3.1 Search results
All 4834 studies were reviewed, and only 20 studies enrolling a total of 21,898 patients with NIMR were included in the final analysis (Table S1 & Fig. 1) [1, 9–25]. Four studies was retrospective study [10–12, 20], 2 studies are case-control study [14, 19], where other studies were prospective cohort study. Two studies exclusively enrolled hypertrophic obstructive cardiomyopathy patients with MR [12, 13] and 2 studies exclusively included Marfan syndrome cases with MR [14, 15], whereas 4 studies excluded ischemic MR [9–11, 22]. The other studies enrolled those with degenerative MR. The mean follow-up duration was from 30 days  to more than 10 years . When available, prevalence of risk factors of surgery is given in Table S2 separately for each study. These risk factors were: heart failure (New York Heart Association (NYHA) class III or IV, low pre-operative LVEF), prior history of hypertension, diabetes, atrial fibrillation.
3.2 Quality Assessment
The quality assessment of 20 studies is shown in Table S3. According to the Newcastle-Ottawa Scale to assess the risk of bias in these studies, 17 studies scored between 6 and 9, indicating high methodologic quality.
3.3 Patient profiles
Age Of total 20 articles, 15 studies provided exact values for mean age of two groups. The average age in most of the studies was between 50 and 78 years (Table S1). There was a study reported using younger subjects ; while another study used very elderly subjects . A pooled analysis demonstrated that the subjects of MVr group is slightly younger than MVR group (1.92; [1.74, 2.11]; p = < 0.0001; Supplemental Figure S1).
Woman Of total 20 articles, 14 studies provided exact numbers of women for two groups. Most of the studies had a preponderance of male patients. Rate of women was significantly higher than men in only 2 studies (Table S2) [13, 26]. Analysis demonstrated that rate of women was lower in the MVr group than MVR group (1.37; [1.28, 1.47]; p < 0.0001; Supplemental Figure S2).
Pre-operative Cardiac function Of total 20 articles, only 6 studies provided precise values of pre-operative LVEF for two groups, 11 studies supplied the numbers of NIMR patients with NYHA functional class (≥ III), and 3 studies offered the numbers of NIMR patients with heart failure. Our results proved that there was no difference in number of pre-operative LVEF (0.03, [-0.16, 0.23]; p = 0.01, Supplemental Figure S3A), or heart failure (1.02, [0.77, 1.35]; p = 0.87, Supplemental Figure S3C) between two group, but the subjects of NYHA functional class (≥ III) in MVr group is slightly less than MVR group (1.19, [1.01, 1.41]; p = 0.0009, Supplemental Figure S3B)
Risk factors Of total 20 articles, only 5 studies provided exact numbers of diabetes patients for two groups, 6 studies supplied the numbers of NIMR patients with hypertension, and 11 studies offered the numbers of MR subjects with atrial fibrillation (AF). It was showed that the rates of diabetes (3.04, [2.03, 4.54]; p = 0.003, Supplemental Figure S4A), hypertension (1.45, [1.33, 1.58]; p < 0.00001, Supplemental Figure S4B) and AF (1.31, [1.22, 1.41]; p = 0.50, Supplemental Figure S4C) were lower in the MVr group than MVR group.
Early mortality Of total 20 articles, OR for early mortality was available in 12 articles. These studies were not precisely consistent in their definition of early mortality. For the purpose of the current analysis, early outcomes labeled ‘early mortality’ (n = 1), ‘hospital mortality’ (n = 2), ‘operative mortality’ (n = 5), and ‘30-day mortality’ (n = 4) were combined under the ‘early mortality’ label. There was no early mortality among the patients in 1 studies . An unadjusted OR for early mortality was available in 11 studies, only two of which demonstrated a significantly lower early-mortality rate in the MVr group than MVR group [11, 22]. The pooled analysis with these studies showed that summary OR is 2.72 ([2.28, 3.24]; p < 0.00001), indicating that the MVR group has significantly increased risk of early dying compared to the MVr group (Fig. 2). Studies were subgroup according to the etiology of MR, and 7 of these studies are about degenerative MR. The heterogeneity test was not significant (p = 0.09, I2 < 50%) for degenerative subgroup, as well as for all of the included studies, suggesting that there was no clear evidence of a major discrepancy among the OR for the studies analyzed.
Late survival Of total 20 articles, HR for late mortality was available in 10 articles, 4 of which offered the adjusted HR [1, 17, 23, 27]. HR for 20 years survival was offered in two studies [1, 17], one study provided 6 years survival data , one study provided 5 years survival data , while HR for about ten years survival was available in five articles [12, 14, 18, 23, 25, 27]. A pooled analysis indicated statistically much higher late-mortality rate in MVR group than MVr group (1.81, [1.59, 2.07]; p = 0.56, Fig. 3). Studies were subgroup according to the etiology of MR, and 7 of these studies are about degenerative MR. The heterogeneity test was not significant (p > 0.10, I2 < 50%) for degenerative subgroup, as well as for all of the included studies, suggesting that there was no clear evidence of a major discrepancy among the hazard ratios for the studies analyzed.
Reoperation Out of the articles reviewed, 9 articles provided information to allow determination of the re-operative OR for mitral valve replacement relative to repair. The results demonstrated that the risk of reoperation is higher in MVR group than in MVr group (1.59; [1.36, 1.86]; p = 0.22; Fig. 4).
Complications Major complications of MV surgery in reviewed studies are post-operative MR (including residual or recurrent MR), thromboembolism, heart failure, infective endocarditis. From all of the reviewed articles, information for determining the hazard ratio for development of these complications is insufficient.
The severe mitral regurgitation after operation were analyzed. Out of the articles reviewed, only four articles provided exact numbers of post-operative MR in two groups. Results demonstrated that the incidence rate of post-operative MVr group is lower than MVR group (1.43; [1.13, 1.82]; p = 0.01; Figure S5).
Besides, other complications were studied in only two articles [22, 24]. Both of them showed beneficial effects from repair compared to replacement for thromboembolism and heart failure. One study  demonstrated that more patients had infective endocarditis in the MVR group than MVr group, while another study  demonstrated that the incidence rate of infective endocarditis in two group is comparable.
3.5 Publication bias
Inspection of the funnel plot (Fig. 5) did not show significant asymmetry for early and late mortality. These results suggest that publication bias was not a significant influencing factor.