The study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (25). On January 14, 2022, a comprehensive literature search was conducted using the PubMed, Embase, and Cochrane Library databases for relevant studies published since 2000 that reported the outcomes of reoperation for failed MVr. The following words found in the title or abstract were used to perform the search: “mitral valve” and “repair” and “recurrent^*” or “reoperation.”

Studies that met the following criteria were included: (1) the population consisted of patients who underwent repeat mitral valve surgery (re-repair or MVR) because of failed previous MVr; (2) there were data on perioperative mortality and long-term survival outcomes; (3) studies that reported the outcomes of the re-repair and MVR groups.

The exclusion criteria were as follows: (1) primary surgery including MVR, (2) reoperation due to causes other than repair failure, (3) transcatheter procedures, (4) second cross-clamp during the same surgery, and (5) studies with ≤ 10 patients included, duplicate publications, and review of articles.

Three authors (ZZ, HX, and WS) screened and assessed the studies independently for inclusion. Disagreements regarding the inclusion of articles were resolved via group consensus.

Two authors (ZZ and HX) reviewed and extracted the reported data from the studies, including details of the study (study design, inclusion criteria, study period, and follow-up duration), baseline demographics, preoperative echocardiographic parameters, causes of repair failure (technique or valve-related), surgical details (re-repair/MVR, and cross-clamp time), perioperative morbidities and mortality, and long-term outcomes (MR recurrence, reoperation, and mortality). For long-term outcomes, the number of observed events and the Kaplan-Meier estimation were extracted and analyzed.

Study quality and risk of bias were assessed using the Newcastle-Ottawa Scale. Disagreements were resolved by consensus.

The analyses were performed utilizing Review Manager 5.4 (Cochrane Collaboration, Oxford, United Kingdom). Statistical heterogeneity was assessed using I² for low (25~49%), moderate (50~ 74%), and high (≥ 75%) heterogeneity. When I² ≥ 50%, random-effects models were utilized.

For the single-arm meta-analysis (re-repair rate and incidence of technical failure), generic inverse variance methods were used. For binary data (incidence of technical failure), the odds ratio (OR) was calculated using the Mantel-Haenszel method. For continuous data (aortic cross-clamp time and hospital stay time), the mean difference (MD) was calculated using the inverse variance method. For binary data (perioperative and follow-up events), the relative risk (RR) was calculated using the Mantel-Haenszel method. For time-to-event data (the long-term outcomes), the hazard ratio (HR) was calculated using the logarithmic scale generic inverse variance method (26). When HR was not provided, it was calculated using the Kaplan-Meier survival P value and other details if possible (26, 27). When the mean and variance were not provided, they were estimated from the median and quartile range (28–30). Forest plots were generated from the present pooled results.

A total of 2,365 studies were identified using the search criteria. Based on the title and abstract, 17 studies were retrieved for full-text review. Follow-up outcomes comparing re-repair and MVR were not reported in six studies. Two studies by Zegdi et al. used patients from the same institute (AP-HP, Assistance Publique-Hôpitaux de Paris) for two different periods: 1991–2004 (31) and 1987–2006 (14), and the first one was excluded. One study by Ma et al. reported a second cross-clamp under reinstituted cardiopulmonary bypass for incomplete initial repair but not repair failure (32). Kwedar et al. included reoperation after both previous MVr and MVR and did not report outcomes after initial repair specifically (33). The remaining eight studies comprised pooled data (Figure 1).

The eight studies were all unmatched retrospective cohort ones. Four studies included only degenerative mitral regurgitation, while the other four studies did not specify the pathological types. A total of 938 patients were included, of whom 295 underwent re-repair and 643 underwent MVR. Six studies compared the preoperative demographic data between the re-repair and MVR groups. All the studies reported follow-up results; mean/median follow-up ranged from 1.8 to 8.9 years. Seven studies reported the Kaplan-Meier survival estimation of the re-repair and MVR groups, while one only reported the follow-up events. The basic characteristics of the studies are listed in Table 1, and the demographic data are shown in Table 2. The Newcastle-Ottawa Scale scores ranged from 7 to 9 out of 9 for the studies. Some points were lost because of a lack of comparability between the groups. The quality assessment of the included studies is listed in Supplementary Table 1.

Failure of MVr can be categorized into technical failure (procedure-related) or progression of disease (valve-related) for recurrent regurgitation (34). Causes of mitral stenosis after MVr included organic and functional (“tunnel effect” due to a small ring) (19).

The proportion of technical failure was reported in six studies (12–17) and showed high heterogeneity (I² = 86%). One study (18) only included reoperation due to ring dehiscence (technical failure). The pooled incidence of technical failure was 41% (RD: 0.41; 95% CI:0.32–0.5; P = 0; I² = 86%; 6 studies, 846 patients).

The incidence of technical failure in the re-repair and MVR groups was reported in four studies (12, 15–17). The incidence of technical failure was higher in the re-repair group, but the difference was not significant (OR: 2.66; 95% CI: 1.02–6.93; P =0.05; I² = 82%; 4 studies, 615 patients) (Figure 2).

The pooled mitral valve re-repair rate was 36% (RD 0.36; 95% CI: 0.26–0.46; P = 0; I² = 91%; 8 studies; 938 patients).

Mean aortic cross-clamp time was reported by three studies (15–17) and was shorter in the re-repair group, but the difference was not significant (MD = −11.5 min; 95% CI: −23.03–0.02; P =0.05; I² = 82%; 3 studies, 470 patients).

Perioperative mortality was reported in all the studies, including two studies (17, 19) without perioperative death. The pooled data of the other six studies (12–16, 18) showed significantly lower perioperative mortality in the re-repair group than in the MVR group (RR: 0.22; 95% CI:0.07–0.66; I² = 0%; P = 0.008; 6 studies, 824 patients).

Major perioperative morbidities were reported in two studies (16, 17). There was a significant difference in the risk of acute kidney injury (AKI) (RR: 0.18; 95% CI:0.05–0.67; I² = 0%; P = 0.01; 2 studies, 384 patients, but not in stroke (RR: 0.37; 95% CI:0.06–2.34; I² = 0%; P = 0.29; 2 studies, 384 patients) and reoperation for bleeding (RR: 0.33; 95% CI:0.1–1.09; I² = 0%; P = 0.07; 2 studies, 384 patients). There was no significant difference in length of hospital stay (MD = −0.96 d; 95% CI: −4.43–2.5; P = 0.59; I² = 68%; 2 studies, 384 patients) (Table 3; Figure 3). Low cardiac output syndrome, blood transfusion, mediastinitis, and other morbidities were not pooled because of heterogeneity in the diagnostic criteria.

Follow-up reoperation was reported in six studies (12–16, 19) and was significantly higher in the re-repair group (RR: 2.11; 95% CI: 1.03–4.28; I² = 0%; P = 0.04; 6 studies, 802 patients).

Follow-up death was reported in six studies (13–15, 17–19) and was significantly lower in the re-repair group (RR: 0.23; 95% CI: 0.13–0.4; I² = 0%; P = 0; 6 studies, 488 patients). In addition, re-repair was associated with significantly lower long-term mortality when HR was pooled for analysis (HR: 0.42; 95% CI: 0.3–0.58; I² = 0%; P = 0; 7 studies, 903 patients). Finally, increase in long-term survival was observed in studies that only reported degenerative diseases (12–15) (HR 0.34; 95% CI: 0.23–0.51; I² = 12%; P = 0.00; 4 studies, 462 patients). Pooled data from studies reporting mixed pathological types (16–18) showed lower long-term mortality in the re-repair group but without significant difference (HR: 0.62; 95% CI: 0.36–1.07; I² = 0%; P = 0.09; 3 studies, 441 patients) (Table 4; Figure 4).

The pooled data in our meta-analysis suggested that mitral valve re-repair was associated with significantly lower perioperative mortality (P = 0.008). For primary surgery, perioperative mortality was significantly lower for MVr than for MVR in both unmatched and matched populations (P < 0.001) (4). This was similar to that of reoperative mitral valve surgery. Kwedar et al. analyzed data from Medicare and found that hospital mortality was 9.8% for re-repair, 12.7% for MVR with bioprosthesis, and 12.2% for MVR with mechanical prosthesis (33). Re-repair has lower perioperative morbidity (16, 17). Relatively simple techniques and preservation of the subvalvular apparatus might contribute to the improved short-term outcomes.

Mitral valve re-repair significantly improved late survival in our meta-analysis (RR and HR, P = 0) despite the higher risk of a third reoperation (P = 0.04). Even in institutes with a “very low threshold for re-repair” and high re-repair rate, re-repair has excellent survival and durability (24). A previous meta-analysis found similar long-term survival between re-repair and MVR, but a second cross-clamp under reinstituted cardiopulmonary bypass was included in the study (23). In a subgroup analysis of studies that did not specify the pathology, the long-term benefit of re-repair was not significant (P = 0.09). However, Aphra et al. reported that patients with rheumatic mitral valve disease showed excellent survival after re-repair, although there was a certain risk of failure over time (35). In addition, patients who underwent re-repair had significantly better ejection fractions and smaller LV end-systolic dimensions during follow-up (12).

Thus, mitral valve re-repair was associated with better immediate and sustained outcomes and should be recommended if technically feasible.

The pooled mitral valve re-repair rate of the studies was 36% (95% CI 0.26–0.46), ranging from 16 to 49%. Anyanwu et al. reported a systemic strategy to re-repair in all cases, and the re-repair rate was 85% (90% for degenerative disease) (24). Even mitral stenosis after MVr could be re-repaired in an experienced mitral valve reference center (19). Several factors can influence the decision to re-repair. When analyzing the mode of repair failure and valve structure, the surgeon should keep in mind that both early and long-term survival would be improved if durable re-repair could be achieved.

In studies that reported re-repair procedures in detail (15, 17, 24, 35), the most frequently used techniques include leaflet resection, ring removal/annuloplasty, edge-to-edge repair, and neochordae. Simple techniques such as edge-to-edge repair may be an effective choice (36). The technique should be selected based on individual mitral valve lesions, while the basic principles and steps of MVr outlined by Carpentier should always be adhered to.

It is not clear what kind of repair failure is suitable for re-repair with long-term durability. Our meta-analysis suggested that the percentage of technical failure might be higher in the re-repair group. This suggests that re-repair should be considered if there is a definite lesion in the previously repaired location.

Percutaneous transcatheter procedures have been increasingly used for failed MVr in selected patients, including mitral clips (20), neochords (21), and valve-in-rings (22). However, the current literature is limited by the small number of patients and short follow-up period. There is a paucity of data on the long-term outcomes of these interventional procedures. Our meta-analysis suggested that transcatheter MVr might be chosen over transcatheter mitral valve replacement.

Our meta-analysis has several limitations. First, there were only limited studies comparing the long-term survival of re-repair and MVR. The included studies were retrospective and observational, with relatively small sample sizes. Second, there was significant heterogeneity among the included studies regarding the patients' baseline characteristics and outcomes. Third, there was no uniform standard for selecting re-repair or MVR, and the surgical approach was selected based on the experience of the surgeon or institution. Fourth, most of the studies included in the meta-analysis focused on reintervention and survival, and few studies reported on the follow-up functional status and echocardiographic evaluation. Finally, the studies included mainly patients with degenerative mitral regurgitation. The optimal treatment for rheumatic and infective disease has yet to be studied.