Article Text

Download PDFPDF

Original research article
Clinical outcomes in 1731 patients undergoing mitral valve surgery for rheumatic valve disease
  1. Wan Kee Kim,
  2. Ho Jin Kim,
  3. Joon Bum Kim,
  4. Sung-Ho Jung,
  5. Suk Jung Choo,
  6. Cheol Hyun Chung,
  7. Jae Won Lee
  1. Department of Thoracic and Cardiovascular Surgery, Asan Medical Center, University of Ulsan College of Medicine, Seoul, South Korea
  1. Correspondence to Dr Joon Bum Kim, Department of Thoracic and Cardiovascular Surgery, Asan Medical Center, University of Ulsan College of Medicine, Seoul 138-736, South Korea; jbkim1975{at}


Objective Unlike degenerative mitral valve (MV) disease, the advantages of valve repair procedure over replacement have been debated in rheumatic MV disease. This study aims to evaluate the impact of procedural types on long-term outcomes through analyses on a large data set from an endemic area of rheumatic disease.

Methods We evaluated 1731 consecutive patients (52.3±12.5 years; 1190 women) undergoing MV surgery for rheumatic MV disease between 1997 and 2015. Long-term survival and valve-related outcomes were compared between repair and replacement procedures. To adjust for selection bias, propensity score analyses were performed.

Results Patients undergoing repair were younger and had more predominant mitral regurgitation than mechanical and bioprosthetic replacement groups (61.6% vs 15.6% vs 24.4%; P<0.001). During follow-up (130.9±27.7 months), 283 patients (16.3%) died and 256 patients (14.8%) experienced valve-related complications. Propensity score matching yielded 188 pairs of repair and replacement patients that were well balanced for baseline covariates. In the matched cohort, there was no significant difference in the mortality risk between the repair and replacement groups (HR, 1.24; 95% CI 0.62 to 2.48). The risk of composite valve-related complications, however, was significantly lower in repair group (HR, 0.57; 95% CI 0.33 to 0.99) principally derived by a lower risk of haemorrhagic events (HR, 0.23; 95% CI 0.07 to 0.70). The incidence of reoperation was not significantly different between groups in the matched cohort (HR, 1.62; 95% CI 0.49 to 5.28).

Conclusion Valve repair in well-selected patients with severe rheumatic MV disease led to comparable survival, but superior valve-related outcomes compared with valve replacement surgery.

  • mitral stenosis
  • mitral regurgitation
  • valve disease surgery

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Despite the significant advantages of the valve repair procedure over replacement in degenerative mitral valve (MV) disease, whether it has similar advantages in rheumatic heart disease has been debated for decades.1–3 Although there is a general agreement that the majority of patients with severe rheumatic MV diseases who require surgery are best treated with valve replacement surgery, valve repair techniques are also frequently used with excellent short-term and mid-term outcomes in experienced centres.4 5 Even with early successful outcomes of the rheumatic mitral repair, however, there still are concerns on its long-term durability because the rheumatic pathology can still progress in the mitral apparatus over time.

To evaluate the impact of procedural types on long-term clinical outcomes, we present here our experience of rheumatic MV surgery over 20 years with a reasonably sized data set.


Study subject and outcome measures

From the Institutional Cardiac Surgical Database of the Asan Medical Center, Seoul, Korea, 1877 adult (>17 years old) patients who underwent MV surgery for rheumatic MV disease between January 1997 and December 2015 were identified. Of these, patients who underwent concomitant cardiac tumour resection or infective endocarditis and those who had previously undergone mitral surgery were excluded. Finally, a total of 1731 patients formed the subjects of this study. The patients were categorised based on the type of MV surgery: repair (n=294), mechanical prosthesis (n=1134) and bioprosthetic prosthesis (n=303).

Primary outcomes of interest were death and valve-related complications as defined in the Society of Thoracic Surgeon Guidelines,6 including reoperation of the MV, thromboembolic events, anticoagulation-related bleeding complications and infective endocarditis. Early mortality was defined as occurring in-hospital or within 30 days of surgery.

The requirement for informed consent from individual patients was waived due to the retrospective nature of the study design.


Preoperative echocardiography was performed in all patients within 3 months before surgery. Echocardiographic parameters were set using standardised methods as previously described.4 Significant pulmonary hypertension was defined as a tricuspid regurgitation peak velocity >3.4 m/s, equal to a pulmonary artery pressure of >50 mm Hg. MV area was estimated by using the pressure half-time method, and significant mitral stenosis was defined by an MV area ≤1.5 cm2.7

In general, a follow-up echocardiographic exam was performed at 6 months after discharge. Thereafter, echocardiographic exams were performed annually or biennially.

Surgical procedures

With a median sternotomy (n=1459) or a thoracotomic approach (n=272), a longitudinal right-sided left atriotomy was the conventional approach. Deciding whether to perform a valve repair technique or fit a replacement depended heavily on the severity and modes of the MV dysfunction. Flexibility of the anterior leaflet was the most important indicator for rheumatic MV repair surgery. In highly selected cases with stenotic MV dysfunction, simple commissurotomy was conducted when fibrotic lesions were limited to the commissural sites (online supplementary video). Age was not used as a sole criterion for rheumatic MV repair, but younger patients tended to have more favourable MV anatomies. Therefore, younger individuals were more likely to undergo repair surgery. On the other hand, those with severe involvement of rheumatic fibrosis extending to the subvalvular apparatus were not regarded as fair candidates for repair. In addition, presentation with a predominant stenotic lesion with diffuse leaflet calcification was another unfavourable indication for a repair procedure. The final decision on reparability, however, depended on the surgeon’s discretion. For those considered appropriate to undergo MV replacement, the choice of the prosthetic types (mechanical vs bioprosthesis) was left to the patient after being given full information regarding the pros and cons of each, which were primarily based on the patient’s age and comorbidities.

Supplementary file 1

Details of the MV repair techniques are as follows: the deformed, restricted chordae were replaced with artificial thread using 4-0 or 5-0 polytetrafluoroethylene. Release of subvalvular apparatus was performed in amenable cases by myotomic incision of fused papillary muscles to obtain appropriate leaflet motion. Leaflet mobilisation, usually using bilateral commissurotomy, was performed to relieve the restrictive leaflet motion and thereby to eliminate type IIIb mitral insufficiency as well as to secure adequate MV opening. The leaflet extension technique was infrequently used as those requiring this technique were regarded as best treated with replacement surgery. Following these techniques, complete rigid annuloplasty rings were inserted in most cases to maintain the anteroposterior dimension of the mitral annulus. Details of the repair techniques are summarised in table 1.

Table 1

Details of MV repair procedures

In cases undergoing valve replacement surgery, subvalvular apparatus was retained whenever possible. MV competency was routinely assessed intraoperatively using transoesophageal echocardiography, after weaning from cardiopulmonary bypass.

Statistical analyses

Categorical variables, presented as frequencies and percentages, were compared using the χ2 test or Fisher’s exact test. Continuous variables, expressed as the mean±SD or the median with the range, were compared using Student’s unpaired t-test (between two groups) or one-way analysis of variance (among three groups) as appropriate. Kaplan-Meier analyses were used to assess conditional probability of survival or freedom from valve-related complications, and log-rank tests were used to compare intergroup differences. This study used the ‘as-treated’ method for categorisation of patients into MV repair, mechanical and bioprosthetic groups.

To determine the predictive factors for MV ’reparability', a stepwise backward multivariable logistic regression analysis was performed. For this, baseline variables with a P value ≤0.20 in univariable logistic regression models were included in the multivariable models, in which a stepwise backward elimination technique was conducted to leave only variables with a P value <0.10 in the final multivariable model.

To reduce the impact of treatment selection bias and potential confounders in the comparisons of outcomes between repair and replacement groups, we adjusted the differences in the patient’s baseline variables with a propensity score (PS).8 The PS was estimated by the logistic regression model incorporating the 33 baseline variables that are listed in online supplementary table 1. For the adjustment of MV pathoanatomic variables, grades of mitral regurgitation and stenosis, MV area, left atrial size, previous mitral valvotomy and pulmonary arterial pressure were used as dummy variables (0 or 1). Discrimination and calibration of the PS were assessed with Hosmer-Lemeshow statistics and C statistics. The PS-matched pairs were created by matching between the repair and the replacement groups on the logit of the PS using callipers of width ≤0.10 of the SD. Balance checking for baseline covariates in the matched cohort was assessed by evaluating the standardised difference (SDi) of mean, in which an SDi of mean <0.1 was regarded as well balanced for each covariate.9

Supplementary file 2

Conditional logistic regression analysis was performed for the evaluation of the early outcomes while fit proportional hazards regression models were used to evaluate the late time-related outcomes (death and valve-related complications) in the PS-matched cohort.

For subgroup analyses for the comparisons between repair and mechanical prosthetic replacement (online supplementary table 2) and between repair and bioprosthetic replacement (online supplementary table 3), two separate PS models were created to include each of the subgroups only, and PS matching procedures as well as outcome analyses were conducted in each of the subgroups.

R statistical software V.3.3.1 was used for statistical analyses. All reported P values were two tailed, and P≤0.05 was considered statistically significant.


Baseline characteristics and predictors of reparability

The mean age of the patients was 52.3±12.5 years and 1190 (68.7%) were female. There were nine patients who initially underwent MV repair and were converted to MV replacement because of repair failure revealed by intraoperative transoesophageal echocardiography. These patients were categorised into the ‘replacement’ group in this study based on the ‘as-treated’ approach. When these nine patients were recategorised into the ‘repair’ group for ‘intention-to-treat’ based analysis, study results regarding all clinical endpoints did not change. The baseline demographic and clinical profiles of patients according to the type of MV are listed in table 2. Patients who underwent repair were younger, biased towards more female gender and more likely to have predominant regurgitation than those in the replacement groups. Patients who underwent bioprosthetic valve replacement were relatively older than those in the other two groups (figure 1).

Figure 1

Distribution of patients according to mitral valve (MV) procedures based on patient age (red=repair; blue=mechanical valve replacement; orange=bioprosthetic valve replacement).

Table 2

Baseline demographics and clinical variables

On multivariable logistic regression analysis, younger age and predominant mitral regurgitation were positively correlated with greater reparability while the history of previous mitral valvotomy and the presence of left atrial thrombi or significant pulmonary hypertension were significantly associated with a lower probability of receiving a valve repair procedure (table 3).

Table 3

Baseline factors to predict repair procedures

Unadjusted clinical outcomes

Early mortality rates were 1.7% (5/294) in the repair group, 4.3% (49/1134) in the mechanical valve group and 6.6% (20/303) in the bioprosthetic valve group (P=0.022). During a mean follow-up of 130.9±27.7 months, overall death occurred in 283 (9.3%/person-years (PY)) and valve-related complications occurred in 37 (2.4%/PY), 173 (4.1%/PY) and 46 (3.6%/PY) in the repair, mechanical valve and bioprosthetic valve groups, respectively (table 4, left set of columns).

Table 4

HRs for adverse clinical outcomes according to surgical procedures: repair versus replacement

Figure 2 depicts Kaplan-Meier plots of cumulative rates of death (A), reoperation (B), valve-related complications (C) and overall adverse events (D). Overall survival rates at 15 years were 57.6%±17.0%, 55.1%±5.5% and 29.2%±7.4% for the repair, mechanical and bioprosthetic valve groups, respectively (P<0.001). Reoperation-free rates at 15 years were 81.9%±5.4%, 96.5%±0.8% and 46.2%±20.0% in the repair, mechanical and bioprosthetic valve groups, respectively (P=0.01).

Figure 2

Unadjusted Kaplan-Meier (KM) plots for cumulative overall mortality (A), reoperation (B) and valve-related complication rates (C) according to the types of surgery. Shaded bands indicate areas within 95% CIs (red=repair; blue=mechanical valve replacement; orange=bioprosthetic valve replacement).

When unadjusted, patients receiving a valve repair procedure showed a significantly lower risk of early mortality, overall mortality and adverse valve-related complications than those undergoing valve replacement surgeries (table 4).

Adjusted clinical outcomes

The PS model showed Hosmer-Lemeshow goodness of fit P value of 0.24 and C statistics of 0.872, indicating reasonable discrimination and calibration profiles of the PS model. By the PS matching, 188 pairs of patients were yielded that are well balanced for most baseline covariates (online supplementary table 1). In the matched cohort, the replacement group comprised 155 (82.4%) patients with mechanical replacement valves and 33 (17.6%) patients with bioprosthetic replacement valves.

Analyses on the PS-matched cohort revealed that there were no significant differences in the early and overall mortality risks between the repair and replacement groups (table 4 and figure 3). The risk of valve-related complications, however, was significantly lower in the repair group principally derived by a lower risk thromboembolic and haemorrhagic events (table 4). There was a trend of high risks for reoperation in the repair group compared with the MV replacement group, but there was no significant statistical difference.

Figure 3

Adjusted Kaplan-Meier plots for cumulative overall mortality (A), reoperation (B) and valve-related complication rates (C) according to the types of surgery. Shaded bands indicate areas within 95% CIs (red=repair; blue=replacement).

Subgroup analyses: ‘Repair vs Mechanical’ and ‘Repair vs Bioprostheses’

The subgroup PS model showed Hosmer-Lemeshow goodness of fit P values of 0.79 and 0.85, and C statistics of 0.876 and 0.887 in subgroup analyses, comparing mechanical valve replacement surgery and bioprosthetic valve replacement surgery, respectively. This indicates that it had reasonable discrimination and calibration profiles (online supplementary figure 1).

Supplementary file 3

When the PS-matched repair and mechanical valve replacement groups (181 pairs) were compared, there were no significant differences in the risks of early mortality (OR, 0.49; 95% CI 0.10 to 1.89; P=0.32) and overall mortality (HR, 0.73; 95% CI 0.35 to 1.54; P=0.41, online supplementary table 2). The risk of valve-related events was significantly lower in the repair group (HR, 0.50; 95% CI 0.28 to 0.89; P=0.02, online supplementary figure 2), and this was specifically derived from the significantly lower risk of anticoagulation-related haemorrhagic events found in the repair group than mechanical replacement group (HR, 0.15; 95% CI 0.04 to 0.66; P=0.01). The result showed a trend of high risk for reoperation in the repair group (HR, 3.30; 95% CI 0.84 to 12.99; P=0.09, online supplementary table 3). Similarly, when the repair and bioprosthetic replacement were compared in another subgroup PS analysis (111 pairs), the risk of overall mortality (HR, 0.32; 95% CI 0.11 to 0.90; P=0.03) and the composite risk of valve-related events (HR, 0.35; 95% CI 0.14 to 0.87; P=0.02) were significantly lower in the repair group than in the bioprosthetic group (online supplementary table 4).

Supplementary file 4

MV function after repair procedures

Up to 251 (85.4%) patients who underwent MV repair procedures were evaluated with echocardiography beyond 6 months, postoperatively. Mean echocardiographic follow-up duration was 71.9±41.0 months, and the values of trans-MV pressure gradients are filled in online supplementary table 5. During this period, 61 (24.3%) patients showed moderate to severe mitral dysfunction either insufficiency (regurgitation grades III and IV; n=51) or stenosis (MV area ≤1.5 cm2; n=11). Among them, 15 (6.0%) patients showed postoperative severe mitral dysfunction (regurgitation grade IV or MV area ≤1.0 cm2). In 61 patients with moderate to severe mitral dysfunction, 27 had New York Heart Association (NYHA) class III and 2 had NYHA class IV. These 29 patients classified as NYHA III/IV were considered for reoperation. Among whom, reoperation was deferred for 10 patients due to the improvement of symptoms with medical treatments. Six of the 10 patients experienced dyspnoea relief after cardioversion, and the symptoms in the other four were alleviated by drainage of intrathoracic fluids. One elderly patient with chronic renal disease declined reoperation despite our recommendation. This patient was still alive at 6 years postsurgery. The remaining 18 (0.86%/PY) patients with symptomatic MV regurgitation underwent reoperation. Of the 294 patients, at 5 and 10 years postoperatively, 86.7%±2.5% and 68.0%±4.3%, respectively, were free from moderate to severe mitral dysfunctions (>mild mitral regurgitations or MV area ≤1.5 cm2); 96.5%±1.4% and 92.9%±2.2%, respectively, were free from severe mitral dysfunctions (severe mitral regurgitations or MV area ≤1.0 cm2; n=15); and 97.9%±9.5% and 92.9%±2.1%, respectively, did not require reoperation (online supplementary figure 3).

Supplementary file 5


Given the associated risks of progression to heart failure, infective endocarditis, stroke and pregnancy-related complications,10 11 advanced rheumatic heart disease remains a major global problem, disabling more than 10 million and killing nearly 275 000 annually, according to a systemic analysis for ‘The World Health Organization (WHO) Global Burden of Disease Study’.12 13 Accordingly, WHO recommends timely interventions.14 In a comprehensive research of rheumatic heart diseases in endemic areas, Zühlke and colleagues showed that lack of proper antibiotic prophylaxes or appropriate timely interventions were strongly associated with unfavourable clinical outcomes.11 While recommendations on antibiotic prophylaxis are well established, there is ongoing debate regarding the best surgical options for the treatment of advanced rheumatic heart diseases. Thus, careful analysis of the impact on overall long-term outcomes according to surgical interventions in advanced rheumatic heart diseases is important.

Unlike degenerative MV insufficiency where valve repair is favoured over replacement surgery with solid scientific evidence, establishing practical recommendations on optimal surgical therapy in rheumatic MV disease has been a conundrum. This is partly attributable to concerns about poor durability of repair techniques in rheumatic diseases.1 2 Although previous studies conducted decades ago had postulated inferior durability after rheumatic MV repair compared with valve replacement,15 16 the repair outcomes from more recent studies have been more favourable than before.17 18 In addition, Yakub et al have shown non-inferior clinical outcomes of rheumatic MV repair procedure compared with degenerative MV repair up to 10 years of follow-up.5 Furthermore, Coutinho et al showed the feasibility of the repair procedure even in cases with prior failed percutaneous mitral valvotomy.19

To the best of our knowledge, the present study included the largest cohorts with long-term follow-up among the studies on outcomes of surgical options in rheumatic MV disease. In addition, comparisons between repair and replacement therapy with rigorous adjustments for potential confounding baseline characteristics have been scarce. In this study, PS matching methodology was used to reduce the baseline confounders. The repair procedure showed a significantly lower rate of composite valve-related complications, in particular a lower risk of anticoagulation-related bleeding events, while survival outcomes were comparable with valve replacement therapy in the PS-matched cohort. Furthermore, the risk of reoperation in the repair group was not significantly higher than that of the replacement group after PS matching (P=0.43). Of note, extensive rheumatic MV repair procedures such as pericardial leaflet extension or calcified MV peelings were limitedly performed in our institution because of the prevailing assumption that far advanced rheumatic pathologies requiring more complex repair procedures are better treated by prosthetic valve replacement surgery, although favourable mid-term outcomes following such extensive rheumatic MV repair have been reported from several centres.20–22 To further validate the results of this study, we conducted subgroup analyses, matching the repair group patients with both the mechanical and bioprosthetic groups. In subgroup analysis comparing repair and bioprosthetic replacement groups, overall long-term clinical outcomes of the repair group were significantly superior to those of the bioprosthetic valve group in terms of mortality as well as valve-related complications. These results, however, could be influenced by the inclusion of the PS matching of younger patients in this subgroup analysis (53.3±12.8 years), for whom bioprosthetic replacement may not be the best option for this age.

Costs of treatment and the management of lifelong anticoagulation therapy are among the major hindrances in the treatment of advanced rheumatic valve disease in low-income countries,11 23 24 where rheumatic disease prevalence is the highest in the world.12 25 The cost of prosthetic valves is known to be a prohibitive burden for many patients in countries with endemic rheumatic heart disease. In this regard, rheumatic valve repair may be an ideal option, especially for those with valvular pathologies favourable for repair.


This study has inherent limitations due to the retrospective and observational nature of the study design. Although we sought to use rigorous statistical adjustment techniques, unmeasured residual biases could still have influenced our results. Since the patients in repair group were selected according to the feasibility of repair, which means they were more favourable conditions at baseline, unwanted selection bias could remain even after PS matching with narrow calibration. There was a statistical difficulty to estimate reparability from highly variable and individualised cardiac anatomic parameters.

Finally, accounting for surgeon factor in statistical models was unavailable given a relatively small sample size in the repair group (n=294) while there were seven attending surgeons involved in this study.


Valve repair in well-selected patients with severe rheumatic MV disease showed comparable survival outcomes and a trend towards more favourable valve-related outcomes than valve replacement surgery. Following repair, the rates of moderate to severe MV dysfunction and reoperation were 4.1%/PY and 0.86%/PY, respectively. The risk of haemorrhagic events was significantly lower than MV replacement. Therefore, MV repair may be a reasonable alternative to replacement surgery in well-selected patients.

Key messages

What is already known on this subject?

The mitral valve (MV) repair procedure in degenerative cardiac valvular disease is known for having better clinical outcomes than MV replacement surgery. However, MV repair durability in rheumatic heart disease has been debated, and long-term benefit of the repair procedure over prosthetic replacement has not been determined.

What might this study add?

Valve repair in well-selected patients with severe rheumatic MV disease showed comparable survival outcomes (HR, 1.24; 95% CI 0.62 to 2.48; P=0.54) and a trend towards more favourable valve-related outcomes (HR, 0.57; 95% CI 0.33 to 0.99; P=0.045) compared with valve replacement surgery.

How might this impact on clinical practice?

The MV repair procedure for rheumatic heart disease may be a reasonable alternative to replacement surgery in well-selected young patients.



  • Contributors JBK conceived the present study. JBK, SHJ and WKK performed statistical analyses. JBK and JWL handled funding and supervision. WKK, SJC and CHC acquired the data of present study. WKK drafted the manuscript. JBK made critical revision of the manuscript for key intellectual content.

  • Funding This study was supported by a grant from the Asan Institute for Life Sciences, Asan Medical Center, Seoul, Korea.

  • Competing interests None declared.

  • Ethics approval The study protocol was approved by the Institutional Review Board of the Asan Medical Center.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles