Author + information
- Received September 8, 2006
- Revision received February 5, 2007
- Accepted February 5, 2007
- Published online June 5, 2007.
- Tomislav Mihaljevic, MD⁎,⁎ (, )
- Buu-Khanh Lam, MD⁎,
- Jeevanantham Rajeswaran, MSc†,
- Masami Takagaki, MD⁎,
- Michael S. Lauer, MD‡,
- A. Marc Gillinov, MD⁎,a,
- Eugene H. Blackstone, MD⁎,† and
- Bruce W. Lytle, MD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Tomislav Mihaljevic, Department of Thoracic and Cardiovascular Surgery, Cleveland Clinic, 9500 Euclid Avenue, Desk F24, Cleveland, Ohio 44195.
This paper was presented at the 55th Annual Meeting of the American College of Cardiology, Atlanta, Georgia, March 11 to 14, 2006.
Objectives The aim of this work was to determine whether mitral valve (MV) annuloplasty benefits patients with moderate/severe (3+/4+) functional ischemic mitral regurgitation (MR) who undergo coronary artery bypass grafting (CABG).
Background Mitral regurgitation is a strong predictor of poor outcomes in patients with ischemic cardiomyopathy; whether correcting it at the time of CABG improves outcomes is less certain.
Methods From 1991 to 2003, 390 patients with 3+/4+ ischemic MR had CABG with (n = 290) or without (n = 100) MV annuloplasty. Groups were propensity-matched using demographics, extent of coronary disease, regional wall motion, and quantitative electrocardiography. Survival, echocardiographic severity of MR, and New York Heart Association (NYHA) functional class were compared.
Results One-, 5-, and 10-year survival was 88%, 75%, and 47% after CABG alone and 92%, 74%, and 39% after CABG + MV annuloplasty (p = 0.6). Mortality was increased in patients with severe lateral wall motion abnormalities (p = 0.05), ST-segment elevation in lateral leads (p < 0.004), and higher QRS voltage sum (p < 0.0001). Patients undergoing CABG alone were more likely to have 3+/4+ postoperative MR than those undergoing CABG + MV annuloplasty (48% vs. 12% at 1 year, p < 0.0001). The NYHA functional class substantially improved in both groups (p < 0.001) and remained improved; at 5 years, 23% of patients having CABG + mitral annuloplasty and 25% having CABG alone were in NYHA functional class III/IV.
Conclusions Although CABG + MV annuloplasty reduces postoperative MR and improves early symptoms compared with CABG alone, it does not improve long-term functional status or survival in patients with severe functional ischemic MR. The MV annuloplasty in this setting, without addressing fundamental ventricular pathology, is insufficient to improve long-term clinical outcomes.
Functional ischemic mitral regurgitation (MR) is strongly associated with poor outcomes in patients with advanced coronary artery disease (1). However, whether its correction improves outcomes is unclear.
Ischemic MR is not caused by gross intrinsic disease of the valve, but by left ventricular (LV) remodeling, dilation, and dysfunction leading to geometric reconfiguration of the mitroventricular apparatus, including papillary muscle displacement and annular dilation. Mitral valve (MV) leaflets become tethered, with failure of anterior-posterior leaflet coaptation, resulting in symmetric or asymmetric regurgitation. Surgical treatment options include coronary artery bypass grafting (CABG) alone or with concomitant MV annuloplasty or replacement. Currently, the most common technique to restore valve competence is placing an undersized annuloplasty ring to reduce mitral annulus size.
Whether or not MV annuloplasty improves outcomes over and above CABG alone is debated, however (2). Therefore, in the absence of evidence-based on randomized trials, this clinical study has 2 objectives: 1) to determine the efficacy of MV annuloplasty in patients who are comparable to those undergoing CABG alone, based on extent of coronary artery disease, regional wall motion abnormalities, severity of MR, quantitative echocardiograms, and clinical characteristics; and 2) to discover clinically useful features and predictors of outcomes that might identify which patients, if any, are more likely to benefit from adding MV annuloplasty to their revascularization procedure.
From 1991 to 2003 at the Cleveland Clinic, 100 patients with chronic, severe (3+ or 4+) functional ischemic MR, previous myocardial infarction, and ejection fraction <45% by preoperative transthoracic echocardiography (TTE) underwent CABG alone (CABG group), and 290 underwent CABG with concomitant MV annuloplasty (CABG + MV annuloplasty group). Ischemic MR was defined as MR resulting from myocardial infarction, not from intrinsic MV disease (Carpentier types I and IIIb). Patients’ clinical records were reviewed to verify that their MR was caused by ischemic heart disease and myocardial infarction and did not just coexist with them. Patients who underwent other concomitant procedures (including LV restoration), those with ruptured papillary muscle, and 49 patients with atrial fibrillation, atrial flutter, or a pacemaker were excluded (because these rhythm disturbances adversely affected the quantitative echocardiogram analysis).
The decision to perform MV surgery was made by the surgeon and cardiologist on the basis of clinical presentation and findings at echocardiography and cardiac catheterization. All operations were performed on cardiopulmonary bypass using antegrade and retrograde cold blood cardioplegia. Coronary artery bypass grafting was performed in standard fashion, using a single-clamp technique. In patients undergoing MV annuloplasty, a left atrial approach was used and the valve inspected to confirm the diagnosis. Size of annuloplasty ring was smaller than the surface area of the anterior MV leaflet to ensure annulus “downsizing” (3).
Annuloplasty technique varied over the course of the study. They included partial flexible annuloplasty band (n = 207, 71%; Cosgrove-Edwards Annuloplasty System, Edwards Lifesciences, Irvine, California), rigid annuloplasty ring (n = 63, 22%; Carpentier-Edwards Classic Annuloplasty Ring, Edwards Lifesciences), and posterior suture plication with autologous pericardium or Peri-Guard graft (n = 20, 6.9%; Baxter Healthcare Corp., Deerfield, Illinois).
Perioperative inotropic therapy was guided by the hemodynamic condition of the patient. Routine postoperative use of dopamine or other inotropic agents was not practiced.
Preoperative and operative variables were retrieved from the cardiovascular information registry, augmented by quantitative echocardiogram (4) and echocardiography databases. All databases were approved for research by the institutional review board, with patient consent waived.
Recent work has highlighted the value of quantitatively assessed echocardiogram measures of LV mass, depolarization, and repolarization as potentially powerful predictors of risk in a variety of cardiovascular conditions (5,6). Because echocardiograms are inexpensive, routinely obtained, and relatively easy to describe quantitatively, we extended these observations to our study to refine propensity matching. Echocardiogram analyses included heart rhythm, right and left bundle branch block patterns, and locations of prior Q-wave myocardial infarctions. Using specialized software (Magellan Echocardiogram Research Work Station, General Electric Marquette Medical Imaging, Milwaukee, Wisconsin), computerized measures were made of P-wave duration; PR, QRS, and QT intervals; Cornell and Sokolow-Lyon voltages as measures of LV mass; and ST-segment changes. ST-segment slope was calculated as the difference of ST-segment deviation at the J-point and at the end of the ST-segment.
Preoperative ventriculograms and TTE were used to assess global LV function, graded regional wall motion, severity of MR, and regurgitant jet direction (Appendix). A postoperative TTE was performed routinely before hospital discharge and during follow-up at the discretion of the referring physician. None of the echocardiographic data in this report is based on intraoperative transesophageal studies (TEE).
Myocardial viability was assessed in only a small number of patients by magnetic resonance imaging (MRI) (33 patients) or positron emission tomography (PET) (76 patients). Only 21 patients (3 MRI, 18 PET) from matched cohorts had undergone myocardial viability assessment preoperatively. There was no statistically significant difference in presence of ischemia versus scar in any of the LV wall segments between groups. These data were insufficient to study the association of myocardial viability with postoperative outcomes.
Outcomes assessed were all-cause time-related mortality, TTE grade of postoperative MR, and postoperative New York Heart Association (NYHA) functional status. Patients undergoing CABG alone were systematically followed by questionnaire and telephone interview every 5 years for vital status, functional status, and cardiac events. Those undergoing CABG + MV annuloplasty were followed similarly at 2-year intervals. All follow-up was performed under institutional review board-approved protocols, with patient consent.
Median follow-up was 5 years for CABG alone and 4 years for CABG + MV annuloplasty; 10% of survivors were followed more than 11 and 9 years, respectively. We considered estimates of outcome reliable to 10 years.
Postoperative TTEs were available for a subset of 247 patients (63%) with 518 assessments of MR. These studies were concentrated early postoperatively, with median time to assessment of 5.3 months after CABG alone and 1.1 months after CABG + MV annuloplasty.
The NYHA functional class was derived algorithmically from a grid of symptoms and functional limitations self-reported by the patient. A subgroup of 257 patients (66%) had 403 postoperative functional status assessments at a median of 5 and 2.3 years for CABG alone and CABG + MV annuloplasty, respectively.
Data analysis first addressed dissimilarities between CABG alone and CABG + MV annuloplasty patients (Tables 1 and 2)⇓so that comparison of outcomes would be fair and based on more extensive risk adjustment than has heretofore been done.
Multivariable logistic regression was used to identify a parsimonious set of preoperative factors (Appendix) associated with performing CABG alone versus CABG + MV annuloplasty. Bagging was used for variable selection (7), with automated analysis of 500 bootstrap samples and a p value for retention of 0.05. Frequency of occurrence of individual factors or clusters of highly correlated factors in these analyses was counted (aggregation step), and factors appearing in 50% or more of the analyses were considered reliably statistically significant at a value of p ≤ 0.05.
To this parsimonious model were added nonsignificant variables representing groups of patient, coronary disease, regional wall motion, and echocardiogram variables that might be related to unrecorded selection factors (saturated propensity model) (8,9). Using this saturated model, a propensity score was calculated for each patient. Both CABG alone and CABG + MV annuloplasty cases were then matched on propensity score alone. This yielded 54 well-matched pairs (Tables 1 and 2).
Survival was estimated nonparametrically by the Kaplan-Meier method and parametrically by a multiphase hazard model (10). Two hazard phases were resolved, early and late, and because of this, proportional hazards modeling would have been inappropriate. Therefore, multivariable analyses were performed in the multiphase hazard function domain simultaneously for each hazard phase using variables in (Appendix), with treatment group (CABG alone) and propensity score forced into each (11). Otherwise, variable selection was by bagging, as described under “Fair comparison.”
Time-related change in postoperative MR among propensity-matched patients was analyzed by longitudinal ordinal logistic regression for repeated measures (PROC GENMOD, SAS Institute Inc., Cary, North Carolina). Because of low frequency, grades 3+ and 4+ were combined. Interval from operation to echocardiographic assessment, CABG alone versus CABG + MV annuloplasty groups, and interaction between interval and group were included in the comparison model to assess differences in rates of MR evolution.
Postoperative Functional Status
Time-related change in NYHA functional class among propensity-matched patients was analyzed using longitudinal ordinal logistic regression for repeated measurements, as described under “Postoperative MR.” Because of low frequency, NYHA functional classes III and IV were combined.
Benefit of MV Annuloplasty
Possible benefit of MV annuloplasty in reducing MR and improving functional class was assessed by examining overlap of confidence limits (CLs) of propensity-matched longitudinal estimates. Search for any survival benefit of MV annuloplasty was made by: 1) multivariable propensity-adjusted analysis for each hazard phase; 2) superimposing propensity-matched hazard functions for death; and 3) calculating the difference in expected 7-year survival with CABG alone versus CABG + MV annuloplasty. The latter was done by solving the multivariable equation for death twice for each patient, first by treatment received and then by treatment not received. These two 7-year estimates were subtracted and the difference examined by quintiles. This analysis suggested that patients with prolonged myocardial ischemic time benefited most from CABG alone. This time, however, is highly correlated with receiving MV annuloplasty and so was purposely not used in any analyses. Nevertheless, we developed another model of mortality that included ischemic times, carefully examining its possible interaction with treatment received.
Categorical data are summarized by frequencies and percentages, with comparisons made using the chi-square test, except as noted. Continuous variables are summarized as means ± SD and as equivalent 15th, 50th (median), and 85th percentiles when data were skewed. Comparisons were made using the Wilcoxon rank sum nonparametric test. All analyses were performed using SAS statistical software (version 9.1, SAS Institute Inc.). For consistency, uncertainty is expressed by CLs equivalent to ±1 standard error (68%).
In propensity-matched patients, those undergoing CABG + MV annuloplasty had longer cardiopulmonary bypass times (137 ± 42 min vs. 107 ± 38 min, p =0.0001) and myocardial ischemic times (97 ± 24 min vs. 74 ± 28 min, p < 0.0001) than those undergoing CABG alone, as expected. Reasons for not performing MV annuloplasty included downgrading of MR by intraoperative TEE (58%), MR not considered clinically significant by surgeon (11%), MR improved after CABG (8%), MR not addressed because of patient’s hemodynamic instability (5%), and reasons not noted by surgeon (18%).
More than 60% of patients in each group received at least 1 internal thoracic artery graft. Hospital mortality was 7.4% (CL 3.9% to 13%) after CABG alone and 3.7% (CL 1.3 to 8.4%) after CABG + MV annuloplasty (p = 0.7).
In both unadjusted comparisons (Fig. 1A)and those of propensity-matched patients (Fig. 1B), survival after CABG + MV annuloplasty was similar to that after CABG alone (p = 0.6). In matched pairs, it was 97% versus 94% at 30 days, 92% versus 88% at 1 year, 74% versus 75% at 5 years, and 39% versus 47% at 10 years. Risk factors associated with early death included severe lateral wall motion abnormality, renal insufficiency, lower hematocrit, and earlier date of operation; CABG + MV annuloplasty versus CABG alone was not a risk factor (p = 0.3) (Table 3).Risk factors associated with late death included older age, insulin-treated diabetes, renal insufficiency, higher QRS voltage sum, and larger positive and negative values of ST-segment slope in the fifth precordial echocardiogram lead (Fig. E1 in the Appendix), but not CABG + MV annuloplasty versus CABG alone (p = 0.2).
Patients undergoing CABG + MV annuloplasty experienced less severe postoperative MR compared with those undergoing CABG alone (p < 0.0001). After CABG alone, early presence and recurrence of MR was particularly pronounced, with the proportion of patients having no MR decreasing from 51% to 8% within the first few weeks of surgery (Figs. E2A and E2B in the Appendix). Recurrence was considerably slower after CABG + MV annuloplasty; the proportion of patients with no MR declined from 86% immediately postoperatively to 43% 6 months after surgery, while the proportion with 3+/4+ MR increased from 1% to 9%. The latter continued to increase gradually thereafter (Fig. 2).
Postoperative NYHA functional class
Substantial improvement occurred pre- to postoperatively in functional status in both groups (Figs. E3A and E3B in the Appendix). Comparison of temporal trends in propensity-matched groups showed that although the CABG + MV annuloplasty group appeared to have a slightly lower percentage of patients with severe symptoms (NYHA functional class III and IV) at 5 years, this difference was not statistically significant (23% for CABG + MV annuloplasty vs. 25% for CABG alone, p = 0.3). Over time, the percentage of patients with severe symptoms in the CABG + MV annuloplasty group remained constant, but the percentage in the CABG alone group increased, although graphically, confidence intervals overlap (Fig. 3).
Benefit of MV annuloplasty
The most apparent value of MV annuloplasty was immediate elimination of 3+ and 4+ MR, which was only partially mitigated by CABG alone. Patients experienced similar 7- to 8-year symptomatic benefit from both procedures (Fig. E3 in the Appendix). The early hazard function for death was more prolonged after CABG alone (Fig. 4),but in multivariable analysis, no early survival benefit of adding MV annuloplasty to CABG was evident (Tables 3and E1 [in the Appendix]). Analysis of predicted 7-year survival differences indicated that older patients with more comorbidities who had a longer period of myocardial ischemia (Fig. E4 in the Appendix) had somewhat worse survival after CABG + MV annuloplasty than after CABG alone; however, after adjusting for aortic clamp time, nearly all differences were <15%, and 60% were <10%.
Key findings were that: 1) CABG + MV annuloplasty almost completely eliminated MR, but this is only partially durable; 2) CABG with or without MV annuloplasty substantially improved symptoms; 3) addition of MV annuloplasty to CABG had no evident survival benefit; 4) survival of patients with ischemic cardiomyopathy and MR was primarily influenced by the presence of ischemia and ventricular dysfunction and hypertrophy at the time of operation.
Ischemic cardiomyopathy with associated functional MR is a leading cause of heart failure, affecting more than 5 million Americans. Limited treatment options for these patients have prompted aggressive surgical and interventional approaches to correct functional MR (2), with the hope of reversing ventricular remodeling, ameliorating symptoms of heart failure, and improving survival (3).
Optimal treatment of important ischemic MR at the time of CABG remains controversial, however. Some report that CABG alone results in improved ejection fraction, reduced MR, and better survival (12,13). Others indicate that CABG alone leaves many patients with clinically significant residual MR, which may be suboptimal therapy (14–16). These latter reports have resulted in wide use of MV annuloplasty at time of CABG, with anticipated improvement in symptoms and late survival (17). Nevertheless, a survival advantage of MV annuloplasty has not been demonstrated in this or other studies, despite reduced MR (18). These conflicting results likely reflect the heterogeneous nature of the disease, nonstandardized surgical treatments, and the absence of randomized study design. In addition, some reports mix ischemic cardiomyopathy with idiopathic dilated cardiomyopathy, making interpretation of results of MV annuloplasty difficult (19).
Choice of Treatment for Functional Ischemic MR
Most surgeons base their choice of procedure for functional ischemic MR on patients’ clinical status (severity of symptoms, risk profile) and MR severity, with great institutional and individual variance. Preoperative TTE is more likely than intraoperative TEE to reflect the true degree of MR, because MR is often artificially downgraded by altered hemodynamics in anesthetized patients. We, therefore, matched patients by severity of MR on preoperative TTE. Choice of procedure was based on the surgeon’s clinical assessment, with a number of patients undergoing CABG alone due to “downgrading” of MR by intraoperative TEE. It is important to emphasize that “downgrading of MR” did not change the underlying severityof MR, but only the perceptionof it.
Although assessment of myocardial viability provides important clinical information, it was not used as a guide in treating patients in our study, nor has it been in any major study on the same subject.
MV Annuloplasty and MR
Reduction of MR by MV annuloplasty using partial or complete annuloplasty rings is accomplished by reducing the septal-lateral diameter of the annulus, with resulting improved leaflet coaptation (3). Previous reports document 17% to 29% prevalence of recurrent MR early or late postoperatively (20,21). Annuloplasty technique in those studies was not uniform. The temporal pattern of developing postoperative MR was similar for rigid or flexible annuloplasty bands, but substantially worse for Peri-Guard (Baxter Healthcare Corp.) annuloplasty (22). Mitral valve annuloplasty in our study was usually performed using a partial flexible annuloplasty ring, our preferred current surgical technique. Mitral valve annuloplasty can also be performed with partial rigid rings or complete rings; however, differences in ring design have not yet translated into demonstrably improved outcomes (19). Recurrence of postoperative MR has also been associated with higher grade of preoperative MR and more severe LV dysfunction. In our study, these factors were controlled for by preoperative matching.
Although prevalence of important (3+ to 4+) postoperative MR (9% at 6 months, 20% at 5 years) in our study compares favorably with previous reports, these proportions remain substantial. This likely is the consequence of LV remodeling with an ongoing increase in papillary muscle displacement (22).
Patients with chronic ischemic MR are at markedly higher risk of heart failure than those with competent valves, and this risk is proportional to MR severity (1,23). The risk persists even after successful percutaneous or surgical revascularization, with resulting decreased postprocedural survival (15,24). This striking, well-documented association of ischemic MR and survival has resulted in the hypothesis that eliminating MR at revascularization improves outcomes. Unfortunately, we and others have failed to demonstrate improved survival of these patients with addition of MV annuloplasty. The reasons are unclear, but may indicate that functional ischemic MR is only a manifestation of advanced infarction-induced ventricular remodeling and that survival of these patients is primarily dictated by extent of their ischemic heart disease (25). This is substantiated by decreased survival associated with MV replacement that permanently eliminates MR but does not address ventricular remodeling (and may contribute to it) (26). Presence of myocardial ischemia and hypertrophy were found to be predictors of early and late risk in our study, further supporting this hypothesis.
Alternatively, this finding may be the result of inability of annuloplasty to completely and permanently eliminate MR. Although annuloplasty reduces MR in all patients in the short term, there is gradual recurrence in the long term. It remains to be seen whether improved annuloplasty ring design will result in long-term durable elimination of MR and improved functional status and survival (19).
Coronary artery bypass grafting with or without MV annuloplasty provides substantial and lasting symptom relief. However, there was a suggestion of progressive return of severe (NYHA functional class III/IV) symptoms late after CABG alone. This may indicate that worsening of heart failure symptoms is associated with residual or recurrent MR.
Benefit of MV Annuloplasty
Adding MV annuloplasty to CABG eliminates MR in the short term, improves symptoms—as does CABG alone—but has no demonstrable survival benefit. Nevertheless, prolonged myocardial ischemic time, particularly in older ill patients, increases the early risk of death.
This is a single-institution, nonrandomized clinical study conducted over a protracted period during which documented and undocumented differences in decision-making and surgical management occurred. We attempted to compensate for lack of randomization by more extensive propensity matching than has been done previously (27). This resulted in 2 groups of patients well matched on demographic characteristics, extent of coronary artery disease, LV wall motion abnormalities, severity of MR, and quantitative echocardiogram abnormalities. Although randomized trials represent the gold standard in clinical research, this approach may be difficult because of the heterogeneity of ischemic cardiomyopathy after myocardial infarction of varying extent, localization, and transmurality; variable LV remodeling in response; and variable remaining coronary artery disease.
We do not routinely obtain preoperative echocardiograms on patients’ underlying CABG, but do so on patients with a history of heart failure. Thus, some patients with important but unrecognized MR may have been missed. However, we believe the sample of patients is representative of the prevalence of severe ischemic MR in patients with coronary artery disease undergoing CABG. This is the largest study of such patients.
The groups were analyzed by treatment received, because we could not reliably ascertain treatment intended.
Nonsystematic echocardiographic follow-up is an important limitation. Postoperative echocardiograms were obtained at the discretion of treating cardiologists and were likely indicated by the clinical condition of the patient. This introduces the potential risk of overrepresentation of patients with severe postoperative MR. However, two-thirds of echocardiograms were obtained within the first year of operation, during which time it appears that most changes in MR grade occurred.
The question of whether recurrent MR directly affects survival is not possible to answer at present because both are outcomes of surgical intervention, one longitudinal and the other an event; therefore, their possible interdependent relationship with surgical technique awaits further development of relevant statistical methodology.
All these findings indicate that myocardial factors, rather than MR, are the primary determinants of outcome in patients with ischemic cardiomyopathy and MR. Any correction of MR, surgical or percutaneous, confined to the level of the annulus will likely result in only temporary reduction in MR, although possibly more rapid recovery from the operative intervention. If the benefit of MV annuloplasty, without addressing myocardial remodeling, occurs primarily early after CABG, one might predict that the benefit of percutaneous MV annuloplasty may also be limited to the period of early recovery after percutaneous coronary intervention. Neither may improve long-term survival.
The authors thank Mark Henderson, BS, BA, for downloading the electrocardiograms from the Marquette/GE system; Derlis Martino, MD, for quantitative echocardiogram analyses; and Kenneth Baker, MS, for assembling the resulting data. Christopher Pierce, PhD, assembled the echocardiographic data. Tanya Ashinhurst, BA, assisted in verifying cases and data and Linda DiPaola, BA, provided statistical programming. Karen Mrazeck, Capri Spencer, Kelly Polasko, and Wanda Weaver participated in patient follow-up, and Tess Parry, BS, provided editorial assistance.
For the variables used in analysis, Table E1, and Figures E1 through E4, please see the online version of this article.
Impact of Mitral Valve Anuloplasty Combined With Revascularization in Patients With Functional Ischemic Mitral Regurgitation
↵a Dr. Gillinov is a consultant to Edwards Lifesciences, Inc. and AtriCure, Inc.
- Abbreviations and Acronyms
- coronary artery bypass grafting
- confidence limits
- institutional review board
- left ventricular
- mitral regurgitation
- magnetic resonance imaging
- mitral valve
- New York Heart Association
- positron emission tomography
- transesophageal echocardiogram
- transthoracic echocardiogram
- Received September 8, 2006.
- Revision received February 5, 2007.
- Accepted February 5, 2007.
- American College of Cardiology Foundation
- Grigioni F.,
- Enriquez-Sarano M.,
- Zehr K.J.,
- Bailey K.R.,
- Tajik A.J.
- Holmvang L.,
- Clemmensen P.,
- Lindahl B.,
- et al.
- Okin P.M.
- Okin P.M.,
- Devereux R.B.,
- Nieminen M.S.,
- et al.
- Breiman L.
- Rosenbaum P.R.,
- Rubin D.B.
- Drake C.,
- Fisher L.
- Grigioni F.,
- Detaint D.,
- Avierinos J.F.,
- Scott C.,
- Tajik J.,
- Enriquez-Sarano M.
- Cohn L.H.,
- Rizzo R.J.,
- Adams D.H.,
- et al.