Mitral valve annuloplasty versus replacement for severe ischemic mitral regurgitation

Although practice guidelines recommend surgery for patients with severe chronic ischemic mitral regurgitation (CIMR), they do not specify whether to repair or replace the mitral valve. 436 consecutive patients with severe CIMR were eligible for inclusion in the study, of which 316 (72.5%) underwent mitral valve annuloplasty (MVA) whereas 120 (27.5%) received mitral valve replacement (MVR). At 59 months (interquartile range, 37–85 months) follow-up, though the left ventricle end-diastolic diameter was markedly larger (P = 0.019) in the MVA group than in the MVR group, no significant difference was observed in overall survival, freedom from cardiac death, or avoidance of major adverse cardiac or cerebrovascular events (MACCE). MVA provides better results in freedom from cardiac death in subgroups of age ≥65years and left ventricular ejection fraction (EF) ≥50% (P = 0.014 and P = 0.016, respectively), whereas MVR was associated with a lower risk of MACCE in subgroups of age <65years, EF <50% and left ventricular inferior basal wall motion abnormality (BWMA) (all P < 0.05). In conclusion, MVR is a suitable management of patients with severe CIMR, and it is more favorable to ventricular remodeling. The choice of MVA or MVR should depend on major high-risk clinical factors.

Risk factors and prespecifed subgroup analysis. Cox proportional hazard model analysis showed that both age and preoperative EF were independent predictors of cardiac death at follow-up (for age: HR, 1.04; 95% CI, 1.01-1.07, P = 0.011; and for EF: HR, 0.95; 95% CI, 0.93-0.97, P < 0.001). Of note, the choice of MVA or MVR was not a significant predictor of late cardiac death (P = 0.233) ( Table 3). We also assessed the relative surgical  procedures effects in subgroups of patients with major high-risk clinical factors. MVA provides better results than MVR in terms of freedom from cardiac death in subgroups of age ≥65years and EF ≥50% (for age: P = 0.014; and for EF: P = 0.016), whereas MVR was associated with a lower risk of MACCE than MVA in subgroups of age < 65years, EF < 50% and BWMA (for age: P = 0.010; for EF: P = 0.007, and for BWMA: P = 0.016) (Fig. 1).
Results of propensity score matching analysis. In the propensity score matching analysis, 109 pairs were extracted by 1:1 manner using nearest neighbor matching without replacement. There was no signifcant difference between the two propensity-matched groups with regard to baseline characteristics (Table 1). However, compared with MVA, the aortic cross-clamp time, cardiopulmonary bypass time and the duration of intensive care unit (ICU) for MVR were markedly longer (P < 0.05) ( Table 1). The incidences of early mortality and postoperative complications (stroke, reoperation for bleeding, application of intra-aortic balloon pump and acute renal failure) were also similar between the two propensity score-matched groups, except for a higher incidence of respiratory complications in the MVR group than in the MVA group (Table 4). During the follow-up, compared with the MVR group, the left ventricle end-diastolic diameter was markedly larger (P = 0.019), and the incidence MVA n (%) MVR: n (%) Adjusted HR # (95% CI) P Value  Table 3. Cox proportional hazard analysis for cardiac death at long-term follow-up. HR: hazard ratio, CI: confdence interval, EF: left ventricular ejection fraction, IABP: intra-aortic balloon pump. * Indicates mitral valve annuloplasty or replacement.  of mitral regurgitation recurrence was significantly higher in the MVA group (P < 0.001) (Table 5), however, we observed no significant difference in overall survival, freedom from cardiac death or MACCE between MVA and MVR (Fig. 2, all P > 0.05).

Discussion
In the present study, the incidences of early mortality and postoperative complications were similar between the two propensity score-matched groups. According to published literatures, compared with MVR, early mortality and complications for MVA fall into 2 categories: no difference and lower incidence. Several recent experiences have failed to detect any substantial difference between the two surgical managements in terms of in-hospital mortality or complications, which are in accordance with our observations 2,9 , whereas several studies showed that mitral valve repair is associated with lower operative mortality 7,8 .
In the present long-term observational study, after adjustment for baseline differences with Cox proportional hazard model and propensity score matching analysis, we observed no significant difference between MVA and MVR in risks of MACCE, cardiac death, or overall death. Follow-up echocardiographic results of propensity score-matched patients showed that, compared with the MVR group, the left ventricle end-diastolic diameter was markedly larger, and the incidence of mitral regurgitation recurrence was significantly higher in the MVA group. The optimal surgical approach to the treatment of severe ischemic mitral regurgitation remains controversial. Published series have provided a wide range of results for long-term outcomes. A multicenter, randomized trial conducted by the Cardiothoracic Surgical Trials Network for severe ischemic mitral regurgitation showed that two-year mortality was 19.0% in the MVA group and 23.2% in the MVR group (P = 0.39) 2 . An important study carried out by Lorusso and colleagues showed that eight-year survival was 81.6% ± 2.8% versus 79.6% ± 4.8% in MVA and MVR, respectively (P = 0.42) 3 . A recent meta-analysis showed that MVA is associated with higher recurrence of MR in patients with CIMR, and no difference was found regarding survival, NYHA class, and functional indicators 7 . Cohn and colleagues 10 reported a 5-year survival of 56% and 91.5% in MVA and MVR, respectively, whereas a meta-analysis showed that the relative long-term risk of death was 35% higher in the MVR group than in the repair group 11 .
Such different conclusions might have derived from the heterogeneity of patient cohorts. Therefore, in the present study, we included only patients undergoing MVA or MVR with complete myocardial revascularization. We also excluded patients with congenital valvular heart disease, rheumatic valvular disease, infective endocarditis, presence of aortic valve regurgitation or stenosis, or receiving other procedures. Moreover, to minimize the effects of confounding variables, a propensity score model was constructed.
An important study showed that both older age and lower preoperative left ventricular EF were independent predictors of cardiac death at follow-up 12 , which was in consistent with the present study. Considering MVA and MVR have different characteristics in terms of operative mortality, long-term correction of mitral regurgitation   and long-term thromboembolism, mitral procedure selection should be individualized and depend on major high-risk clinical factors. As a result, we assessed the effects of both surgical procedures in subgroups of patients with major high-risk clinical factors. In this study, prespecifed subgroup analysis showed that, during long-term follow-up, MVR was associated with a lower risk of MACCE than MVA in subgroups of age < 65years, EF < 50% and BWMA. Compared with MVA, MVR provides a considerably more durable correction of MR 2,8 , which may have a beneficial effect on long-term outcomes. However, this effect must be weighed against any potential adverse consequences of a prosthetic valve, such as long-term thromboembolism, endocarditis, and structural valve deterioration 2 . Recent studies have showed that, compared with MVR, MVA provides worse long-term MR correction with risks of adverse left ventricular remodeling, readmission and poor survival 2 . Multiple studies have attempted to develop predictive models of MR recurrence based on preoperative parameters [13][14][15][16] . Some studies pointed out that poor LV function and preoperative basal aneurysm/dyskinesis were key predictive factors of MR recurrence 13,17 . BWMA reflects severe LV ischemic remodeling including papillary muscle displacement, and leaflet tethering, all of which influence CIMR. Kron and colleagues concluded that the presence of preoperative basal aneurysm/dyskinesis was strongly associated with MR recurrence, and the mechanism for MR recurrence was largely mitral valve leaflet tethering 16 .
There are still some limitations. First, this study reports retrospective data from a single center and is subject to all the limitations inherent to this design. Second, the small study sample might have led to type II statistical errors. An appropriately powered, randomized, controlled trial evaluating the optimal management of CIMR would be useful to confirm our results. Third, pre-, intra-, and postoperative information about the exact mechanisms and characteristics of MR were not available in all patients.
In the current study, though MVR was more favorable to ventricular remodeling than MVA, they provided comparable results in terms of overall survival, freedom from cardiac death and avoidance of MACCE at follow-up. Mitral procedure selection should be individualized and depend on major high-risk clinical factors. With age ≥65years or EF ≥50%, MVA is the optimal management of severe CIMR, whereas MVR would be a better alternative under the condition of age < 65 years, EF < 50% or BWMA.

Methods
Patients and study design. This study was approved by the Human Research Ethics Committee of the Fuwai Hospital and was performed in accordance with the Declaration of Helsinki and the approved guidelines. Oral informed consent was obtained from all of the patients via a telephone questionnaire. CIMR was defined by coronary angiographic and echocardiographic findings according to accepted criteria, i.e., 1) Mitral regurgitation (MR) occurring more than 16 days after myocardial infarction; 2) type I/IIIb leaflet dysfunction following Carpentier's classification; and 3) 70% or greater stenosis of at least one coronary artery, with wall motion abnormalities of the corresponding left ventricular segment 3 .
Between January 2002 and December 2014, a total of 1066 patients with CIMR were hospitalized for undergoing coronary artery bypass grafting (CABG) combined with MVA or MVR. From the initial cohort, 630 patients were excluded for various reasons 3 , i.e., 1) Preoperative MR ≤2+, congenital valvular heart disease, rheumatic or degenerative valvular disease, infective endocarditis, presence of aortic valve regurgitation or stenosis, emergency surgery, or repeat operation; or 2) performance of other procedures, such as left ventricular reconstruction/ reshaping, or procedures other than mitral ring annuloplasty for the treatment. Thus, the final study cohort comprised 436 patients: 316 patients (72.5%) underwent MVA whereas 120 (27.5%) underwent MVR. 7 (1.6%; 5 in MVA group and 2 in MVR group) patients were lost to follow up.
Baseline patient characteristics, echocardiography data, operative data, and surgical techniques were collected from the division of cardiovascular surgery's database and individual medical records. Patients were followed up through internet or telephone interviews and the outpatient department records. All collected data were sent to a core lab (State Key Laboratory of Cardiovascular Disease, Beijing, China) for analysis. Surgical technique. All surgical procedures were performed with standard bypass techniques through median sternotomy by senior surgeons with special interest in mitral valve surgery. The decision to perform MVA or MVR was at the surgeon's discretion. Downsizing ring annuloplasty (2 sizes) was used in all patients subjected to MVA 3,18 . Subvalvular apparatus were preserved for MVR whenever possible (94/120, 78.3%), including posterior leaflet preservation, posterior and partial anterior leaflet preservation, both leaflets preservation. The decision to perform which kind of procedure was at the surgeon's discretion according to the condition. The posterior leaflet preservation was performed in 58 patients undergoing MVR. In 8 of patients undergoing MVR, the middle portion of the anterior leaflet was resected and the remaining leaflet tissue was plicated with the individual valve sutures. In 28 of the patients undergoing MVR, the anterior leaflet of the valve was partly or completely detached from the mitral annulus and divided in the middle at the 12 o' clock position, and the leftward portion of the anterior leaftlet was plicated to the anterolateral commissure with a pledgetted 4-0 polypropylene suture. The rightward portion of the anterior mitral leaflet was similarly plicated to the posteromedial commissure. Complete revascularization was achieved in all patients with arterial conduits or saphenous vein grafts. All patients received the same perioperative care and medical therapy according to guidelines. A stepwise multivariable Cox proportional hazards model was developed to determine the independent risk factors. Variables with a P value less than 0.15 in the univariate analyses were entered into multivariable models. Differences in risk-adjusted, long-term rates of study outcomes among patients underwent different surgical procedures were assessed by use of multivariable Cox proportional hazards regression with adjustment for all patient-level variables in Table 1. Cumulative event rates were calculated using a Kaplan-Meier method, and different event curves of outcomes were compared using a Log-Rank test. Surgical procedures related differences in long-term outcomes were also analyzed in high-risk clinical subgroups.

Echocardiography.
To reduce the impact of treatment selection bias and potential confounding in the observational study, we performed rigorous adjustment for baseline differences by use of propensity score matching 22 . A propensity score representing the probability of having MVR as opposed to MVA was calculated for each patient by using a nonparsimonious multivariable logistic regression model. Variables used in the model are shown in Table 1. Pairs of patients with MVA and MVR were matched using calipers of width 0.2 standard deviations of the logit of the propensity score 23 . Model discrimination was assessed with C statistics, and model calibration was assessed with Hosmer-Lemeshow statistics. Finally, 109 pairs of patients were matched to obtain risk-adjusted outcome comparisons between the two groups.