Author + information
- Received March 7, 2017
- Revision received July 3, 2017
- Accepted July 27, 2017
- Published online September 25, 2017.
- Abhinav Sharma, MDa,
- Susanna R. Stevens, MSa,
- Joseph Lucas, PhDa,
- Mona Fiuzat, PhDa,
- Kirkwood F. Adams, MDb,
- David J. Whellan, MDc,
- Mark P. Donahue, MD, MHSd,
- Dalane W. Kitzman, MDe,
- Ileana L. Piña, MDf,
- Faiez Zannad, MD, PhDg,
- William E. Kraus, MDd,
- Christopher M. O’Connor, MDh and
- G. Michael Felker, MD, MHSa,∗ ()
- aDuke Clinical Research Institute, Durham, North Carolina
- bDivision of Cardiology, University of North Carolina Chapel Hill, Chapel Hill, North Carolina
- cDivision of Cardiology, Thomas Jefferson University, Philadelphia, Pennsylvania
- dDivision of Cardiology, Duke University School of Medicine, Durham, North Carolina
- eDivision of Cardiology, Wake Forest School of Medicine, Winston-Salem, North Carolina
- fDivision of Cardiology, Albert Einstein College of Medicine/Montefiore Medical Center, Bronx, New York
- gINSERM, CIC 1433, Université de Lorraine, CHRU, Nancy, France
- hInova Heart and Vascular Institute, Fairfax, Virginia
- ↵∗Address for correspondence:
Dr. G. Michael Felker, Duke Clinical Research Institute, 2400 Pratt Street, Durham, North Carolina 27705.
Objectives This study sought to determine the relationship between growth differentiation factor (GDF)-15 and clinical outcomes in ambulatory patients with heart failure and reduced ejection fraction (HFrEF).
Background The prognostic utility of GDF-15, a member of the transforming growth factor-β cytokine family, among patients with HF is unclear.
Methods We assessed GDF-15 levels in 910 patients enrolled in the HF-ACTION (Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training) trial, a randomized clinical trial of exercise training in patients with HFrEF. Median follow-up was 30 months. Cox proportional hazard models assessed the relationships between GDF-15 and clinical outcomes.
Results The median GDF-15 concentration was 1,596 pg/ml. Patients in the highest tertile of GDF-15 were older and had measurements of more severe HF (higher N-terminal pro–B-type natriuretic peptide [NT-proBNP] concentrations and lower peak oxygen uptake on cardiopulmonary exercise testing [CPX]). GDF-15 therapy was a significant predictor of all-cause death (unadjusted hazard ratio [HR]: 2.03 when GDF-15 was doubled; p < 0.0001). This association persisted after adjustment for demographic and clinical and biomarkers including high sensitivity troponin T (hs-TnT) and NT-proBNP (HR: 1.30 per doubling of GDF-15; p = 0.029). GDF-15 did not improve discrimination (as measured by changes in c-statistics and the integrated discrimination improvement) in addition to baseline variables, including hs-TnT and NT-proBNP or variables found in CPX testing.
Conclusions In demographically diverse, well-managed patients with HFrEF, GDF-15 therapy provided independent prognostic information in addition to established predictors of outcomes. These data support a possible role for GDF-15 in the risk stratification of patients with chronic HFrEF. (Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training [HF-ACTION]; NCT00047437)
Chronic heart failure (HF) affects more than 5 million people in the United States and is one of the most common causes of rehospitalization (1–3). The overall prognosis remains poor, with 5-year mortality exceeding 50%, despite advances in therapy (2,4,5). As the number of invasive and noninvasive HF therapies increase, risk stratification and prognostication become essential to identify patients who would most benefit from these treatments. Currently the U.S. Food and Drug Administration has approved 4 biomarkers to aid in the prognostication for HF patients: B-type natriuretic peptide (BNP), N-terminal pro–B-type natriuretic peptide (NT-proBNP), galectin-3, and ST-2. Current HF guidelines indicate that measurements of other clinically available tests such as natriuretic peptides (NT-proBNP and BNP) and markers of myocardial injury (cardiac troponin T or I) may be considered for risk stratification (5). Although a variety of other potential prognostic biomarkers have been identified in HF, their clinical value remains uncertain.
Growth differentiation factor (GDF)-15, a member of the transforming growth factor-β family, is secreted from a range of cells such as adipocytes and myocytes in response to cellular ischemia, mechanical strain, and oxidative stress (6–9). Prior studies have suggested that GDF-15 provides prognostic utility across a spectrum of cardiovascular diseases in addition to existing clinical risk factors and biomarkers (10–15). However, robust data for GDF-15 in well-treated patients with chronic HF are limited, especially in the context of other guideline-recommended HF biomarkers such as natriuretic peptides and high-sensitivity troponin. A prior analysis from the ValHeFT (Valsartan Heart Failure Trial) (16) identified the fact that GDF-15 was independently associated with mortality; however, there was low use of evidence-based medical therapy and a demographically homogenous patient population. An evaluation of the prognostic role of GDF-15 in a cohort of patients more reflective of current HF practice was needed. Furthermore, there was a paucity of data for the association between GDF-15 and other critical measurements of HF status, such as exercise capacity. In the current study, we evaluated, first, the association between GDF-15 and other biological covariates; second, the association between GDF-15 functional status and exercise capacity; and third, the relationship between GDF-15 and clinical outcomes in a well-characterized cohort of ambulatory patients with HF with reduced ejection fraction (HFrEF) enrolled in the HF-ACTION (Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training) trial.
The design, rationale, and primary results of the HF-ACTION (NCT00047437) trial have been previously reported (17,18). Briefly, HF-ACTION was a randomized clinical trial funded by National Heart, Lung, and Blood Institute, evaluating the effects of exercise training in addition to usual care versus usual care alone on long-term morbidity and mortality in patients with symptomatic chronic HF and left ventricular systolic dysfunction (New York Heart Association [NYHA] functional classes II to IV and left ventricular ejection fraction [LVEF] of <35%). Patients were receiving guideline-based HF therapy prior to randomization. The study enrolled 2,331 patients and had a primary composite endpoint of time to all-cause death or all-cause hospitalization. Patients had a median follow-up of 30 months. All deaths and first cardiovascular hospitalizations were adjudicated by a blinded, independent clinical event committee. HF-ACTION was approved by local institutional review boards, and all enrolled patients provided written informed consent.
A subset of patients enrolled in the HF-ACTION study agreed to participate in the biomarker substudy. Blood samples were obtained on the same day as baseline cardiopulmonary exercise (CPX) testing was conducted but before exercise. Samples were collected from the peripheral vein into EDTA-containing tubes, centrifuged immediately, and stored at −70°C for subsequent analysis.
GDF-15 concentrations were measured in a core laboratory from baseline samples (n = 910), using sensitive sandwich-immunoassay monoclonal antibodies (Elecsys GDF-15 assay, Roche Diagnostics, Indianapolis, Indiana). Details of this assay have been previously described (19). The assay has a lower limit of detection of 400 pg/ml, an upper limit of detection of 20,000 pg/ml, an intra-assay coefficient of variation of ≤3.0%, and an interassay coefficient of variation of ≤4.6%. The core laboratory was blinded to all clinical data.
The primary clinical outcome of interest for the current analysis was the relationship between GDF-15 concentrations and all-cause mortality. The secondary outcomes of interest were, first, the composite of all-cause death or all-cause hospitalization (HF-ACTION primary endpoint), and second, the composite of cardiovascular death or HF hospitalization. We also assessed the relationship between baseline GDF-15 and baseline assessments of functional capacity including NYHA functional class, 6-min walk distance, and maximal oxygen consumption (peak Vo2) by cardiopulmonary exercise testing (CPX).
Baseline characteristics are presented as medians (25th, 75th percentiles) and numbers (percentages). The associations between GDF-15 tertile and baseline characteristics were compared using the Kruskal-Wallis test for continuous and ordinal variables and the chi-square test for categorical variables. Event rates by tertile of GDF-15 are shown with Kaplan-Meier curves. To assess the relationship between GDF-15 and baseline clinical and biomarker variables, a model was developed using linear regression with backward elimination and alpha of <0.10 for retention. Modeling for time-to-event variables were done with complete case analysis. There were 716 patients with complete data for the all-cause death model, 646 for the all-cause death or hospitalization model, and 640 for cardiovascular death or HF hospitalization model. GDF-15 was a continuous variable in all models but was log transformed for analysis because it was not normally distributed. For the clinical outcomes of interest, Cox proportional hazards models were used. Hazard ratios (HRs) were calculated for the log2 such that the HR represented the risk per 2-fold greater value of GDF-15; in place of to create a clinically more interpretable reference for the increase in GDF-15.
Adjustment variables included demographics (age, sex, and race) and a comprehensive set of predictors that had previously been identified in the HF-ACTION cohort for each endpoint (1,20) (Online Table 1a). Because CPX testing is not routinely available in some clinical settings, the adjusted models were examined with and without CPX variables. The baseline CPX variables used in the adjustment models included peak Vo2 by Weber class, ventricular conduction abnormality, minute ventilation/carbon dioxide production, and exercise duration (Online Table 1b). Models were repeated with the inclusion of GDF-15 × randomized treatment interaction term to assess for evidence of a differential treatment response to the study intervention based on GDF-15 concentrations. Model discrimination and risk predictions with and without GDF-15 were evaluated by using the c-index, continuous net reclassification index (NRI), and integrated discrimination improvement (IDI). Individual biomarkers were also added to the clinical model separately, and the measurements of discrimination were compared. The proportional hazards assumption was checked for each endpoint in the full models (including demographic, clinical, CPX, and biomarker data), and no deviations were identified. For each model comparison, the summary measurements were reported along with 95% bootstrap confidence intervals (CIs) based on 999 replications.
Data were analyzed using SAS version 9.4 software (SAS, Cary, North Carolina). Statistical significance was based on a p value of ≤0.05.
Baseline characteristics for the HF-ACTION study population are shown in Table 1. The biomarker substudy cohort was broadly similar to the overall HF-ACTION population (data not shown). Generally, the study population was similar to that from other chronic HFrEF clinical trials, with the exception of a high representation of nonwhite patients (34%) and women (29%) in the current study. Overall, the patients were medically well managed: 95% were receiving beta blockers; 94% were receiving angiotensin-converting enzyme inhibitors or angiotensin receptor blockers; and 45% were receiving aldosterone antagonists.
GDF-15 tertiles were identified as follows: lowest tertile was <1,173 pg/ml; middle tertile was 1,173 to 2,252 pg/ml; and the highest tertile was >2,252 pg/ml. Compared to patients in the lowest tertile of GDF-15 values, patients in the highest tertile of GDF-15 were older and had more markers of severe HF, including higher prevalence of comorbidities, higher NYHA functional class, higher NT-proBNP, and higher high sensitivity troponin T (hs-TnT). Use of evidence-based medical therapies tended to be lower in patients with higher GDF-15 levels, and loop diuretic doses tended to be higher, again, consistent with more severe HF in the patients with the highest GDF-15 values. In the overall HF-ACTION cohort, some differences in baseline characteristics were seen among patients who were included and excluded from multivariate modeling due to missing data; however, missing data did not have an impact on clinical outcomes (data not shown).
Biological correlates of GDF-15
The independent variable most strongly associated with higher GDF-15 concentration was higher hs-TnT (p < 0.0001) (Table 2). In addition, older age, lower peak Vo2, and higher creatinine and NT-proBNP concentrations were also found to be associated with higher GDF-15 concentrations after multivariate adjustment.
GDF-15 and symptom status, functional capacity, and exercise performance
Higher GDF-15 concentrations were consistently associated with shorter 6-min walk distances, lower exercise durations, lower peak Vo2, and higher VE/Vo2 slope (Figure 1). Those in the highest tertile of GDF-15 had the lowest exercise duration and lowest peak Vo2 (Table 1). When we assessed higher GDF-15 concentration as a continuous variable, it was associated with a lower baseline peak Vo2 (r = −0.40; p < 0.0001) and shorter 6-min walk distance (r = −0.32; p < 0.0001).
GDF-15 and clinical endpoints
The median follow-up was 32, months and there were 171 deaths, 647 all-cause deaths or hospitalization events, and 301 cardiovascular deaths or HF hospitalization events in our study cohort. Generally, higher tertiles of GDF-15 were associated with worse clinical outcomes across all clinical endpoints of interest (Figure 2). In univariate analyses, a 2-fold increased concentration of GDF-15 was associated with a 203% increase in all-cause mortality, a 32% increase in the composite of all-cause death or all-cause hospitalization, and a 61% increase in the composite of cardiovascular death or HF hospitalization (p < 0.0001) for all.
In order to assess the independent association of GDF-15 in the context of other common clinically available data, we created a series of multivariate models that sequentially included clinical and demographic variables, other guideline-recommended biomarkers (NT-proBNP and hs-TnT), and variables obtained from CPX testing. After we adjusted for demographic and clinical variables only, we found a 2-fold increase in the concentration of GDF-15 was associated with a 71% increased risk of all-cause death (p < 0.0001) (Figure 3A), a 27% increased risk in all-cause death or rehospitalization (p < 0.0001) (Figure 3B), and a 34% increased risk of cardiovascular death or HF rehospitalization (p = 0.0005) (Figure 3C). After additionally adjusting for both hs-TnT and NT-proBNP in addition to clinical and demographic variables, we found a doubling of the baseline concentration of GDF-15 was still associated with a 30% increased risk of all-cause death (p = 0.03) (Figure 3A). After we made further adjustments for CPX variables in addition to the clinical model and biomarkers, GDF-15 was no longer significantly associated with any of the clinical endpoints of interest. There were no significant interactions between GDF-15 and randomized treatment assignment (exercise training vs. control) for any of the clinical endpoints (all: p > 0.14), suggesting that GDF-15 did not identify patients more or less likely to respond to exercise training.
Prognostic utility of GDF-15
The prognostic utility of GDF-15 in addition to the clinical models, biomarkers (NT-proBNP and hs-TnT), and CPX variables are shown in Table 3. GDF-15 improved discrimination of all-cause death in addition to the clinical model, particularly the discrimination of nonevents. When we compared the addition of biomarkers to the clinical model, all biomarker improved discrimination of all-cause death individually. However, NT-proBNP appears to have the greatest magnitude of improvement in discrimination, followed by GDF-15 (Online Table 2). Although the addition of GDF-15 incrementally increased the c-index value for all-cause death in addition to clinical, biomarker, and CPX variables, the improvement was not significant overall. Similarly, GDF-15 did not significantly improve discrimination for the composite outcomes of all-cause death or hospitalization and cardiovascular death or HF hospitalization.
Assessing the clinical utility of new and existing biomarkers remains an important goal in HF. The present analyses from the HF-ACTION study adds a number of important findings regarding the role between GDF-15 and HF: 1) a 2-fold greater concentration of GDF-15 is associated with a 30% greater risk of all-cause death, despite comprehensive multivariate adjustment including hs-TnT and NT-proBNP; 2) greater GDF-15 concentrations were associated with worse symptom burden and functional capacity, including baseline NYHA functional class, 6-min walk distance, and peak Vo2; and 3) the strongest baseline predictor of higher GDF-15 concentrations was greater hs-TnT. Because it was a large multicenter cohort of demographically diverse, ambulatory, chronic HFrEF patients, the HF-ACTION trial allowed for a comprehensive assessment of GDF-15’s association with cardiovascular outcomes. The range of outcomes in HF-ACTION, including CPX data available through HF-ACTION, provides a unique opportunity to evaluate the association between GDF-15, functional capacity, and exercise performance.
Association of GDF-15 with baseline clinical and biomarker variables
Cellular expression of GDF-15 increases in response to various stressors including reactive oxygen species and proinflammatory cytokines (8,21). In murine models, GDF-15 inhibits apoptosis, hypertrophy, and adverse cardiac remodeling (8,9,22). GDF-15 is not normally expressed in the heart; however, it can be induced in experimental models of myocardial ischemia, pressure/volume overload, and dilated cardiomyopathy (8,9). Our study expands upon these findings: we identified the fact that GDF-15 concentrations were closely associated with hs-TnT and NT-proBNP (Table 2), suggesting that cardiac ischemia and cardiomyocyte stretch play a role in GDF-15 expression. Our study also aligns with previous studies suggesting that GDF-15 is closely associated with multiple biologic pathways including renal dysfunction and abnormal glucose regulation (11,16,23,24).
Association between GDF-15 and cardiovascular outcomes
Current HF guidelines indicate that natriuretic peptides and biomarkers of myocardial injury (cardiac troponin T or I) may be considered for ambulatory risk stratification (5). Our findings suggest that GDF-15 may also play an important role in ambulatory risk stratification. In the ValHeFT analysis, after adjustment for clinical and biomarker variables, an increase of 100 pg/ml GDF-15 was associated with an increased risk of all-cause death (HR: 1.007; 95% CI: 1.001 to 1.014; p = 0.02) (16). Compared to that of ValHeFT, our patient population was younger (median 59 vs. 63 years of age, respectively), had increased nonwhite participants (36% vs. 16%, respectively), a higher proportion of patients with hypertension (64% vs. 7%, respectively), lower median GDF-15 concentrations (1,596 pg/ml vs. 2,040 pg/ml, respectively), and greater use of evidence-based medical therapy (use of beta-blocker therapy: 95% vs. 33%, respectively; and use of spironolactone: 45% vs. 2%, respectively). Despite these differences, our study still demonstrated that, in a demographically diverse, well-medically managed group of chronic HF patients, GDF-15 was associated with an increased risk of all-cause mortality, even after adjusting for commonly available demographic, clinical, and laboratory variables.
Compared to other biomarkers, GDF-15 appears to have a robust association with all-cause mortality. Prior studies from HF-ACTION have evaluated the fibrosis marker galectin-3 (25) and the interleukin signaling ligand-soluble ST-2 (26). Galectin-3 was not associated with an increased risk of all-cause death after adjustment for clinical variables and NT-proBNP (adjusted HR: 1.06; p = 0.30) (25). A doubling of ST-2 was associated with increased risk of all-cause death (adjusted HR: 1.42; p = 0.007) (26); however, this study did not adjust for troponin, which is a powerful predictor of mortality in patients with HF (26).
In our study, GDF-15 did not provide any additional prognostic information when CPX variables were added to demographic, clinical, laboratory, and biomarker data. Similar results have been seen with other biomarkers, including ST-2 and galectin-3 (25,26). Our results suggest that the unique prognostic information provided by GDF-15 is also captured by variables derived from CPX testing. Abnormal CPX measurements indicate worse HF and may reduce the association between GDF-15 and cardiovascular outcomes. As CPX testing is not widely available or performed in the routine management of HF patients in clinical practice (27), our results demonstrate the strength of GDF-15’s association with clinical outcomes of interest in addition to routinely available clinical and biomarker variables.
GDF-15 significantly improved discrimination for all-cause mortality in addition to clinical variables; however, when CPX variables or additional biomarkers were added to the clinical model for all-cause death, model discrimination did not improve after the addition of GDF-15. In the ValHeFT study, the addition of GDF-15 to clinical and biomarker variables (BNP, high sensitivity c-reactive protein, and hs-TnT) did improve the c-index for mortality (c-index value from 0.73 to 0.76; p = 0.02), and the IDI for mortality was borderline significant (p = 0.06) (16). Among patients with chronic HF in a Singapore cohort, GDF-15 also improved discrimination in addition to a clinical model with NT-proBNP and hs-TnT (c-index from 0.72 to 0.74; p = 0.0019) (28). In our study, the discriminatory ability of biomarkers (hs-TnT and NT-proBNP) and CPX variables in addition to baseline clinical variables for the outcome of all-cause mortality was already high, as reflected by a c-index of 0.77 (Table 3). Although GDF-15 improved the c-index for all-cause mortality, in addition to baseline biomarkers and CPX variables, this increase was not significant. This finding likely reflects the challenge of improving discrimination in addition to an already robust model. However, combined with the prior analyses of GDF-15 among patients with HF, the totality of evidence suggest that GDF-15 may improve discrimination above and beyond clinical variables and established biomarkers.
This was a post hoc study and was subject to the limitations of these types of analyses. Patients enrolled in the biomarker substudy were predominantly from the United States. As with most clinical trials, our patient population differed from the broader HF population seen in HF registries (29), although HF-ACTION was notable for enrolling a higher proportion of nonwhite patients and women than most other HF clinical trials. Furthermore, the patients in HF-ACTION demonstrated higher use of evidence-based therapies than those in other studies evaluating the prognostic role of GDF-15 among patients with HF. Attempting to assess the association of GDF-15 and individual outcomes relative to an optimized set of variables might have contributed to the difficulties in showing a significant association with outcomes besides all-cause mortality. Patients enrolled in the HF-ACTION trial had impaired ejection fraction (LVEF: <0.35%); as such our results cannot be extrapolated to other HF populations such as patients with HF and preserved ejection fraction. Our study revealed that patients with higher GDF-15 had worse peak Vo2; however, all patients had to consent and be able to perform a CPX test, which may not reflect the overall HF population.
In this analysis of a large clinical study with well-treated chronic HFrEF patients, increasing baseline GDF-15 was associated with long-term adverse cardiovascular outcomes. Even after comprehensive multivariate adjustment, including demographics, clinical variables, and biomarkers (hs-TnT and NT-proBNP), doubling the concentration of GDF-15 was significantly associated with all-cause mortality. Further adjustment for CPX-derived variables attenuated the value of GDF-15, but CPX testing is not widely used in routine clinical practice. These data suggest a potential role for GDF-15 in risk stratification among patients with chronic HFrEF.
COMPETENCY IN MEDICAL KNOWLEDGE: GDF-15, a member of the transforming growth factor-β cytokine family, is an emerging biomarker used for risk stratification among patients with cardiovascular diseases. In our analysis of a demographically diverse, well-managed patients with chronic HFrEF, GDF-15 provides independent prognostic information in addition to established predictors of outcomes including hs-TnT and NT-proBNP.
TRANSLATIONAL OUTLOOK: Our findings suggest a possible role for GDF-15 in the risk stratification of patients with chronic HFrEF. Future studies evaluating the mechanistic role of GDF-15 in patients with HF may lead to new therapeutic targets.
For supplemental tables, please see the online version of this article.
Dr. Sharma has received funding from the Alberta Innovates Health Solution Clinician Scientist fellowship, Roche Diagnostics, and a Bayer-Vascular Canadian Cardiovascular Society research grant. Dr. Fiuzat has received research funding from Roche. Dr. Adams has received research funding from and is a consultant for Roche. Dr. Zannad has received personal fees from Roche Diagnostics, Janssen, Bayer, Pfizer, Novartis, Boston Scientific, Resmed, Amgen, CVRx, Quantum Genomics, Takeda, General Electric, Boehringer Ingelheim, Relypsa, ZS Pharma, AstraZeneca, GlaxoSmithKline, Cardiorenal, and Cardiovascular clinical trials. Dr. O’Connor has received research funding from Roche. Dr. Felker has received research funding from Roche. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. John R. Teerlink, MD, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- B-type natriuretic peptide
- growth differentiation factor
- heart failure
- N-terminal pro–B-type natriuretic peptide
- Received March 7, 2017.
- Revision received July 3, 2017.
- Accepted July 27, 2017.
- 2017 American College of Cardiology Foundation
- Writing Group Members,
- Mozaffarian D.,
- Benjamin E.J.,
- et al.
- Yancy C.W.,
- Jessup M.,
- Bozkurt B.,
- et al.
- Kempf T.,
- Eden M.,
- Strelau J.,
- et al.
- Xu J.,
- Kimball T.R.,
- Lorenz J.N.,
- et al.
- Wallentin L.,
- Hijazi Z.,
- Andersson U.,
- et al.
- Wollert K.C.,
- Kempf T.,
- Lagerqvist B.,
- Lindahl B.
- Daniels L.B.,
- Clopton P.,
- Laughlin G.A.,
- Maisel A.S.,
- Barrett-Connor E.
- Eggers K.M.,
- Kempf T.,
- Lagerqvist B.,
- et al.
- Dallmeier D.,
- Brenner H.,
- Mons U.,
- Rottbauer W.,
- Koenig W.,
- Rothenbacher D.
- Anand I.S.,
- Kempf T.,
- Rector T.S.,
- et al.
- Wollert K.C.,
- Kempf T.,
- Giannitsis E.,
- et al.
- O'Connor C.M.,
- Whellan D.J.,
- Wojdyla D.,
- et al.
- Kim H.-N.,
- Januzzi J.L.
- Kempf T.,
- Haehling von S.,
- Peter T.,
- et al.
- Felker G.M.,
- Fiuzat M.,
- Shaw L.K.,
- et al.
- Felker G.M.,
- Fiuzat M.,
- Thompson V.,
- et al.
- Balady G.J.,
- Arena R.,
- Sietsema K.,
- et al.
- Chan M.M.Y.,
- Santhanakrishnan R.,
- Chong J.P.C.,
- et al.
- Fonarow G.C.,
- Albert N.M.,
- Curtis A.B.,
- et al.