A systematic database search was performed in MEDLINE and Embase for studies published from January 2010 to December 2020 reporting on TEER or SMVr for treatment of FMR. Due to limited data regarding use of the novel PASCAL^® system in FMR patients and its early implementation, this investigation focused on the well-established MC. MeSH terms included “mitral valve insufficiency,” “ventricular dysfunction,” “cardiac surgical procedures,” and “secondary mitral regurgitation” identifying 1,703 studies. Further details about the search strategy can be found in Supplementary Figure 1.

We included studies reporting in-hospital, 30-day or 1-year death figures for MC or SMVr. Studies reporting data of patients with DMR and FMR were only included if event numbers were quoted separately for each etiology. The flow chart of literature search is presented in Figure 1.

All included studies used MR grading 1–4 or mild, moderate, moderate-to severe and severe, which were assessed by contributed echocardiographers according to current European or American guidelines, vena contracta, jet size or eyeballing. Few studies did not explicitly report grading criteria.

Two investigators (DF and RF) independently reviewed all articles, selected eligible studies and collected data of interest. Baseline characteristics such as age, gender, LVEF, EuroSCORE, and clinical outcomes (in-hospital, 30-day, 1-year mortality and reoperation or reintervention) were extracted from each study, if available. A third investigator (LS) reviewed differences between the collected data.

Available baseline characteristics are displayed as weighted percentage means with standard deviation for dichotomous variables and weighted means with standard deviation for continuous variables. Categorial variables were treated as quasi-continuous by percentages ranging between 0 and 1 considering a different size of cohorts and studies reporting each variable and treatment. The number of studies reporting the analyzed variable is displayed in Table 1. Mean with standard deviation was calculated using Microsoft Excel (Version 16) for each cohort and compared with the unpaired t-test using the t-test Calculator by GraphPad online.

Since only few studies were available that enabled a direct comparison of TEER and SMVr, a classical meta-analysis was not feasible. In order to combine results from all identified studies regarding in-hospital, 30-day and 1-year mortality as well as reoperation and reintervention rates, New York Heart Association (NYHA) class and MR grade reduction, a meta-analytic approach was used. In particular, the reported overall proportions from the included single-arm studies were combined using the inverse variance method, which is available in the R package “meta”. Study heterogeneity was assessed using the I² measure leading to a fixed effects combination model in case of I² < 50 and to a random effects combination model otherwise. Furthermore, meta-regression (R package “metafor”) was applied in order to account for possible confounding of the results by different patient characteristics. For a statistical comparison of the overall proportions calculated for each procedure separately, the 95% confidence intervals (CI) were used, where non-overlapping CIs indicated statistical significance (p < 0.05) (18).

The study quality was assessed as described in detail by the National Institutes of Health Quality Assessment Tool. Studies were rated as being of either “good”, “fair” or “poor” quality (19). Statistical supervision was provided by the Institute of Epidemiology and Medical Biometry of Ulm University.

All analyses were based on previous studies, therefore neither patient consent nor ethical committee approval was required for this analysis. The investigation conforms with the principles outlined in the Declaration of Helsinki.

The included surgical studies mainly used annuloplasty for MV repair (20–22). Several studies additionally reported chordal cutting (23), chordal replacement (21), papillary muscle relocation (20, 24) or left ventricular remodeling (24, 25). Only one study explicitly reported edge-to-edge repair in combination with annuloplasty (26). Regarding TEER, all included studies used first and second generation MC exclusively.

Patients in the TEER cohort were significantly older (71.9 ± 1.7 vs. 64.7 ± 4.7 years, p < 0.001) and presented significantly more often with hypertension (74.5 ± 9.8 vs. 58.9 ± 12.8%, p < 0.001), atrial fibrillation (51.6 ± 13.4 vs. 25.9 ± 14.8%, p < 0.001), history of coronary artery bypass graft (37.9 ± 10.7 vs. 9.1 ± 8.2%, p < 0.001), chronic obstructive pulmonary disease or lung disorder (23.2 ± 5.5 vs. 12.0 ± 5.9, p < 0.001), previous percutaneous coronary intervention (40.7 ± 12.1 vs. 28.8 ± 16.9%, p = 0.048), and chronic renal disease (33.8 ± 14.3 vs. 16.8 ± 11.1%; p = 0.005). Moreover, patients in the TEER cohort showed poorer LVEF (30.9 ± 5.7 vs. 36.6 ± 5.3%, p < 0.001; see Figure 2). Accordingly, logistic EuroSCORE was significantly higher in the TEER cohort (22.4 ± 4.4 vs. 9.1 ± 3.2, p < 0.001; see Table 1). No significant differences were observed regarding the rate of prior myocardial infarction (p = 0.29) as well as NYHA class (p = 0.24), diabetes (p = 0.06), severity of tricuspid regurgitation ≥ grade 2 (p = 0.31), and MR grade (p = 0.70; see also Table 1).

A total of 33 of the 56 included studies (21 SMVr, 12 TEER) reported in-hospital mortality, which was significantly lower in the TEER compared to the SMVr cohort [3% (95%-CI 0.02–0.03, I² = 0%) vs. 5% (95%-CI 0.04–0.07, I² = 45%); see also Figure 3]. 30-day mortality was reported in 35 of the included studies (15 TEER, 20 SMVr) and comparable between both groups [4% in TEER (95%-CI 0.03–0.05, I² = 45%) vs. 4% in SMVr (95%-CI 0.03–0.06, I² = 49%); see also Figure 4].

A total of 8 TEER and 10 SMVr studies explicitly reported 30-day cardiac death, which was comparable between both treatment strategies [3% (95%-CI 0.02–0.05, I² = 60%) vs. 4% (95%-CI 0.02–0.06, I² = 0%); see also Figure 5]. Regarding in-hospital and long-term mortality cardiac death was not analyzed due to limited data.

A total of 9 TEER and 6 SMVr studies containing 1,994 and 509 patients separately reported on 1-year mortality (Figure 6). 1-year mortality was 18% in the TEER and 11% in the SMVr cohort [(95%-CI 0.15–0.21, I² = 67%) vs. (0.07–0.18, I² = 70%)]. As the 95%-CI of the TEER and SMVr groups did overlap by a small margin, a tendency toward lower 1-year mortality in SMVr was observed, however, this finding lacks statistical significance (see also Figure 2).

Within a mean follow-up period of 13.3 ± 4.4 months in TEER and 20.2 ± 22.9 months in SMVr, NYHA class reduction was comparable between both groups [1.06 (95%-CI 0.87–1.26, I² = 98%) vs. 1.15 (0.74–1.56, I² = 98%); see also Figure 7]. Likewise, there was no difference in MR grade reduction within a mean follow-up period of 13.9 ± 5.1 months in TEER and 26.9 ± 23.1 months in SMVr [1.74 (95%-CI 1.52–1.97, I² = 99%) vs. 2.08 (1.57–2.59, I² = 99%); see also Figures 2, 8].

The reported necessity of reoperation or reintervention within a mean follow-up period of 18.1 ± 14.2 months in TEER and 55.7 ± 28.3 months in SMVr did not differ significantly and was comparably low in both cohorts [3% (95%-CI 0.02–0.05, I² = 68%) vs. 3% (0.01–0.05, I² = 76%); see also Supplementary Figure 2].

Meta-regression analysis of 30-day mortality was performed for all potential variables if at least 10 studies individually reported on the variable of interest according to the Cochrane Handbook for Systematic Reviews of Interventions (27).

Baseline NYHA class remained a significant correlate of 30-day mortality. In the TEER cohort, NYHA IV (estimate: 3.1, 95%-CI 1.33–4.93, p < 0.001) and in the SMVr cohort, NYHA III + IV (estimate: 2.4, 95%-CI 0.82–4.04; p = 0.003) presented as a significant moderator of 30-day mortality (Table 2).

One limitation of the meta-analytic approach is the comparison of patients included in studies, which did not primarily compare SMVr and TEER. Consequently, the presented estimators are not adjusted and risk-of-bias-assessment was not feasible due to the use of single-arm studies. However, this can also be considered a strength resulting in large datasets of patients undergoing TEER or SMVr. Additionally, adjustment for baseline characteristics might presumably be in favor of TEER. Since some variables are reported by a limited study number, meta-regression was not performed for all variables. Potential moderating influence of variables not examined by meta-regression analysis cannot be ruled out. Our studies’ endpoint is restricted to 1 year because many studies report different long-term follow-up periods impeding interpretation. Likewise, follow-up periods of the reported post-interventional NYHA and MR grades substantially differed between TEER and SMVr, which might affect comparability. NYHA and MR grade reduction were treated as numeric variables because many studies reported mean with standard deviation only, which necessitated a general comparison of means (22, 23, 40). Eventually, the fact that only studies reporting on first and second generation MC devices were included in this investigation limits its value. However, latest results of the novel PASCAL^® system as well as recent MC device generations would presumably emphasize the benefits of TEER.

Evidence regarding favorable treatment of FMR using TEER is constantly growing, whereas surgical MV repair and replacement have repeatedly shown moderate outcomes in this high-risk population. However, there still is an ongoing debate about the optimal treatment strategy due to a lacking direct comparison. With this meta-analytic approach, we were able to show similar mid- and long-term prognostic and symptomatic outcomes of TEER compared to SMVr in >8,000 FMR patients, despite an unfavorable baseline risk-profile of TEER patients. Results of a direct head-to-head comparison of both therapies are eagerly awaited upon completion of the randomized controlled MATTERHORN trial.