did not have HFpEF. Patients either at risk for HFpEF or with HFpEF demonstrated similarly higher prevalence/extent of MF and worse prognosis compared with patients with no HFpEF. Among those at risk for HFpEF or with HFpEF, the actual diagnosis of HFpEF was not associated with significant differences in MF (median ECV, 28.2%; IQR, 26.2%-30.7% vs 28.3%; IQR, 25.5%-31.4%; P = .60) or prognosis (log-rank 0.8; P = .38). Over a median of 1.9 years, 61 patients at risk for HFpEF or with HFpEF experienced adverse events (19 hospitalization for heart failure, 48 deaths, 6 with both). In those with HFpEF, ECV was associated with baseline log BNP (disease severity surrogate) in multivariable linear regression models, and was associated with outcomes in multivariable Cox regression models (eg, hazard ratio 1.75 per 5% increase in ECV, 95% CI, 1.25-2.45; P = .001 in stepwise model) whether grouped with patients at risk for HFpEF or not.

Conclusions and Relevance  Among myriad changes occurring during the apparent evolution of HFpEF where elevated BNP is prevalent, MF was similarly prevalent in those with or at risk for HFpEF. Conceivably, MF might precede clinical HFpEF diagnosis. Regardless, MF was associated with disease severity (ie, BNP) and outcomes. Whether cells and secretomes mediating MF represent therapeutic targets in HFpEF warrants further evaluation.

Among myriad cardiac and noncardiac changes that occur during the evolution of heart failure with preserved ejection fraction (HFpEF), myocardial interstitial disease (MID) from myocardial fibrosis (MF) may precede the clinical diagnosis of HFpEF and be associated with higher disease severity and worse subsequent outcomes. The cells and secretomes involved in MF might therefore represent promising therapeutic targets specific to the heart, especially during the evolution of HFpEF, from being “at risk” for HFpEF—manifest by elevated brain-type natriuretic peptide (BNP) levels—to clinical diagnosis of HFpEF. Myocardial fibrosis indicates interstitial expansion from excess collagen and represents the most common form of MID, but misfolded light chain or transthyretin protein in cardiac amyloidosis (CA) represents a less common, more extreme form of MID, potentially confounding HFpEF studies because CA can be challenging to diagnose. Myocardial fibrosis can affect microvascular function, mechanical function, electrical function, and myocyte energetics,¹ reflecting cardiomyocyte-extracellular matrix interactions beyond the interstitium.² These interactions include (1) capillary rarefaction and perivascular fibrosis³ limiting perfusion reserve,^4-6 (2) myocardial stiffening^7,8 from titin and collagen expansion with increased cross-linking in MF leading to systolic and diastolic dysfunction^6,9-11 and increased filling pressures,⁷ (3) impaired electrical conduction from disarray in the collagen network architecture predisposing to reentrant arrhythmia and sudden death,^12,13 and (4) likely impaired cardiomyocyte or mitochondrial energetics if interposing excess collagen isolates cardiomyocytes from capillaries in the setting of decreased perfusion reserve and myocardial stiffening.¹⁴

Beyond these deleterious MF-cardiomyocyte interactions, further evidence supports MF as a promising therapeutic target during the evolution of HFpEF: MF may be prevalent,¹⁵ strongly associated with outcomes in general cohorts,^15-18 and reversible.^4,9,10,19-21 Yet, HFpEF remains an incompletely understood, etiologically heterogeneous prevalent syndrome in need of efficacious therapies to reduce mortality and hospitalization. Trials of HFpEF using conventional, modestly “antifibrotic” renin-angiotensin-aldosterone system inhibitors^2,14 have mostly had neutral results but have been confounded by substantial methodologic issues, such as the inadvertent inclusion of patients with CA whose condition would not respond to antifibrotic therapy. Targeting specific phenotypes in the spectrum of HFpEF with specific therapies instead of a “one-size-fits-all” treatment approach may become increasingly important to address unmet needs.²²

To investigate MF during the apparent evolution of HFpEF, we enrolled 1174 consecutive patients referred for cardiovascular magnetic resonance (CMR) with preserved systolic function in a single-center observational study. We quantified MF severity using robust extracellular volume (ECV) CMR measures, and grouped patients according to HFpEF categories.¹⁵ We identified clinical HFpEF if patients’ physicians documented heart failure signs and symptoms. In the absence of clinical HFpEF, we identified those “at risk” for HFpEF by elevated BNP levels because BNP specifies cardiac dysfunction and is robustly associated with adverse outcomes. We hypothesized that their extent of MF and clinical course would resemble that of patients with clinically diagnosed HFpEF. Thus, after excluding the small but important subset of HFpEF patients with evident CA among those with HFpEF or at risk for HFpEF, we examined the association of MF with baseline disease severity measures of hemodynamic stress, that is, BNP levels, and subsequent outcomes, namely, the combined end point of hospitalization for heart failure and all-cause mortality.

After approval by the UPMC institutional review board , we recruited 2316 consecutive adult patients referred to the UPMC CMR Center at time of clinical CMR from June 1, 2010, to September 17, 2015, observed until December 14, 2015. Inclusion criteria were written informed consent and completion of a gadolinium contrast–enhanced CMR. Exclusion criteria included (1) hypertrophic cardiomyopathy (n = 221), (2) stress-induced cardiomyopathy (n = 14), (3) adult congenital heart disease (n = 339), (4) inadequate image quality (eg, coil malfunction [n = 4]), and (5) individuals with siderosis (n = 5) or Fabry disease (n = 3). Because the goal of this work was to examine those with preserved systolic function, we also excluded participants with overt systolic dysfunction determined by CMR, defined as left ventricular ejection fraction less than 50% (n = 556). The final cohort included 1174 patients.

Data elements have been described previously.¹⁵ Data were managed using REDCap (Research Electronic Data Capture) hosted at the University of Pittsburgh. Baseline comorbidity data at the time of CMR were determined from the medical record. Investigators classified race.

We divided the cohort into 3 main categories: (1) clinical HFpEF, (2) no clinical HFpEF but at risk for HFpEF given elevated BNP levels (>100 pg/mL; to convert to nanograms per liter, multiply by 1.0) at time of CMR, and (3) neither. Clinical HFpEF diagnosis required medical record documentation of heart failure signs and symptoms from physicians responsible for the patient’s care using a definition from prior epidemiologic studies²³: (1) documented symptoms and physical signs (eg, edema), (2) supporting clinical findings (eg, radiography), or (3) therapy for heart failure (eg, diuresis).²³ First hospitalization for heart failure (HHF) after CMR included any HHF event after CMR scanning (regardless of any prior HHF) applying the same criteria. Vital status was ascertained by means of Social Security Death Index queries and medical record review.

Patients received clinical CMR scans from a 1.5-T scanner (Magnetom Espree, Siemens Medical Solutions). Examinations included standard cine imaging with steady-state free precession as we have described previously.^24,25 Left ventricular volumes, mass, and ejection fraction were measured by experienced readers from short-axis stacks of cine frames that covered the ventricles (6-mm slice, 4-mm gap).

Late gadolinium enhancement (LGE) imaging was performed 10 minutes after a 0.2–mmol/kg intravenous gadoteridol bolus (Prohance, Bracco Diagnostics) with a motion-corrected phase-sensitive inversion recovery pulse sequence²⁵ matching the cine imaging planes. The extent of myocardial infarction and LGE was assessed visually in terms of the extent of LGE (none, <25%, 26%-50%, 51%-75%, >75%), rendering 5 categories for each of the 17 segments to compute extent of LGE.²⁵

We identified CA according to the clinical report, based on prominent diffuse LGE in a nonischemic pattern with other associated features (eg, poor annular motion, diffuse subendocardial enhancement, and increased myocardial thickness).²⁶ Most patients with CA had ancillary biopsy data supporting the CA diagnoses. After CA was excluded, those with elevated ECV were assumed to have MF.

We used reproducible²⁷ and validated ECV measures after a gadolinium bolus described previously.^16,17 We did not exclude foci of nonischemic scar on LGE images (ie, atypical of myocardial infarction) from ECV measures acquired in noninfarcted myocardium,^16,17,28 which would bias ECV measures. We measured the middle third of myocardium to avoid partial-volume effects.

We quantified MF with ECV²⁸ defined as ECV = λ × (1 − hematocrit), where λ = (ΔR1_myocardium)/(ΔR1_bloodpool) before and after administration of gadolinium contrast (where R1 = 1/T1) from basal and mid-ventricular short-axis slices in noninfarcted myocardium as described previously.¹⁵ Hematocrit measures were acquired on the day of scanning. We defined MF as ECV greater than 29%, approximately the upper 95th percentile based on 16 healthy volunteers (median age, 23 years; interquartile range [IQR], 21-33 years),¹⁵ which agreed with prior reports.

χ² tests or Fisher exact tests were used to compare categorical variables. Nonparametric Wilcoxon rank sum tests or Kruskal-Wallis tests were used to compare continuous variables given skewed nonnormal distributions based on the Kolmogorov-Smirnov test. There was no correction for multiple comparisons. Patients with CA (n = 19) were excluded from all multivariable analyses. Linear regression models were used to assess associations with log-transformed BNP levels ignoring 13 participants with missing BNP. Survival analyses examined a combined end point of time to either first HHF or death (all-cause mortality) because ECV shows similar relationships when each event is modeled separately.¹⁵ Kaplan-Meier curves used the log-rank test with ECV categorized arbitrarily in 5% intervals to demonstrate dose-response relationships.

Cox regression analysis was used to examine associations between MF and outcomes in those with clinically diagnosed HFpEF and those at risk for HFpEF (BNP > 100 pg/mL) because these groups were suspected to have similar risk profiles and MF burden. Further analyses were limited to only the cohort with clinically diagnosed HFpEF. Extracellular volume was expressed as a continuous variable (percentage) and reported as a 5% hazard ratio increment to scale the hazard ratio to a clinically meaningful interval. Similarly, all continuous variables in regression models were scaled to clinically meaningful intervals. To benchmark ECV against other clinically important variables, we compared Cox regression χ² values.

Given limited numbers of events, we created 2 principal parsimonious Cox regression models acknowledging alternate valid methodologies. The first “clinical” model attempted thoughtful variable selection informed by clinical judgement, prior literature, and inspection of univariable models, specifically selecting variables representing separate independent disease processes distinct from MF that also may relate to outcomes. The second model used automated “stepwise selection” from the pool of available variables using a typical threshold of P = .10 to enter and remain in the model. To conserve degrees of freedom while maximizing risk adjustment, we constrained the number of covariates to minimize overfitting and stratified all multivariable Cox models by “risk marker” frailty variables such as hospitalization status and hematocrit (categorized as quartiles) that do not illuminate etiology in HFpEF. In analysis stimulated by clinical interest, these models included BNP as a covariate, ignoring the lack of independence between BNP and ECV whereby prognostic associations can be shared among nonindependent variables. We confirmed the proportional hazards assumption. Extracellular volume did not interact with myocardial infarction size, focal nonischemic LGE, or left ventricular mass index. Statistical tests were 2 sided, and P < .05 was considered significant. Statistical analyses were performed using SAS, version 9.4.

The baseline characteristics of the patients with preserved ejection fraction (≥50%) are summarized in Table 1. Those with HFpEF were older and more often female. Median ECV was significantly higher, and MF (ie, ECV > 29%) was more prevalent in those with HFpEF (41% prevalence [n = 65]) or at risk for HFpEF (42% prevalence [n = 106]) compared with those without HFpEF who were not at risk based on BNP levels (25% prevalence [n = 189]). There were no significant differences in either ECV levels or MF prevalence of elevated ECV between those with HFpEF and those at risk for HFpEF, consistent with ECV elevations preceding the clinical diagnosis of HFpEF. Those with clinically diagnosed HFpEF had lower BNP but higher rates of loop diuretic use than those at risk for HFpEF; 93 of 160 HFpEF patients (58%) without evident CA had elevated BNP greater than 100 pg/mL. Among patients with HFpEF (Figure 1), median ECV was highest in the 19 patients with suspected CA (46.2%; IQR, 33.8%-52.6% vs 27.1%; IQR, 25.0%-29.8%; P < .001), who were then excluded from further multivariable analysis.

Extracellular volume measurement of MF was the variable most strongly associated with log-transformed BNP levels in the cohort of patients combining clinically diagnosed HFpEF and those at risk for HFpEF given elevated BNP values (Table 2). Similar ECV-BNP associations were found in only clinically diagnosed HFpEF. Significant associations between log BNP and ECV remained after adjustment in multivariable models.

Over a median follow-up period of 1.9 years (IQR, 0.9-2.9 years), event rates were higher in either those with clinically diagnosed HFpEF (n = 24 [15.0%]) or those at risk for HFpEF given elevated BNP levels (n = 37 [14.8%]) compared with patients without HFpEF or elevated BNP (n = 26 [3.5%]). Yet, similar to the lack of intergroup differences in ECV between those with clinically diagnosed HFpEF and those at risk for HFpEF, there were also no significant differences in the survival analyses (log-rank 0.8; P = .38) (Table 3 and Figure 2). Among these patients, the actual clinical diagnosis of HFpEF was not associated with outcomes.

Combining those with HFpEF or at risk for HFpEF, we observed strong associations between ECV measures of MF and outcomes (Figure 2B and Table 3), where higher ECV was associated with higher event rates in a dose-response fashion. Sixty-one patients experienced events after CMR (19 HHF events and 48 deaths in which 6 patients died after HHF). The patients with CA demonstrated the highest event rates (Figure 2B), demonstrating their ability to confound HFpEF trials through inadvertent inclusion and justifying their exclusion from the main analysis.

Among the various indicators of myocardial disease, MF was among the cardiac variables most strongly associated with adverse outcomes in univariable Cox models and the 2 principal multivariable models (Table 3), even when the cohort included only those with HFpEF. While the clinical diagnosis of HFpEF was not associated with outcomes after combining these 2 groups, ECV provided robust risk stratification.

When BNP was excluded as a covariate in these models (given the lack of independence observed between ECV and BNP), ECV exhibited even stronger associations in multivariable models (data not shown). Extracellular volume was associated with outcomes more strongly than log BNP based on χ² values in univariable models (Table 3) or when both variables were included in a Cox model (17.9 vs 7.8 among 397 patients with HFpEF or at risk for HFpEF). There was no statistical interaction (P = .47). Stepwise variable selection identified ECV as a robust risk stratifier but not BNP.

Our data leveraging CMR to characterize patients with HFpEF generate several novel observations. First, despite the inherent heterogeneity of patients with HFpEF, our results emphasize the potential for abnormalities located specifically in the myocardium—as opposed to the periphery—to mediate disease severity and outcomes in HFpEF. Second, patients with HFpEF or at risk for HFpEF demonstrated high prevalence of elevated BNP and similarly worse prognosis and similarly higher prevalence and extent of MF compared with patients without HFpEF, who fared significantly better. These data imply but do not prove that during the apparent evolution of HFpEF, MF might precede the clinical diagnosis of HFpEF. Indeed, once elevated BNP appeared, the actual clinical diagnosis of HFpEF was not associated with significant differences in either MF or subsequent prognosis among those with HF or at risk for HFpEF, perhaps reflecting the clinical challenge of establishing the HFpEF diagnosis and distinguishing from other comorbidity. Third, MF was strongly associated with (1) myocardial disease severity measured by BNP, and (2) outcomes such as subsequent death and hospitalization for heart failure in proportion to MF severity. In fact, ECV measures provided unprecedented risk stratification in those with HFpEF or those at risk in whom event-free survival curves varied widely according to ECV strata, and we observed a dose-response relationship between MF and outcome, even after adjustment for several important variables.

Despite the “neutral” results of the TOPCAT trial, in which significant results were obtained only in secondary end point or post hoc analyses, the strength of associations between MF and disease severity and subsequent outcome suggests that MF may be a promising therapeutic target for future trials and a causal disease pathway, mediating outcome in HFpEF, and not simply a risk stratifier. Indeed, spironolactone reverses MF in animal models. Several possibilities might explain the TOPCAT results: (1) limited antifibrotic efficacy of spironolactone in humans²; (2) inadvertent inclusion of patients with unsuspected CA (challenging to diagnose without CMR or bone scintigraphy) whose condition would not respond to spironolactone; and (3) heterogeneous patient populations, without adequate prevalence of MF or even HFpEF, diluting therapeutic responses in the overall study. Resolving these issues requires further study, including studying the degree to which therapies regress MF.

Considerable data emphasize the potential for cardiomyocyte–extracellular matrix interactions in MID leading to organ dysfunction and ultimately adverse outcomes. There is biologic precedence in other organs (eg, lung, kidney, liver) where interstitial disease leads to organ dysfunction and vulnerability. Elegant work highlights the etiologic potential of MF in MID as well as its reversibility in animals and humans with resultant improvement in cardiac function.^{4,7,9,10,19,20,29} Cardiac amyloidosis exemplifies well the relationship between MID, disease severity, and outcomes, because this group had the highest ECV, high BNP levels, and the worst outcomes. In HFpEF trials, optimal screening for exclusion of patients with CA, and optimal identification of MF where CMR is not available, for example, with biomarker panels, requires further investigation. Our work builds on a smaller study by Duca et al³⁰ that also suggested a potential relationship between MF and outcomes. Now, our larger data set emphasizes MF as a potential mediator of disease, possibly even preceding the clinical diagnosis of HFpEF although this issue requires further confirmation. Myocardial fibrosis prevalence and its associations with outcomes likely vary across cohorts, which emphasizes the need for personalized medicine: treat MF in those likely to have it.

Our study has limitations. First, associations in single-center observational data do not establish causality and could represent unmeasured confounders perhaps related to referral biases. We did not adjust for the Seattle Heart Failure Model, Heart Failure Survival Score, or Medicare readmission models, but we are uncertain whether these scores derived mostly from those with reduced ejection fraction generalize to HFpEF with adequate discrimination and calibration. Still, we used various multivariable models including many of the same covariates in these risk scores and attempted to maximize risk adjustment with diverse covariates while avoiding overfitting to minimize this possibility. We also attempted to minimize exclusions and maximize the size of the cohort to maximize generalizability. Trials with efficacious antifibrotic therapies are ultimately required to establish a causal role of MF in HFpEF. Second, the definition of HFpEF was primarily clinical, reflecting local practice, which may not be generalizable. Regardless, the clinical recognition of HFpEF offered little for risk stratification. Elevated BNP level seemed to have more robust risk stratification, potentially reflecting how HFpEF symptoms can be subjective, nonspecific, and challenging to discern from other comorbidity. Third, inferences about MF during apparent HFpEF evolution only arose from cross-sectional comparisons in which serial ECV MF measures did not occur; further study is required. Finally, we did not have histologic confirmation for ECV measures of MF, and CA was not always validated with histologic analysis, but ECV is well validated, and the poor survival curves for those with evident CA support their clinical classification.

Our data add to the growing literature promoting MF as a promising therapeutic target in HFpEF trials. Myocardial fibrosis seemed to precede the clinical diagnosis of HFpEF. Extracellular volume MF measures were strongly associated with disease severity and vulnerability to adverse outcomes such as death and HHF in those with HFpEF or at risk for HFpEF manifest by elevated BNP levels. Cardiovascular magnetic resonance detected a small but important CA subgroup with high event rates that could confound trials given its potential to escape clinical recognition. Given the biologic plausibility of MF mediating adverse outcomes, the issue of whether the cells and secretomes underlying MF represent potential therapeutic targets for future HFpEF trials warrants further evaluation.

Corresponding Author: Erik B. Schelbert, MD, MS, Cardiovascular Magnetic Resonance Center, Heart and Vascular Institute, UPMC, University of Pittsburgh School of Medicine, 200 Lothrop St, PUH A349, Pittsburgh, PA 15101 (schelberteb@upmc.edu).

Accepted for Publication: June 9, 2017.

Published Online: August 2, 2017. doi:10.1001/jamacardio.2017.2511

Author Contributions: Dr Schelbert had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Schelbert, Miller, Ugander, Quarta, Senni, Butler, Redfield.

Acquisition, analysis, or interpretation of data: Schelbert, Fridman, Wong, Abu Daya, Piehler, Kadakkal, Miller, Ugander, Maanja, Kellman, Shah, Abebe, Simon, Quarta, Butler, Diez, Gheorghiade.

Drafting of the manuscript: Schelbert, Kadakkal, Senni, Butler.

Critical revision of the manuscript for important intellectual content: Schelbert, Fridman, Wong, Abu Daya, Piehler, Miller, Ugander, Maanja, Kellman, Shah, Abebe, Simon, Quarta, Senni, Butler, Diez, Redfield, Gheorghiade.

Statistical analysis: Schelbert, Fridman, Kadakkal, Abebe.

Obtained funding: Schelbert.

Administrative, technical, or material support: Schelbert, Wong, Abu Daya, Miller, Simon.

Supervision: Schelbert, Abu Daya, Ugander, Quarta, Butler.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Schelbert has accepted contrast material from Bracco Diagnostics for research purposes beyond the scope of this work, and he has served on advisory boards for Bayer Healthcare and Merck. Dr Ugander’s institution has a research and development agreement with Siemens for CMR, outside the present work. Dr Ugander has served on a scientific advisory board of Sanofi Genzyme regarding CMR imaging in Fabry disease with minor compensation. Dr Simon reports personal fees from Gilead, United Therapeutics, Actelion, St. Jude Medical, Hovione, and Third Pole, outside the submitted work. Dr Butler reports other from Amgen, Bayer, Boehringer Ingelheim, Cardiocell, Merck, Novartis, Relypsa, ZS-Pharma, and Astra Zeneca, and grants from the European Union and NHLBI, outside the submitted work. Dr Gheorghiade is a consultant for Abbott Laboratories, AbbVie, Astellas, Astra-Zeneca, Bayer Schering Pharma AG, Cardiocell, Cardiorentis Ltd, CorThera, Inc, Cytokinetics, Inc, DebioPharm SA, Errekappa Terapeutici (Milan, Italy), GlaxoSmithKline, Ikaria, Intersection Medical, Inc, Johnson & Johnson, Medtronic, Merck, Novartis Pharma AG, Ono Pharmaceuticals USA, Otsuka Pharmaceuticals, Palatin Technologies, Pericor Therapeutics, Protein Design Laboratories, Sanofi-Aventis, Sigma Tau, Solvay Pharmaceuticals, Stealth BioTherapeutics, Sticares Inter-ACT, Takeda Pharmaceuticals North America, Inc, and Trevena Therapeutics. No other disclosures are reported.

Funding/Support: This work was supported by a grant from the Pittsburgh Foundation (Pennsylvania), grant M2009-0068 to Dr Schelbert; and an American Heart Association Scientist Development grant (09SDG2180083) including a T. Franklin Williams Scholarship Award; funding provided by Atlantic Philanthropies, Inc, the John A. Hartford Foundation, the Association of Specialty Professors, and the American Heart Association (Dallas, Texas) to Dr Schelbert. Dr Ugander was supported by a grant from Karolinska Institutet. Dr Wong was supported by grant K12 HS19461-01 from the Agency for Healthcare Research and Quality. This work was also supported by grant UL1-TR-001857 from the National Center for Research Resources, a component of the National Institutes of Health (NIH), and NIH Roadmap for Medical Research.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank the patients who volunteered to participate in this study. The authors also wish to acknowledge and thank the following individuals for their assistance in supporting this research: Joan Lacomis, MD, Christopher Deible, MD, PhD, Kathy Puntil, RN, Elizabeth Ruhl, RN, Stephen Mancuso, RT (R)(MR), and Marie-Therese Najjar, RT (R)(MR), BS, of the UPMC Cardiovascular Magnetic Resonance Center. They were not compensated beyond their salaries for their contributions.