| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2005;79:1260-1267
© 2005 The Society of Thoracic Surgeons
a Division of Cardiovascular Surgery, Toronto General Hospital, and the Department of Surgery, University of Toronto, Toronto, Ontario, Canada
b Institute of Clinical Evaluative Sciences, Toronto, Ontario, Canada
Accepted for publication September 21, 2004.
* Address reprint requests to Dr Yau, 200 Elizabeth St, EN 13–239, Toronto, ON M5G 2C4, Canada (E-mail: terry.yau{at}utoronto.ca).
Presented at the Fortieth Annual Meeting of The Society of Thoracic Surgeons, San Antonio, TX, Jan 26–28, 2004.
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and MethodsResultsCommentDISCUSSIONReferences |
|---|
METHODS: Patients with IMR undergoing repair (n = 65) or replacement (n = 137) from 1990 to 2001 were evaluated. There were 87 patients with LV-IMR, and 115 patients with PM-IMR. Patients presenting in cardiogenic shock were excluded. Outcomes were evaluated by Cox survival analysis with propensity score adjustment and bootstrap validation.
RESULTS: Survival at 3, 5, and 9 years was, respectively, 0.94, 0.79, and 0.63 in the repair group, and 0.73, 0.67, and 0.59 in the replacement group. The hazard ratio (HR) of death for mitral repair versus replacement was not constant over the period of follow-up. Repair was associated with better early survival in the PM-IMR group, with an adjusted HR of 0.25 (95% confidence interval: 0.09 to 0.71) at 1 year. In the LV-IMR group and in patients with PM-IMR with high acuity and comorbidity, there was no significant survival advantage associated with repair. The beneficial effect of repair was not evident at late follow-up in either group. These findings were independent of the surgeon. Need for reoperation was more common after repair than after replacement (14% versus 3%, p = 0.003).
CONCLUSIONS: Patients with PM-IMR benefit from mitral repair with a significantly better early survival. However, the benefit of repair is not evident at longer follow-up. There was a nonsignificant trend toward greater early survival among patients with LV-IMR who underwent repair.
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDISCUSSIONReferences |
|---|
Studies that have examined this issue previously have showed variable results. That may be due to differences in clinical presentation and the pathophysiologic types of MR studied. In some cases, patients with nonischemic MR due to degenerative leaflet or chordal disease (Carpentier types Ia, Ib, IIa, IIb, IIIa; Table 1) [4] may have been misclassified as IMR. Most observational studies published before 2001 did not differentiate between true IMR and degenerative MR coexistent with CAD. Thus these studies reported variable results, with hazard ratios (HR) of repair versus replacement ranging from 0.39 [5] to 4.7 [6] and were generally inadequately adjusted for patient differences between study groups. The considerable heterogeneity of these patients contributes to difficulties in analysis. A randomized controlled trial, which would answer the question of repair versus replacement more definitively, has yet to be performed and may be difficult to implement.
|
To permit a meaningful comparison between repair and replacement, we attempted to adjust for differences in acuity, comorbidity, pathophysiologic type of IMR, extent of coronary disease, and experience of the surgeon. The change in outcome over time was also evaluated. Until a large, well-designed, randomized, controlled trial can be performed, this type of analysis, incorporating propensity score adjustment and a time-dependent effect of treatment, may yield the best available comparison between mitral repair and replacement for ischemic MR.
|
Patients and Methods |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDISCUSSIONReferences |
|---|
Based on all the data available, IMR was classified as one of two types: papillary muscle–related (PM-IMR), with Carpentier type IIc, IId, or IV pathophysiology; and left ventricle–related (LV-IMR), with Carpentier type Ia or IIIb pathophysiology. Patients with type Ia MR were included in this group only if there was no evidence of associated degenerative valve disease.
The primary endpoint was overall time-related survival. Secondary endpoints were hospital mortality, reoperation, anticoagulation-related hemorrhage (requiring hospitalization or blood transfusion), and stroke (neurologic deficit lasting more than 24 hours). The varying techniques of mitral repair or the type of mitral prosthesis used were described but were not included in the analysis because of inadequate power to evaluate their effects independently.
Statistical Methods
The study was powered to detect a HR of death, related to repair versus replacement, of 0.52 or more extreme. Exploratory univariable analyses were done with Wilcoxon's rank-sum and Pearson's 
2 tests for continuous and categorical variables, respectively. The variable selection methods described by Harrell [7] were used to specify an a priori model. The Cox proportional hazards (PH) model extended to handle variation in hazards over time with propensity score adjustment was used for the primary outcome [8]. Frailty and propensity score models were used to adjust for the surgeon effect [7, 9]. Hospital mortality was analyzed by logistic regression and the remaining secondary outcomes were described as frequencies and tested by Fisher's exact test. (See the Appendix for details of statistical methods.)
|
Results |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDISCUSSIONReferences |
|---|
|
|
|
|
|
|
|
Adding the propensity score with surgeons as a regressor in the model did not significantly change the HRs of repair or the repair-by-time interaction or their standard errors. Therefore, the propensity analysis also suggested that the benefit of mitral repair did not depend on the operating surgeon.
Secondary Outcomes
The frequencies of the remaining secondary outcomes, including reoperation, stroke, and major hemorrhage, were insufficient for multivariable modeling. The crude rates and Fisher's exact tests are therefore presented in Table 4.
|
|
|
Comment |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDISCUSSIONReferences |
|---|
It is important to note that the reduction in risk of mortality associated with mitral valve repair for IMR was not uniform over the duration of follow-up, but was most marked during the index hospitalization and the subsequent 2 to 3 years. The survival advantage of mitral repair was no longer evident over extended follow-up, a finding that is demonstrated graphically in Figure 4. Gillinov and colleagues [10], in the only other study to evaluate hazard ratios over time, noted a similar nonproportionality over follow-up time, with an early HR of death of 0.21 (CI: 0.11 to 0.35) in favor of repair, but a late HR of death of 1.01 (CI 0.94 to 1.11). Their analysis suggested, as does ours, that mitral repair was associated with lower operative mortality and better short-term survival, but that this survival advantage did not persist. However, Gillinov and colleagues [10] also concluded that there was no difference between repair and replacement in the subgroup of patients with highest acuity and comorbidity. Similarly, in our analysis the initial survival benefit from mitral repair in PM-IMR patients was not seen in patients with high acuity and comorbidity scores. Moreover, for patients with LV-IMR and high acuity and comorbidity (Fig 5), there may be a trend for an early survival disadvantage with mitral repair.
Other studies comparing mitral repair and replacement for IMR have not reported any tests of the assumption of proportional hazards, and one cannot therefore exclude variations of the hazard ratio over time. Grossi and associates [11] and Tavakoli and associates [12] compared mitral repair and replacement in patients with IMR with adjustment for baseline differences and reported conflicting results. In the study of Grossi and colleagues [11], the HR was 0.45 (CI: 0.27 to 0.94) in favor of repair if New York Heart Association (NYHA) class was removed from the analysis. In a small study, Tavakoli and colleagues [12] reported a HR of death of 0.72 (CI: 0.40 to 1.25) favoring repair. Other studies, with less explicit adjustment for differences between groups, have drawn conflicting conclusions, perhaps indicating the importance of evaluation of risk over time.
The disappearance of the survival advantage initially associated with mitral repair over longer-term follow-up in this population may be related to the dynamic nature of ischemic MR. As the left ventricle continues to adversely remodel after infarction, and coronary or graft arteriosclerosis progresses, symptomatic mitral regurgitation may recur. This recurrence resulted in reoperations on the mitral valve in a total of 14% of patients who had undergone initial repair. These reoperations, it should be noted, were associated with no operative mortality. In contrast, only 3% of the valves replaced at the time of initial operation required rereplacement. However, anticoagulant-related hemorrhage tended to be more prevalent in this group, although the difference did not reach statistical significance (p = 0.06).
Almost all patients who underwent mitral repair had an annuloplasty ring, and a significant number had only an annuloplasty ring placed (48%). The median annuloplasty ring size was 30 mm, not quite as undersized as recommended by Bolling and colleagues [13]. Unfortunately, the small number of outcomes prohibited us from testing more detailed hypotheses including the effect of annuloplasty ring size and type of repair.
Evaluation of Effect of Surgeon
Clearly one of the most important predictors of outcomes after mitral repair is the expertise of the operating surgeon. It is therefore important to determine whether our conclusions were skewed by above-average results after repair in the hands of one or two surgeons. We utilized both a random effects (frailty) model and propensity score adjustment to account for this effect. The frailty model is intermediate between adding each surgeon as a variable and ignoring the effect of surgeon. The random effect variable included adjustment for between-surgeon variation, allowing one to comment on within-surgeon variation. The propensity score model, on the other hand, balances the effect of the surgeon (in this case the two highest volume surgeons) between groups. Neither analysis influenced the conclusions drawn. Although patient characteristics and overall outcomes did vary between surgeons, the relationship between the results of mitral repair and those of mitral replacement was remarkably consistent across surgeons. These findings suggested that our conclusions applied to at least the majority of the surgeons at this institution, and would suggest that they are generalizable to others as well.
Limitations of This Study
This was an observational and not a randomized study, conducted by retrospective review of prospectively gathered data. By using several balancing methods, however, we attempted to adjust for baseline differences and selection bias. The limited number of outcomes was approached by reducing the number of potentially predictive variables included in the model without significant data loss, by the use of hierarchical variable analysis and principal component scores. However, because of the limited number of outcome events, some important clinical questions that require testing for specific interactions within specific subgroups may have not been answered, eg, whether different types of repair influenced outcomes.
Propensity scores were added as a regressor in the Cox model to adjust for selection bias. Using the propensity score to divide the population into quintiles and repeating the Cox model within each quintile, instead of using the propensity score as an additional predictor within the regression model, may be methodologically preferable. However, due to the limited number of patients in each quintile (approximately 40 patients) this technique could not be reliably employed.
Conclusions
In patients with chronic ischemic MR resulting from PM infarction or partial rupture who are not presenting in cardiogenic shock, mitral repair is associated with a significantly reduced risk of hospital and short-term mortality. Although patients with LV-related IMR did have a trend toward a short term survival benefit from repair, this trend did not reach statistical significance. However, this survival advantage does not persist in either group, and may reach equivalence with mitral replacement after the fourth postoperative year.
Despite the lack of persistence of this initial survival benefit due to repair, this early difference will translate into many patient-years of increased survival, especially in PM-related IMR. The likelihood of reoperation after initial mitral repair is low, and was not associated, in our experience, with prohibitive risks.
|
Appendix |
|---|
The extended Cox model (extended to handle nonproportional hazards) [8] was used to study the primary outcome. A time-dependent covariate was included to assess the variation of treatment effect over time. To adjust for imbalance between the comparison groups, covariate adjustment and propensity score adjustment were used. The assumption of proportional hazards, ie, the assumption that the effect of repair versus replacement on mortality was constant over time, was tested graphically and by inclusion of a time-dependent covariate. Propensity model assumptions were tested by assessing group differences within each quintile of data generated by the propensity score.
Evaluation of Effect of Surgeon
Convergence between three methods was used to assess the surgeon effect. First, a random effects variable (frailty) was used as a regressor in the Cox model [8]. Second, the two surgeons with the highest volume were included (as binary variables) in the propensity score model. The propensity scores were then used as a covariate in the Cox model. Finally a sensitivity analysis was conducted including only patients treated by the highest volume surgeon, and by all other surgeons.
Model Validation
Bootstrap resampling was used to validate the model coefficients and discrimination statistics. One thousand resamples were obtained with replacement and the models were repeated. The original model coefficients and discrimination statistics were compared with the bootstrap means of the respective parameter. Parameters within one standard deviation of the bootstrap mean were considered valid.
Jackknife after bootstrap analysis was done to exclude overly influential observations. This was done by repeating all analyses and model validations after removing one observation (patient) at a time.
All the analyses were conducted with the R (www.r-project.org) and S-Plus (Insightful Corp) statistical package. The additional functions of the Survival library [8], and the Hmisc and Design libraries [7] were used. The resample (beta version) library of S-Plus was used for model validation.
|
DISCUSSION |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDISCUSSIONReferences |
|---|
In this group they compare the outcome of mitral valve reconstruction versus mitral valve replacement in an ischemic group. They determined that they thought they saw an early benefit of mitral valve repair which tapered off over the course of 7 to 9 years of follow-up. This is a retrospective, propensity matched, bootstrapped, validated, nonrandomized, nonprospective study, and that should be remembered.
In the manuscript there is very little discussion about the decision to repair or replace each individual mitral valve. How was that decided?
Secondly, there was very little discussion at the presentation about the operative mortality, which in the manuscript was presented as 1 of 65 patients in the repair group, for 2%, and 29 of 137, for a 22% operative mortality in the mitral valve repair group.
Also, what was the mechanism of death in reoperation in the mitral valve repair group? There was very little echocardiography data that we saw presented here and very little presented in the manuscript. Was it the mechanism of recurrent or residual mitral regurgitation? Was there echo follow-up for all the patients? When were the echoes obtained? Were there different geometric distortions in the echoes? Was the papillary muscle displaced? We would like to see that the mechanism of death in the patients who had mitral valve repair was not just a poor or failed mitral valve repair.
It is well known that residual or recurrent mitral regurgitation is not tolerated by the poor ischemic ventricle, and it may be, if these groups are separated into those who had a good repair versus those who had a failed repair, the good repair may have a sustained effect.
Were all the patients who had residual or recurrent mitral regurgitation reoperated upon, or were some patients not reoperated upon and it resulted in their death?
And as a final question, a ring size of 30 appears to be quite large. This is a historical retrospective review. Did you have enough numbers in this group to correlate ring size with outcome, ring size with recurrent or residual mitral regurgitation?
I thank the authors for the privilege of looking at this manuscript ahead of time and also thank The Society.
DR AL-RADI: I thank Dr Bolling for his comments and questions. The decision whether repair or replacement was done and what repair was done was made by surgeon, therefore it was an important part of our analysis to adjust for the effective of the surgeon. Multiple techniques were used to make sure that there is minimal confounding effect by surgeon, ie, decision-making, on the outcomes. We have shown that the outcomes were not dependent on who was operating and who made the decision to repair.
Regarding operative mortality, the odds ratio for operative mortality was strongly in favor of repair, with an odds ratio of 0.12.
In regard to echocardiography data, unfortunately, we didn't have follow-up echo data and a sufficient number of patients to perform a multivariable analysis; however, none of the patients had reoperations at other institutions, and it is highly unlikely that a patient had significant recurrent MR and did not undergo reoperation at our institution. If such a patient did succumb to their disease, they would be represented in the survival data as a death.
In terms of reoperation for residual MR and the effect of ring size, in the 9 patients who required reoperation after mitral repair—and there were 3 patients who had an annuloplasty ring only as the original repair, and the ring size in those patients was also about the same—the mean was 31 mm, and it was 30 mm in all patients who had repair. The 1 patient who had papillary muscle reimplantation failed, and 2 patients had chordal transfer, and 1 of them failed. So people who required more complex repairs were probably more likely to fail if they had IMR. It is also interesting that rigid rings like the Carpentier ring were less likely to fail; however, the numbers were small and not enough to even do a univariate test.
|
References |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. Magne, N. Girerd, M. Senechal, P. Mathieu, F. Dagenais, J. G. Dumesnil, E. Charbonneau, P. Voisine, and P. Pibarot Mitral Repair versus Replacement for Ischemic Mitral Regurgitation: Comparison of Short-Term and Long-Term Survival Circulation, September 15, 2009; 120(11_suppl_1): S104 - S111. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Kempfert, J. Blumenstein, M. W.A. Chu, P. Pritzwald-Stegmann, T. Kobilke, V. Falk, F. W. Mohr, and T. Walther Minimally invasive off-pump valve-in-a-ring implantation: the atrial transcatheter approach for re-operative mitral valve replacement after failed repair Eur. J. Cardiothorac. Surg., June 1, 2009; 35(6): 965 - 969. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. D. Barnett and N. Ad Surgery for aortic and mitral valve disease in the United States: A trend of change in surgical practice between 1998 and 2005. J. Thorac. Cardiovasc. Surg., June 1, 2009; 137(6): 1422 - 1429. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. A. Milano, M. A. Daneshmand, J. S. Rankin, E. Honeycutt, M. L. Williams, M. Swaminathan, L. Linblad, L. K. Shaw, D. D. Glower, and P. K. Smith Survival Prognosis and Surgical Management of Ischemic Mitral Regurgitation Ann. Thorac. Surg., September 1, 2008; 86(3): 735 - 744. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Fukuoka, M. Nonaka, S. Masuyama, T. Shimamoto, K. Tambara, H. Yoshida, T. Ikeda, and M. Komeda Chordal "translocation" for functional mitral regurgitation with severe valve tenting: An effort to preserve left ventricular structure and function J. Thorac. Cardiovasc. Surg., April 1, 2007; 133(4): 1004 - 1011. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Magne, P. Pibarot, F. Dagenais, Z. Hachicha, J. G. Dumesnil, and M. Senechal Preoperative Posterior Leaflet Angle Accurately Predicts Outcome After Restrictive Mitral Valve Annuloplasty for Ischemic Mitral Regurgitation Circulation, February 13, 2007; 115(6): 782 - 791. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A. Borger, A. Alam, P. M. Murphy, T. Doenst, and T. E. David Chronic Ischemic Mitral Regurgitation: Repair, Replace or Rethink? Ann. Thorac. Surg., March 1, 2006; 81(3): 1153 - 1161. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
