Author + information
- Received July 13, 2017
- Revision received September 18, 2017
- Accepted October 3, 2017
- Published online May 6, 2019.
- Wojciech Kosmala, MD, PhDa,b,c,
- Monika Przewlocka-Kosmala, MD, PhDa,b,c,
- Aleksandra Rojek, MD, PhDa and
- Thomas H. Marwick, MBBS, PhD, MPHc,∗ ()
- aCardiology Department, Wroclaw Medical University, Wroclaw, Poland
- bCardiovascular Imaging Research Group, Menzies Institute for Medical Research, University of Tasmania, Hobart, Australia
- cCardiovascular Imaging Research Group, Baker Heart and Diabetes Institute, Melbourne, Australia
- ↵∗Address for correspondence:
Dr. Thomas H. Marwick, Baker Heart and Diabetes Institute, P.O. Box 6492, Melbourne, Victoria 3004, Australia.
Objectives This study sought to establish the diagnostic and prognostic value of a strategy for prediction of abnormal diastolic response to exercise (AbnDR) using clinical, biochemical, and resting echocardiographic markers in dyspneic patients with mild diastolic dysfunction.
Background An AbnDR (increase in left ventricular filling pressure) may indicate heart failure with preserved ejection fraction as the cause of symptoms in dyspneic patients, despite a nonelevated noncardiac at rest. However, exercise testing may be inconclusive in patients with noncardiac limitations to physical activity.
Methods In 171 dyspneic patients (64 ± 8 years) with suspected heart failure with preserved ejection fraction but resting peak early diastolic mitral inflow velocity/peak early diastolic mitral annular velocity ratio (E/e′) <14, a complete echocardiogram (including assessment of myocardial deformation and rotational mechanics) and blood assays for biomarkers were performed. Echocardiography following maximal exercise was undertaken to assess AbnDR (exertional E/e′ >14). Patients were followed over 26.2 ± 4.6 months for endpoints of cardiovascular hospitalization and death.
Results AbnDR was present in 103 subjects (60%). Independent correlates of AbnDR were resting E/e′ (odds ratio [OR]: 8.23; 95% confidence interval [CI]: 3.54 to 9.16; p < 0.001), left ventricular untwisting rate (OR: 0.60; 95% CI: 0.42 to 0.86; p = 0.006), and galectin-3—a marker of fibrosis (OR: 1.80; 95% CI: 1.21 to 2.67; p = 0.004). The use of resting E/e′ >11.3 and galectin-3 <1.17 ng/ml to select patients for further diagnostic processing would have allowed exercise testing to be avoided in 65% of subjects, at the cost of misclassification of 13%. The composite outcome of cardiovascular hospitalization or death occurred in 47 patients (27.5%). The predictive value of an AbnDR response and the combined strategy (resting echocardiography and galectin-3 or exercise testing in case of an inconclusive first step) showed similar event prediction (36 vs. 34; p = 0.95).
Conclusions The implementation of a 2-step algorithm (echocardiographic evaluation of resting E/e′ followed by the assessment of galectin-3) may improve the diagnosis and prognostic assessment of individuals with suspected heart failure with preserved ejection fraction who are unable to perform a diagnostic exercise test.
The use of exercise testing is appropriate to confirm the presence of heart failure with preserved ejection fraction (HFpEF) in some dyspneic patients (1). The recognition of mild left ventricular (LV) diastolic dysfunction at rest may not coincide with a diminished functional reserve during exercise, and, if not supported by other tests, may falsely lead to overdiagnosis of HFpEF. This is particularly the case for patients with grade 1 diastolic dysfunction and no evidence of elevated left ventricular filling pressure (LVFP) at rest, who require further clinical evaluation including diastolic stress testing (1). The identification of abnormal diastolic response to exercise, defined as the average ratio of peak early diastolic mitral inflow velocity (E) to the average peak early diastolic mitral annular velocity (e′) (E/e′) > 14 at peak exertion, with an accompanying restriction of aerobic capacity, is strongly suggestive of the most common HFpEF phenotype, characterized by increased exercise LVFP (2). Nonetheless, exercise testing may be inconclusive in patients unable to perform adequate exercise effort due to extracardiac limitations. In view of this, an augmentation of resting assessment, identifying the subset of patients among whom diastolic impairment might worsen with exercise, might be valuable. Such an approach could address the pathophysiologic heterogeneity of HFpEF and allow better understanding of the relationship between demonstrated organic pathologies and exercise intolerance (3–6).
Accordingly, we sought to investigate a range of clinical, biochemical, and resting echocardiographic markers to predict an abnormal LV diastolic response to exercise (based on exertional E/e′) in the HFpEF subset with a mild diastolic dysfunction and estimated nonelevated LVFP at rest. To validate this new approach, we sought its ability to predict adverse outcomes in follow-up.
We enrolled 171 symptomatic patients with suspected HFpEF, satisfying the following diagnostic criteria: 1) signs or symptoms of heart failure (dyspnea, fatigue, and exercise intolerance) categorized as New York Heart Association functional class II or III, exercise capacity reduced from age- and sex-predicted normal ranges; 2) preserved LV ejection fraction (≥50%); 3) grade 1 diastolic dysfunction; and 4) nonelevated (≤14) average E/e′ at rest.
The exclusion criteria comprised atrial fibrillation or flutter, ischemic heart disease (defined by the presence of atherosclerotic lesions on coronary angiography or inducible ischemia during exercise testing), moderate and severe valvular heart disease, diagnosed or suspected pulmonary diseases (vital capacity <80% or forced expiratory volume in 1 s <80% of age- and sex-specific reference values), hemoglobin ≤11 g/dl, and other significant comorbidities, including malignancy, renal failure with estimated glomerular filtration rate <30 ml/min/1.73 m2, infections, and autoimmune, skeletal, and thyroid illnesses.
The study protocol comprised cardiopulmonary exercise testing, resting and immediate post-exercise echocardiogram, and the blood biomarkers serum galectin-3 and plasma B-type natriuretic peptide (BNP).
All enrollees were informed of the purpose of the study and provided written informed consent. Investigations were in accordance with the Declaration of Helsinki and were approved by the institutional review board.
A standard clinical history was taken, including information about risk factors and etiology. The Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC) risk score was calculated for each participant to assess prognosis on the basis of clinical parameters (7).
Standard equipment (Vivid e9, GE Medical Systems, Milwaukee, Wisconsin) with a phased array 2.5-MHz multifrequency transducer was used for echocardiographic imaging. Ultrasound acquisitions were saved in digital format on a secure server and analyzed offline. LV dimensions and wall thicknesses were measured according to standard recommendations. LV volumes were assessed by the biplane Simpson method and left atrial volume by area-length method (8). All cardiac volumes were indexed for body surface area. Cardiac output was calculated as the product of heart rate and stroke volume.
LV inflow parameters including peak early and late diastolic flow velocity (A), and deceleration time of early diastolic flow were assessed by pulsed wave Doppler from the apical 4-chamber view with the sample volume placed between the tips of mitral leaflets. Peak early diastolic tissue velocity at the septal and lateral portions of the mitral annulus were measured by pulsed-wave tissue Doppler. E/e′ was computed to approximate LV filling pressure, and this was considered to be elevated at E/e′ >14 (1). An abnormal diastolic response to exercise (AbnDR) was defined as exertional average E/e′ >14.
LV diastolic dysfunction was graded in accordance with the current recommendations (1).
LV myocardial deformation was assessed using a semiautomated 2-dimensional speckle-tracking technique (Echopac PC version 113, GE Healthcare, Horten, Norway) in the 3 apical views (longitudinal strain and strain rate) and 2 parasternal short-axis views at the basal and apical levels (rotational deformation) at a temporal resolution of 60 to 90 frames/s. After initial manual tracing of the endocardial border, the software identified 6 segments in each view and tracked the movement of acoustic markers. Manual readjustment was used in segments that tracked inadequately and if this was ineffective, they were excluded from further analysis. The greatest negative value on the strain curve and on the strain rate curve during LV systole were measured and averages from all segments subjected to analysis were presented as global longitudinal strain and strain rate.
The basal level was identified by the presence of mitral leaflets while excluding the mitral annulus, and the apical level was identified by the presence of LV cavity in the absence of papillary muscles. LV rotation/rotation velocity at each level was determined as the average angular displacement/displacement velocity of the 6 myocardial segments. LV twist was obtained by subtracting basal rotation from apical rotation at corresponding time points in the cardiac cycle, and the greatest positive value on the curve was assessed. LV torsion was derived as the peak LV twist divided by LV end-diastolic longitudinal length measured in the apical 4-chamber view. LV untwisting rate (UT) was defined as the peak untwisting velocity occurring in early diastole, analogous to calculation of LV twist as a net difference between basal and apical short-axis planes. Counterclockwise rotation was marked as a positive value and clockwise rotation as a negative value when viewed from the apex. The timings of cardiac events (aortic valve closure, mitral valve opening) were estimated from spectral Doppler recordings.
The apical 4- and 2-chamber views were used to evaluate left atrial longitudinal strain and the onset of the P-wave was accepted as a zero-reference point to determine deformation at atrial contraction (the first negative component) and total left atrial deformation (the sum of peak negative and peak positive components).
All echocardiographic parameters were averaged over 3 consecutive cardiac cycles.
Cardiopulmonary exercise testing
Each subject underwent symptom-limited cardiopulmonary exercise testing using a modified Bruce treadmill protocol. Ventilation, oxygen uptake, and carbon dioxide production were monitored continuously, and peak oxygen uptake was calculated as the average oxygen consumption during the last 30 s of exercise. To obtain a parameter independent of the potential errors in direct gas analysis, exercise capacity was also assessed in metabolic equivalents on the basis of the peak exercise intensity from treadmill speed and grade.
Echocardiographic evaluation of LV volumes, longitudinal and rotational deformation, and diastolic function (including E/e′ ratio) was undertaken at rest. E/e′ was also measured immediately after termination of exercise to assess LV diastolic response to exertion. In case of overlapping early and late diastolic Doppler waves (E and A and/or e′ and a′) at high heart rates, images were acquired after heart-rate deceleration led to separation of the respective waves.
Peripheral blood samples were drawn between 8:00 am and 9:00 am, after a 30 min rest in the supine position, and then frozen at –70°C until assayed. Serum galectin-3 levels were assessed with enzyme-linked immunoadsorbent assay kits from BioVendor, Inc. (Brno, Czech Republic). Intra- and interassay coefficients of variation were 6.3% and 8.7%, respectively. BNP levels were quantified using a commercially available fluorescence immunoassay (Triage BNP Test, Biosite Diagnostics Inc., San Diego, California).
Patients were followed over 26.2 ± 4.6 months for endpoints of cardiovascular hospitalization and death.
Data are presented as mean ± SD for normally distributed variables, as median and interquartile range for skewed variables (BNP and galectin-3), and as counts and percentages for categorical variables. BNP and galectin-3 were log-transformed before being analyzed. A change in E/e′ ratio with exercise was calculated by subtracting the pre-test value from the post-test value. Between-group comparisons were performed with an unpaired 2-sided Student’s t-test and the chi-square test for categorical variables. Homogeneity of variances was evaluated by the Levene test. Associations between variables were assessed by Pearson or Spearman correlation coefficients and stepwise multiple regression analysis. Univariate and multivariable logistic regression analysis models were developed to elucidate the associations of predictors with an abnormal diastolic response to exercise. The C-statistic was used to evaluate model performance. The ability of resting E/e′, UT, and galectin-3 to predict AbnDR was assessed using receiver-operating characteristic analysis, which also assisted the determination of the discriminatory cutoff points for these variables (obtained from all patients), defined as the maximum Youden index. Differences in the area under the receiver-operating characteristic curves were analyzed using the DeLong method. Because of an unavoidable positive bias linked with developing and evaluating a discriminant rule using the same dataset, cross-validation was used as an unbiased method to attest the reliability of the obtained cutoff values: one-half of the cases were randomly selected as the development set and one-half as the validation set. Event-free survival curves were estimated using the Kaplan-Meier method, and differences were assessed by the log-rank test. The net reclassification improvement, calculated on the basis of a categoryless approach, was used to assess the effect of reclassification obtained after addition of AbnDR or the constituents of the new diagnostic approach (E/e′ at rest and galectin-3) to the Cox proportional hazards models including MAGGIC risk score and BNP. The reproducibility of E/e′ ratio and UT was evaluated by the Bland-Altman method (mean difference and 95% confidence interval [CI]) and intraclass correlation coefficient. To assess the beat-to-beat variability of UT, the coefficient of variation was calculated, and the analogous coefficient was computed for longitudinal strain to be used as a frame of reference. All calculations were carried out using standard statistical software (Statistica for Windows 12, StatSoft Inc., Tulsa, Oklahoma). The level of statistical significance was set at a 2-sided p value <0.05.
The studied population demonstrated demographic and clinical features typical of HFpEF, especially a high frequency of female sex, hypertension, diabetes mellitus, overweight, and obesity. The subset with an abnormal diastolic response to exercise (higher E/e′ at exercise and change in E/e′ from rest to exercise) was characterized by lower exercise capacity and functional impairment by New York Heart Association, higher MAGGIC risk score and galectin-3 level, more impaired resting LV diastolic properties (lower e′ and UT, and higher E/e′) and left atrial function (lower atrial strain), than their counterparts with a normal exertional diastolic response (Tables 1 and 2⇓⇓).
Correlates of AbnDR
In unadjusted logistic regression models, galectin-3, E/e′, e′, UT, left atrial strain, and MAGGIC risk score were significantly associated with an abnormal diastolic response to exercise (Table 3). Univariable correlations among E/e′, UT, galectin-3, and AbnDR were presented in Supplemental Table 1.
Multivariable models for prediction of AbnDR were developed on the basis of anticipated associations—including the parameters mentioned in Table 3, as well as prescribed medications. In the best-performing model, resting E/e′, UT, and galectin-3 were significant independent predictors of AbnDR (Table 4). The independent contribution of these variables to the absolute exercise-induced increase in estimated LVFP was confirmed in a linear regression model (Supplemental Table 2).
Discriminatory significance of E/e′, UT, and Galectin-3
The ability of resting E/e′, UT, and galectin-3 to predict AbnDR was estimated using receiver-operating characteristic curves (Figure 1). The optimal cutpoint for E/e′ was 11.3 (sensitivity 53%, specificity 94%), for galectin-3 1.17 ng/ml (sensitivity 70%, specificity 62%), and for UT 59 rad/s (sensitivity 29%, specificity 87%).
Comparison of the areas under the curve demonstrated a significant difference between E/e′ and UT (p = 0.008). No significance for the other comparisons was found (E/e′ vs. galectin-3: p = 0.31, and galectin-3 vs. UT: p = 0.08). The cross-validation analysis revealed similar sensitivity and specificity in the development and validation subsets for E/e′ and galectin-3 indicating that the discriminant rule was created without significant bias. The inter-subset differences were more pronounced for UT, suggesting a lesser reliability of the determined cutpoint for this variable (Supplemental Table 3).
Strategy for prediction of AbnDR
Identification of predictors of AbnDR measured on evaluation at rest might allow a diagnostic strategy to be set up without carrying out stress testing. The initial selection step (Figure 2) was based on resting echocardiographic assessment of E/e′ ratio. The positivity of this parameter (>11.3) was consistent with the presence of AbnDR in 93% of subjects. The second step was needed in individuals with negative echocardiographic findings (66% of the overall population) and included further stratification by blood galectin-3 levels. The low concentration of this marker (below 1.17 ng/ml) was associated with a normal diastolic response to exertion in 79% of patients. The diagnostic process was inconclusive in the remainder of subjects (35% of the entire population), which may warrant exercise testing to establish diagnosis of AbnDR.
Associations with outcomes
The composite outcome of cardiovascular hospitalization or death occurred in 47 patients (27.5%). Subgroups with AbnDR identified either by the initial exercise testing approach or by the combined strategy (resting echocardiography and galectin-3 or exercise testing in case of the inconclusiveness of the first step) were characterized by a higher risk of cardiovascular hospitalization or death (Supplemental Figure 1). The predictive value of both approaches was similar (36 events identified by the former vs. 34 by the latter; p = 0.95) (Figure 3).
The addition of AbnDR as well as the components of the new approach (E/e′ at rest and galectin-3) to the model including MAGGIC risk score and BNP resulted in significant reclassification improvements in prediction of the studied outcome (net reclassification index 45%; p = 0.009, and 35%; p = 0.038, respectively).
The level of agreement in measurements of the echocardiographic predictors of AbnDR was assessed in 15 randomly selected examinations, analyzed twice by 2 observers (W.K. and M.P.K.) blinded to the patients’ clinical data on 2 separate days. The intraobserver and interobserver variability as expressed by intraclass correlation coefficient was for UT 0.96 and 0.92, for E/e′ at rest 0.94 and 0.89, and for E/e′ at exercise 0.94 and 0.96, respectively. Mean difference was for UT 6 rad/s (95% CI: −5 to 17) and 3 rad/s (95% CI: −11 to 18), for E/e′ at rest 0.7 (95% CI: 0.3 to 1.2) and −0.2 (95% CI: −0.6 to 0.2), and for E/e′ at exercise −0.4 (95% CI: −1.0 to 0.3) and −0.8 (95% CI: −1.3 to −0.3), for intraobserver and interobserver comparisons, respectively. The beat-to-beat variation of UT was 12.3%, exceeding that of longitudinal strain (5.2%).
This study revealed that in HFpEF patients with exertional dyspnea and nonelevated LVFP at rest, abnormal diastolic response to exercise as expressed by an exertional increase in E/e′ can be predicted by resting echocardiographic indices—E/e′ and UT, and circulating galectin-3—a marker of fibrosis. The implementation of a 2-step algorithm, with the echocardiographic evaluation of resting E/e′ followed by the laboratory assessment of galectin-3 in the subset with a negative imaging finding, may improve the diagnosis of HFpEF in two-thirds of individuals who are unable to perform adequate exercise effort, at the cost of misclassification of 13%.
E/e′ ratio has been recognized as a major echocardiographic parameter reflecting LVFP (9). Pathophysiologically, it is linked with the passive properties of the heart muscle as indicated by associations with LV stiffness and myocardial collagen content (10,11). E/e′ is central to the framework for the recognition of AbnDR. Although resting measurements of E/e′ were predictive of its rise during exertion, the modest correlation between resting E/e′ and AbnDR may indicate that the exertional increase in E/e′ is not predominantly determined by baseline E/e′. The fact that the optimal cutpoint of E/e′ was in the indeterminate range (i.e., 10 to 14) may reflect the extent of myocardial disease that is insufficient to induce apparent hemodynamic derangements at rest, but large enough to provide a substrate for the pathological response to stress.
LV untwisting is a first mechanical event in diastole. Previous studies have demonstrated a reduction and delay of untwisting in a number of clinical pathologies associated with diastolic dysfunction, including HFpEF, hypertrophic cardiomyopathy, aortic stenosis, and severe left ventricular hypertrophy (5,12–15). The physiologic basis of untwist is elastic recoil associated with the release of energy stored during systole, both within sarcomeres by the cardiomyocyte protein titin, and within the interstitium by the shear strain and torsion resulting from the contraction of the subendocardial and subepicardial helixes of myocardial fibers (16–18). Untwist is the precursor to isovolumic pressure decay, generating a base-to-apex intraventricular pressure gradient responsible for diastolic suction facilitating LV filling.
Invasive investigations have revealed that untwisting parameters are related to LV relaxation and suction indices but not to LV passive compliance (19). LV untwisting abnormalities occurring at rest are likely to further exacerbate with exercise. In the setting of the tachycardia-induced exertional shortening of diastole, the deficient contribution from untwisting to LV filling may lead to an overt hemodynamic compromise with exertional dyspnea. Altogether, LV untwisting represents a separate aspect of LV diastology from E/e′, and these results indicate that it may affect the LVFP response to exercise. However, the use of UT in clinical practice is problematic. We did not include UT in the proposed diagnostic algorithm because of the poor test-retest reproducibility, as well as the fact that it is not routinely measured in the echo lab.
Galectin-3 is a marker reflecting the progression of cardiac fibrosis—a pathological basis for increased myocardial stiffness and LV filling abnormalities. An extensive body of evidence indicates that this protein is involved in a variety of pathophysiologic processes in HF, including fibroblast proliferation, collagen synthesis, inflammation, and cardiac remodeling (20–22). Previous studies in HFpEF cohorts demonstrated prognostic significance and associations of elevated galectin-3 levels with impaired exercise capacity, severity of LV diastolic dysfunction, and myocardial collagen content (23–25).
Our analysis showed an independent contribution of galectin-3 to prediction of exertional LV filling disturbances. The usefulness of galectin-3 to identify patients without AbnDR after an initial echocardiographic pre-selection was demonstrated in the proposed algorithm. This finding is consistent with the link between a lesser extent of fibrosis (reflected by lower galectin-3) and greater LV functional reserve under an exercise load.
Association with cardiovascular risk
We provided a clinical validation of the new strategy by demonstrating the association of separated subsets “at risk” (compatible with AbnDR) with adverse outcomes. The proposed approach can offer additional prognostic information beyond traditional clinical and biochemical indicators and may contribute to improvement in risk stratification in HFpEF.
In our experience, at least 20% of dyspneic patients are unable to exercise, and Ingle et al. (26) reported that in addition to 15% of HF patients who are unable to exercise, over 40% cannot exercise maximally. In addition, as the galectin-3 test costs $10 per sample, which is <5% of the cost of an exercise echocardiogram, at some stage, the approach might be considered in patients who are able to exercise. However, while the value of this new diagnostic strategy is emphasized by its similar predictive value to AbnDR, there are a number of benefits of exercise testing—such as the detection of unexpected ischemia—that would be lost if the proposed strategy were to be adopted in patients who are able to exercise.
First, the exclusion of patients with atrial fibrillation and myocardial ischemia was necessary because of the confounding effect of the former on the accuracy of myocardial deformation and Doppler measurements, and the fact that the latter could per se limit exercise capacity—but these exclusions might constrain the external validity of our results. Second, the recruitment of patients limited to a single academic cardiology center might affect the generalizability of findings. Third, we did not use temporal normalization of LV rotational data; however, in the absence of variation in the duration of cardiac cycle between apical and basal acquisitions, it seems unlikely that this had much impact. On the other hand, the increased beat-to-beat variation of UT (as compared to longitudinal deformation—a parameter believed to be relatively stable in consecutive cardiac cycles) might have affected the adequacy of measurements. Finally, despite the lack of significant associations between medications used by patients and AbnDR, as well as the identified predictors, we cannot definitely rule out the possibility of influence of concomitant pharmacotherapy on the study results.
We demonstrated that AbnDR in HFpEF is associated with abnormal metrics of LV filling at rest, as well as with elevated galectin-3—a marker of myocardial fibrosis underlying diastolic defects. Disturbances of both active and passive components of LV diastolic physiology form the underpinnings for an exercise-induced increase in estimated LVFP. The assessment of resting E/e′ and blood assay of galectin-3 may be helpful in the diagnostic process of the separation of HFpEF subsets with and without exertional worsening of LV diastolic properties. Given that the pathophysiologic heterogeneity of HFpEF is likely to elicit differential responses to treatment, this may facilitate the recruitment process in future trials of drug testing in this disease condition.
COMPETENCY IN MEDICAL KNOWLEDGE: The recognition of HFpEF as the cause of dyspnea remains difficult. Exercise testing may clarify the situation but adds cost and complexity and is not suited to many elderly HFpEF patients. A combination of biomarkers and resting echocardiography could be useful.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: The measurement of serum galectin (a marker of fibrosis) with echocardiographic indices of myocardial relaxation and filling pressure provides analogous accuracy to exercise echocardiography for the recognition of HFpEF.
TRANSLATIONAL OUTLOOK: The therapeutic implications of this method of HFpEF diagnosis remain to be defined.
This research was supported by grant ST-678 from Wroclaw Medical University and grant 13-024 from the Royal Hobart Hospital Foundation. Dr. Marwick has received research grant support for the SUCCOUR trial (a trial of strain for detection of cardiotoxicity), unrelated to this topic, from GE Medical Systems. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Sherif Nagueh, MD, served as the Guest Editor for this paper.
- Abbreviations and Acronyms
- abnormal diastolic response to exercise
- B-type natriuretic peptide
- confidence interval
- peak early diastolic mitral inflow velocity
- peak early diastolic mitral annular velocity
- heart failure with preserved ejection fraction
- left ventricular
- left ventricular filling pressure
- Meta-Analysis Global Group in Chronic Heart Failure
- left ventricular untwisting rate
- Received July 13, 2017.
- Revision received September 18, 2017.
- Accepted October 3, 2017.
- 2019 American College of Cardiology Foundation
- Nagueh S.F.,
- Smiseth O.A.,
- Appleton C.P.,
- et al.
- Kosmala W.,
- Rojek A.,
- Przewlocka-Kosmala M.,
- Mysiak A.,
- Karolko B.,
- Marwick T.H.
- Lam C.S.,
- Roger V.L.,
- Rodeheffer R.J.,
- Borlaug B.A.,
- Enders F.T.,
- Redfield M.M.
- Shah A.M.,
- Claggett B.,
- Sweitzer N.K.,
- et al.
- Tan Y.T.,
- Wenzelburger F.,
- Lee E.,
- et al.
- Persson H.,
- Lonn E.,
- Edner M.,
- et al.,
- for the CHARMES Investigators
- Ommen S.R.,
- Nishimura R.A.,
- Appleton C.P.,
- et al.
- Kasner M.,
- Westermann D.,
- Steendijk P.,
- et al.
- Kasner M.,
- Westermann D.,
- Lopez B.,
- et al.
- Stuber M.,
- Scheidegger M.B.,
- Fischer S.E.,
- et al.
- Maier S.E.,
- Fischer S.E.,
- McKinnon G.C.,
- Hess O.M.,
- Krayenbuehl H.P.,
- Boesiger P.
- Notomi Y.,
- Martin-Miklovic M.G.,
- Oryszak S.J.,
- et al.
- Notomi Y.,
- Srinath G.,
- Shiota T.,
- et al.
- Sharma U.C.,
- Pokharel S.,
- van Brakel T.J.,
- et al.