Author + information
- Received June 28, 1996
- Revision received October 15, 1996
- Accepted November 12, 1996
- Published online March 1, 1997.
- Bruno Pinamonti, MDA,*,
- Massimo Zecchin, MDA,
- Andrea Di Lenarda, MDA,
- Dario Gregori, MA, PhDA,
- Gianfranco Sinagra, MDA and
- Fulvio Camerini, MDA
- ↵*Dr. Bruno Pinamonti, Department of Cardiology, Ospedale Maggiore, Piazza Ospedale 1, 34129 Trieste, Italy.
Objectives. We sought to assess the prognostic implications of the evolution of restrictive left ventricular filling pattern (RFP) in dilated cardiomyopathy (DCM).
Background. Previous work has demonstrated that a RFP in DCM is associated with a poor prognosis. Few data are available on the prognostic implications of the evolution of this pattern.
Methods. The evolution of left ventricular filling was studied by Doppler echocardiography in 110 patients with DCM. According to the left ventricular filling pattern at presentation and after 3 months of treatment, the patients were classified into three groups: Group 1A (n = 24) had persistent restrictive filling; Group 1B (n = 29) had reversible restrictive filling; and Group 2 (n = 57) had nonrestrictive filling.
Results. During follow-up (41 ± 20 months), mortality plus heart transplantation was significantly higher in Group 1A than in Groups 1B and 2 (p < 0.0001). On multivariate analysis, the model incorporating E wave deceleration time at 3 months was more powerful at predicting mortality with respect to this variable at baseline (p = 0.0039). Clinical improvement at 1 and 2 years was significantly more frequent in Groups 1B and 2 than in Group 1A (p < 0.0001 at 2 years).
Conclusions. In patients with DCM, the persistence of restrictive filling at 3 months is associated with a high mortality and transplantation rate. The patients with reversible restrictive filling have a high probability of improvement and excellent survival. Doppler echocardiographic reevaluation of these patients after 3 months of therapy gives additional prognostic information with respect to the initial study.
(J Am Coll Cardiol 1997;29:604–12)
A wide spectrum of left ventricular (LV) filling patterns was found during Doppler echocardiography in dilated cardiomyopathy (DCM) ([1–9]). Previous works ([3–12]) demonstrated that in DCM a “restrictive” (LV) filling pattern (RFP), characterized on Doppler echocardiography by a shortened transmitral E wave deceleration time (EDT), is associated with a more severe clinical and hemodynamic status and with increased mortality and heart transplantation rates. However, few data are presently available on the RFP changes during the course of the disease ([7, 9, 13, 14]) and on their clinical and prognostic implications. Therefore, the objectives of this study were to assess the evolution of the LV filling pattern in DCM and to evaluate its clinical and prognostic implications in comparison with the initial evaluation.
1.1 Study patients.
A consecutive series of 110 patients (89 men, mean age 40 ± 14 years) with DCM were evaluated by Doppler echocardiography at presentation and during follow-up. The diagnosis of DCM was made according to usual criteria (), in the presence of a depressed LV ejection fraction (EF) (<50% by both echocardiography and angiography) and in the absence of significant coronary artery disease and other specific heart muscle diseases. In addition, an adequate transmitral Doppler curve was required.
1.2 Clinical and hemodynamic evaluation.
The study patients were evaluated at clinical presentation, after 3 ± 2 months of tailored medical treatment and at long-term follow-up (1 year ± 3 months and 2 years ± 6 months after presentation).
Clinical data included New York Heart Association functional class, third heart sound and cardiothoracic ratio on chest X-ray film. Ventricular arrhythmias were evaluated by Holter electrocardiographic monitoring. Exercise tolerance was evaluated using a symptom-limited bicycle ergometer test in the upright position, with increments of 10 W/min.
Hemodynamic study, including coronary angiography and endomyocardial biopsy, was performed in all patients at diagnosis. Significant coronary stenosis (>50%) and myocarditis were excluded.
At the baseline evaluation, 76 patients (69%) were treated with angiotensin-converting enzyme (ACE) inhibitors (captopril or enalapril), 103 (94%) with digoxin, 76 (69%) with diuretic agents, 20 (18%) with amiodarone and 6 (5%) with beta-adrenergic blocking agents (metoprolol).
1.3 Echocardiographic study.
Conventional M-mode, two-dimensional and Doppler variables () were measured in all patients at presentation and at 3 ± 2, 12 ± 3 and 24 ± 6 months of follow-up. Left ventricular volumes and EF were calculated from the apical four-chamber view using the ellipsoid, single-plane formula. Right ventricular areas and the shortening fraction of areas, as well as the end-systolic left atrial area, were also measured with the same approach.
Mitral regurgitation (MR) was semiquantitatively graded from 1+ to 4+ considering the regurgitant jet area at color Doppler imaging (). Mitral regurgitation graded >1+ was considered significant.
The transmitral flow velocity curve was obtained by pulsed Doppler imaging, positioning the sample volume between the tips of the mitral leaflets. E wave and A wave peak velocities, EDT and the ratio of early transmitral flow velocity to atrial flow velocity (E/A ratio) were measured, as described previously ([1, 5]).
According to our previous report (), RFP was diagnosed when the EDT was <115 ms. All measurements were obtained from the mean of 3 beats for the patients with sinus rhythm, and 5 beats for those with atrial fibrillation. Chamber diameters, areas and volumes were normalized for body surface area.
Reproducibility of Doppler echocardiographic data (EDT and EF) were assessed in 25 randomly selected study patients. These measurements were repeated (mean of 3 beats) by the same examiner (B.P.) and by another author (M.Z.), who were unaware of the initial measurements. Intraobserver and interobserver variabilities (mean ± SD) were 0 ± 22.6 ms and 8 ± 13.6 ms for EDT and 0.07 ± 3.3% and 0.02 ± 4.0% for EF, respectively. In addition, we tested for a intrarater and inter-rater variability using a linear random effect model. According to the model, no rater effect exists (p = 0.61), and the observations are consistent across measurements (p = 0.74).
1.4 Classification of patients.
According to the Doppler LV filling pattern at baseline and at short-term follow-up, the patients were classified into three groups: Group 1A had persistent RFP (EDT <115 ms both at presentation and at 3-month follow-up); Group 1B had reversible RFP (EDT <115 ms at presentation and >115 ms at 3-month follow-up); and Group 2 had non-RFP (EDT >115 ms both at presentation and at 3-month follow-up).
1.5 Analysis of follow-up data.
The follow-up data were obtained by regular follow-up visits at our institution. Additional information on survival was collected by telephone contact with patients or their relatives and physicians.
Primary end points for follow-up were cardiac death or heart transplantation, or at the end of the study. For the purpose of analysis, patients who died or underwent cardiac transplantation were considered as a single group.
A secondary objective was the assessment of clinical improvement at 1- and 2-year follow-up. Clinical improvement was arbitrarily considered to be present in case of normalization of LV function (EF >52%) in asymptomatic patients (functional class I) or of an EF increase >10%, in association with at least two of the following criteria: decrease of at least one functional class; >10% decrease of cardiothoracic ratio on chest x-ray film; >10% decrease of LV end-diastolic diameter on M-mode echocardiography; or >120 s increase in exercise time on effort test.
1.6 Statistical analysis.
See Appendix for more detailed information. Given the small number of cases with missing values in one or more covariates (maximum 8%), and to avoid a possible bias due to discarding patients, the data were imputed using a psychometric scaling method (). Main descriptive statistics were computed for each group. Differences between group means were compared using the unpaired Student ttest, and frequency of events using the chi-square test with the Yates correction, where applicable. Longitudinal analyses within the same group were performed using paired ttests and the McNemar test. Bonferroni correction was applied for multiple analyses. The survival curves were estimated using the Kaplan-Meier approach. Confidence bounds were constructed using Greenwood’s formula. To compare the patients of Group 1A with those of Group 1B, a paired ttest was performed on the ranks of the Kaplan-Meier estimated survival times. To compare the prognostic significance of RFP at baseline with the persistence of RFP at 3 months, diagnostic accuracy in predicting mortality and heart transplantation, sensitivity, specificity, positive and negative test values, confidence bounds and standard errors was constructed using a bootstrap approach (). To give a first indication of the effect of the covariates on survival, the unadjusted relative risks were estimated. To evaluate the impact of EDT after adjusting for the effect of other covariates, we used the model presented in the previous report by Pinamonti et al. (), in which a Cox proportional hazards model was employed (); the covariates were considered in both the original (continuous) and the categorized scale (for the cutoff points, refer to Table 3). In both models, the two covariates that were inserted in the previous model mainly for model stability (i.e., age and functional class) did not turn out to be significant in our data at a Wald and a score tests. Thus, we decided to drop them. Then, a forward stepwise analysis was performed to see if any other variable could produce a significant improvement in the fit. Finally, we decided to consider the categorized model for further analysis, mainly because of its easier clinical interpretation. The fact that the two fitted models were basically the same, and that the estimated effect of covariates did not change when categorizing the variables, was a strong suggestion that no improper overestimation of the true effect of the covariates occurred because of ad hoc cutting off. In comparing the baseline model with the 3-month model to assess their predictive power, the prediction error was estimated using the bootstrap approach (). The overall effect of EDT was assessed using a proportional hazards model with EDT as time-dependent covariate.
2.1 Study group characteristics at presentation (Table 1).
An RFP was observed at presentation in 53 (48%) of 110 patients. Similar to the results of the previous study (), their clinical data and hemodynamic status were more severe than in patients without an RFP (Table 1).
After initial assessment, treatment with ACE inhibitors or beta-blockers, or both, was started or tailored according to clinical status up to the maximal tolerated dosage. At 3 months, 94% of patients were taking ACE inhibitors and 87% beta-blockers (metoprolol 110 ± 58 mg/day).
2.2 Short-term follow-up.
The first follow-up echocardiographic data—3 ± 2 months after presentation (mean 111 ± 54 days)—were available in all patients.
An RFP persisted in 24 patients classified in Group 1A (Fig. 1, upper tracings), whereas it disappeared in the other 29 patients classified in Group 1B (Fig. 1, lower tracings). At the initial evaluation these two groups were similar with regards to the majority of clinical, echocardiographic and hemodynamic variables (Table 1). The only statistically significant differences were age (Group 1A: 32 ± 18 years; Group 1B: 43 ± 11 years; p = 0.03) and history of mild (<170/100 mm Hg) hypertension (4% in Group 1A vs. 31% in Group 1B; p = 0.02).
After 3 ± 2 months, a significant improvement in most clinical and Doppler echocardiographic variables was observed in the whole group of patients with DCM, particularly evident in Groups 1B and 2 (Fig. 2). Notably, Group 1B patients could tolerate a higher beta-blocker dosage (Table 1) (120 ± 68 vs. 93 ± 46 mg/day; p = 0.0009).
2.3 Long-term follow-up—analysis of improvement.
According to the aforementioned criteria, 39 patients (35%) were considered improved at 1-year and 48 (44%) at 2-year follow-up.
An RFP persisted for 1 and 2 years in the majority of Group 1A patients who survived until the controls (10 of 15 patients at 1 year and 7 of 12 patients at 2 years, respectively). In contrast, an RFP was not present in any of the patients of Groups 1B and 2.
Considering the three groups of patients, improvement was significantly more common in Group 1B (46% at 1 year and 67% at 2 years) and in Group 2 (40% at 1 year and 55% at 2 years) than in Group 1A (12% at 1 year and 0% at 2 years) (p = 0.006 at 1 year and p < 0.0001 at 2 years).
Fig. 2shows the evolution during long-term follow-up of functional class, LV end-diastolic diameter index, EF and EDT in the three groups of patients. Significant differences in these variables versus those at baseline were present at 1 year and 2 years in Groups 1B and 2, but not in Group 1A.
2.4 Long-term follow-up—analysis of survival.
In April 1995, the mean follow-up duration was 40.7 ± 20.2 months. No patient was lost to follow-up; two died of noncardiac causes. Twenty-six patients (24%) died of a cardiac cause (13 patients: sudden death in 10 and heart failure in 3) or underwent cardiac transplantation (13 patients) 25.5 ± 19.6 months after presentation.
Fig. 3shows the Kaplan-Meier survival curves (free from death and heart transplantation) of the three groups of patients. After 1, 2 and 4 years, survival of patients with persistent RFP (Group 1A: 65%, 46% and 13%, respectively) was significantly lower (p < 0.0001) than that of the other two groups (Group 1B: 100% at 1 and 2 years and 96% at 4 years; Group 2: 100% at 1 and 2 years and 97% at 4 years). All but 1 of the 16 patients who died (3 patients) or underwent heart transplantation (12 patients) for refractory heart failure belonged to Group 1A, as well as 6 of the 10 patients who had a sudden death.
Table 2compares sensitivity, specificity, predictive positive and negative values and accuracy in predicting cardiac mortality and heart transplantation between the RFP present at diagnosis and the persistence of RFP after 3 months of tailored medical therapy. The presence of RFP at first evaluation showed a good sensitivity, whereas the persistence of RFP was significantly more specific and accurate.
On univariate Cox analysis, several variables, both at baseline and at 3 months, significantly correlated with survival free from heart transplantation (Table 3).
The EDT taken as a time-dependent covariate showed a very highly significant value (odds ratio 0.12, p < 0.0001, Table 4). Table 4shows the two Cox proportional hazards models (multivariate analysis) that considered the EDT at baseline and at 3 months, respectively. As mentioned in statistical analysis section earlier, this model was derived from the previous model (), which was simplified by dropping the two variables inserted for model stability (i.e., age and functional class). In addition, on forward stepwise analysis, the only other variable that showed a significant improvement in the model (p = 0.043) on the Wald test was the fractional shortening of right ventricular areas. However, this variable was highly correlated to EDT () and nonsignificant on a score test. The respective predictive powers of the model incorporating EDT at baseline (M1) and at 3 months (M2), estimated using the bootstrap approach ([17, 20]), showed a significant improvement of M2 with respect to M1 (95th percentile of theta = −2.655, p = 0.0039) (Table 4).
3.1 Evolution of left ventricular filling in dilated cardiomyopathy.
Previous work has demonstrated the importance of the assessment of diastolic function and LV filling in patients with DCM. Although this disease is mainly characterized by systolic dysfunction, abnormalities in diastolic function were also reported. In fact, earlier hemodynamic studies demonstrated both an impairment of relaxation and an increase in chamber stiffness (). Furthermore, from a clinical standpoint, in patients with heart failure, dyspnea and functional impairment mainly depend on an increase of LV filling pressure (), whereas exercise intolerance is more related to diastolic than systolic dysfunction (). These data were confirmed more recently by several studies using Doppler echocardiography. Appleton et al. () defined the RFP of the transmitral Doppler curve, characterized by a shortened EDT and frequently by a predominant E wave. The RFP can be considered a useful noninvasive sign of increased LV stiffness, as recently confirmed by experimental data ([23, 24]). The present study confirms previous investigations that showed that the RFP is common in DCM ([3–5, 7, 9–11]), particularly in patients with more severe clinical and hemodynamic abnormalities ([3–5, 7, 9–12]), and must be considered a powerful independent adverse prognostic factor ([5, 9–11]). Similar data were obtained by other investigators in patients with “dilated myocardial disease” () and with heart failure of various causes ().
Some workers ([7, 9, 11, 13, 14]) found significant changes in LV filling patterns during the course of different heart diseases. Klein et al. () observed in serial echocardiographic studies of patients with cardiac amyloidosis the frequent appearance of a RFP that was correlated with an adverse prognosis. In experimental works a progressive decrease of the EDT with the appearance of a RFP was observed during the development of heart failure ([23, 25]).
A significant relation between LV filling patterns and the clinical course of DCM was noticed by Werner et al. (): A filling pattern with a predominant E wave (presumably restrictive) was more frequent in patients symptomatic for heart failure than in stable patients with a longer duration of the disease. Moreover, during follow-up, an increase in the A wave velocity and a decrease in the E wave and E/A ratio were noted in patients who showed clinical improvement, whereas opposite changes were present in those patients who deteriorated. Similar data were described by Shen et al. (), associated with a prolongation of the EDT.
Our data confirm these preliminary observations and give additional information on the evolution of LV filling in patients with DCM. After 3 ± 2 months of treatment with vasodilators, digitalis and diuretics, frequently associated with beta-blockers, the RFP disappeared in 53% of patients. This evolution was associated with clinical improvement of heart failure and frequently with an increase of the EF (Fig. 2).
3.2 Prognostic significance of evolution of filling.
The present study demonstrates that the evolution of LV filling pattern at short-term follow-up is very helpful from a prognostic point of view. In fact, the patients with RFP at diagnosis—with a bad prognosis according to our previous results ()—can be further classified into two subgroups with different outcomes: Group 1A, in which RFP persisted, with a very high mortality and transplantation rate (35% at 1 year, 54% at 2 years and 87% at 4 years); and Group 1B, with reversible RFP, with an excellent long-term survival (96% at 4 years), comparable to that of patients without RFP at diagnosis (Group 2: 97% at 4 years) (Fig. 3). The presence of RFP after 3 months of treatment was more specific (94%) and accurate (90%) in predicting mortality and heart transplantation in comparison to the presence of RFP at first evaluation (62% and 67%, respectively), whereas both study points had a good sensitivity (77% and 81%, respectively) (Table 2). Moreover, on multivariate analysis, the model incorporating the EDT at 3 months showed a significantly higher predictive value with respect to this variable at baseline (Table 4).
3.3 Proposed mechanisms.
The reasons for the different evolutions of LV filling among patients with DCM are complex, and presumably caused by different mechanisms, such as the heterogeneity of the structural and functional characteristics of the disease and probably the effects of treatment. In our study no data at initial evaluation could predict the subsequent evolution of LV filling. The only two differences that were statistically significant were age (younger in Group 1A) and a history of mild hypertension (more frequent in Group 1B). Several studies demonstrated that LV filling is load dependent. In fact, a decrease in preload decreases the E velocity and E/A ratio and probably prolongs the EDT, whereas an increase in afterload prolongs relaxation ([26–31]). Severity of MR also may influence LV filling (), probably acting on filling pressure.
Thus, the Doppler mitral curve changes in some of our patients may be due at least in part to drug treatment of heart failure, which influenced preload (), afterload, MR () and presumably also ventricular interaction and pericardial restraint (). Furthermore, a favorable action on the diastolic transmitral Doppler curves may be due to treatment with beta-blockers ([33, 34]). Very recently, Andersson et al. () demonstrated that in patients treated with metoprolol, the maximal effects on diastolic variables were reached during the first 3 months of treatment, whereas improvement in the systolic function was observed later.
Although the present study was not specifically designed to assess the influence of beta-blockers on diastolic function and prognosis, and treatment with these drugs was not randomly assigned, it must be noted that the majority of our patients were treated with beta-blockers (Table 1), and that patients with reversible RFP (Group 1B) could tolerate higher doses of beta-blockers than those with persistent RFP (Group 1A). These observations appear to be in agreement with those of Bristow et al. (), who demonstrated that in patients with heart failure the functional improvement due to beta-blocker treatment appears to be dose related. Nevertheless, because our patients were not randomized but titrated to the maximal tolerated dose, we cannot exclude that those receiving a higher dose of beta-blockers were the same who had the best prognosis.
Unfortunately, in Group 1A patients the RFP persisted despite tailored drug treatment. Presumably in these patients the disease was more severe or progressive. As suggested by Bortone et al. (), more severe structural myocardial abnormalities, such as fibrosis, were hypothesized. Further studies using morphometric analysis of endomyocardial biopsy are necessary to answer this question ().
3.4 Study limitations.
This study has several potential limitations. First, the number of patients with persistent RFP was quite small. Because hemodynamic study was not repeated systematically after 3 months, we cannot know whether persistent RFP was related to a persistent increase in LV filling pressure, and which one was the best prognostic factor.
In DCM the evolution of the left ventricle during the course of the disease is presently unknown. We arbitrarily considered the changes in the LV filling pattern at a mean interval of 3 months after the first evaluation. Further studies may be helpful to identify the optimal interval for the evaluation of the response to treatment on LV filling. Moreover, because DCM is probably a heterogeneous disease, different evolutions could be expected among different patients. In our series the persistence of RFP was a valuable predictor of refractory heart failure, leading frequently to cardiac transplantation or death (15 of 16 patients). However, sudden death is also a major problem in patients with DCM, and it was confirmed in our patients. Although the risk of sudden death appears to be related to the severity of ventricular arrhythmias () and systolic dysfunction (), it is uncertain whether the presence of RFP could also have a prognostic role. Finally, diastole is a complex phenomenon that is difficult to analyze from any single variable. Recent data showed that the evaluation of pulmonary venous Doppler curves gives additional information on LV diastolic pressures (). Systematic analysis of these curves was not performed in this study.
3.5 Clinical implications.
Our study demonstrates that in patients with DCM the Doppler echocardiographic assessment of the evolution of LV filling during tailored medical therapy is clinically helpful in predicting subsequent prognosis and it might be used in conjunction with the clinical assessment to select patients for a more strict follow-up or for cardiac transplantation.
From the present study the structural substrate of persistent RFP remains undefined and further studies are necessary to elucidate this topic.
A.1 Statistical Analysis
Group comparison.To derive test statistics for the hypothesis that the sensitivity and specificity and other statistics are different if based on the baseline EDT or the 3-month EDT, we used a bootstrap approach ().
Let (x1i, y1i, x2i, y2i) be a vector of the observations for the ith subject, where x1refers to the baseline EDT, y1to the baseline outcome (dead/alive) and the same for y2, but for the EDT at 3 months. We are interested in the statistics: θ = θBase− θ3m, to test the null hypothesis that θ = 0.
If n is the number of observations, then we can decide to extract the B bootstrap sample from )}for j1, … , jn a random sample of integers from 1 to n.
Then, the bootstrap estimate is where with b = 1, … , B, and Then a test and a confidence bound can be derived using the normal approximation: where SE represents standard error, and z represents normal (1 − α).
The asymptotic normality of the bootstrap estimate has been checked using quantile–quantile (QQ) plots.
Predictive value of EDT at baseline and 3 months.To address whether the EDT at 3 months has a higher discriminant predictive power than the EDT at baseline, after adjusting for the relevant confounding effects, we used an approach introduced by Efron and Tibshirani (), which allows comparisons in terms of accuracy of predictions between two different models.
Let us call the two models we are interested in M1and M2, where M1is the baseline model and M2is the model for the 3-month EDT, both adjusted for the baseline pulmonary wedge pressure and mean aortic pressure (see Results and Table 4). Using the same idea as outlined earlier, a formal comparison can be based on the statistics for the generic jth model (j = 1, 2): (where RSE represents residual square error), which indicates good prediction for small values and a perfect fit in correspondence with the 0. Thus, M1and M2were compared with respect to their predictive power, by comparing RSE(M1) and RSE(M2).
A statistic like θ̂ can then be introduced to compare the two previous statistics: A positive value of θ̂ would indicate that the M1model is not as good as the M2model for predicting long-term survival.
We used the bootstrap BC or ABC methods to derive the confidence bounds of θ. For details refer to Efron and Tibshirani (). The models M1and M2are proportional hazards regressions, and RSE is the deviance residual for them.
This study was presented in part at the 68th Scientific Sessions of the American Heart Association, Anaheim, California, November 1995.
- angiotensin-converting enzyme
- dilated cardiomyopathy
- ratio of early transmitral flow velocity to atrial flow velocity
- E wave deceleration time
- ejection fraction
- left ventricular
- mitral regurgitation
- restrictive filling pattern
- Received June 28, 1996.
- Revision received October 15, 1996.
- Accepted November 12, 1996.
- The American College of Cardiology
- Vanoverschelde J-LJ,
- Raphael DA,
- Robert AR,
- Pinamonti B,
- DiLenarda A,
- Sinagra G,
- Camerini F
- Xie G-Y,
- Berk MR,
- Smith MD,
- Gurley JC,
- DeMaria AN
- Rihal CS,
- Nishimura RA,
- Hatle LK,
- Bailey KR,
- Tajik AJ
- Klein AL,
- Hatle LK,
- Taliercio CP,
- et al.
- Richardson P,
- McKenna W,
- Bristow M,
- et al.
- Sarle M
- Cox DR
- Efron B,
- Tibshirani RJ
- Grossman W,
- McLaurin LP,
- Rolett EL
- Braunwald E
- Packer M
- Ohno M,
- Cheng C-P,
- Little WC
- Little WC,
- Ohno M,
- Kitzman DW,
- Thomas JD,
- Cheng C-P
- Kono T,
- Sabbah HN,
- Rosman H,
- Alam M,
- Stein PD,
- Goldstein S
- Choong CY,
- Abascal VM,
- Thomas JD,
- Guerrero JL,
- McGlew S,
- Weyman AE
- Stoddard MF,
- Pearson AC,
- Kern MJ,
- Ratcliff J,
- Mrosek DG,
- Labovitz AJ
- Stevenson LW,
- Bellil D,
- Grover-McKay M,
- et al.
- Masuyama T,
- Goar FGS,
- Alderman EL,
- Popp R
- Janicki JS
- Kim MH,
- Starling MR
- Andersson B,
- Caidahl K,
- diLenarda A,
- et al.
- Bristow MR,
- O’Connel JB,
- Gilbert EM,
- et al.
- Bortone AS,
- Hess PM,
- Chiddo A,
- et al.
- for the V-HeFT VA Cooperative Studies Group,
- Goldman S,
- Johnson G,
- Cohn JN,
- Cintron G,
- Smith R,
- Francis G
- Rossvoll O,
- Hatle LK