中文摘要

第一部分 退行性二尖瓣返流行二尖瓣成形术后不良预后事件的危险因素探索

目的：本研究旨在评估对本中心退行性二尖瓣返流（DMR）患者接受二尖瓣成形术（MVr）后的长期随访结果，总结发生主要不良心血管事件（MACE）及中度以上二尖瓣返流（MR）的情况，探索MVr不良预后事件的危险因素，并比较不同亚组患者的远期预后。

方法：回顾性连续纳入本中心自2017年11月1日至2023年6月30日接受MVr的DMR患者共1354例。生存数据的分析通过 Kaplan-Meier 方法绘制生存曲线，并利用 Log-rank 检验评估生存时间的差异性，远期预后的影响因素采用 Cox 比例风险回归模型进行评估。

结果：本研究中整体患者的年龄中位数为54岁（IQR 45-62），其中女性患者410名（30%），MACE组患者的身体质量指数（BMI）较高、术前多合并房颤且左室舒张末径和左房前后径较大（54.7 ± 9.6 mm vs. 57.5 ± 10.4 mm，55 ± 29 mm vs. 73 ± 75 mm，P < 0.05）。MACE组中有12名患者（27.9%）的左室射血分数（LVEF）低于60%。中位随访时间为40.2个月。从生存分析结果提示，合并术前房颤的患者较无房颤的患者远期更容易发生MACE事件（P = 0.0038），左心室射血分数小于 60% 的患者较LVEF ≥ 60% 的患者发生远期MACE事件的风险更高（P = 0.0017），前叶病变组患者较无前叶病变组患者远期更易发生MACE事件（P = 0.0042）。有前叶病变的患者术后中度以上MR复发的风险随时间推移而升高（P = 0.022）。简单组、中等组、复杂组三组患者的远期免于中度以上MR生存情况无显著差异。Cox多因素回归预测模型分析中研究结果显示术前心房颤动、LVEF < 60%以及前叶病变是MACE发生的重要独立危险因素，在调整了年龄、性别、BMI及手术难度等混杂因素后，术前心房颤动的HR为2.20（95%置信区间: 1.17-4.14，P = 0.014），LVEF < 60%的HR为2.76（95%置信区间: 1.41-5.40，P = 0.003），前叶病变的风险比为2.24（95%置信区间: 1.25-4.01，P = 0.007）。

结论：MVr是治疗DMR患者的有效手段，在有经验的医学中心可达到满意的中远期预后。术前房颤、前叶脱垂以及LVEF < 60% 是中度以上MR复发的独立危险因素。

关键词：退行性二尖瓣返流，二尖瓣成形术，远期预后，MACE，危险因素

第二部分 探讨退行性二尖瓣病变患者成形术失败原因分析及再次修复策略的选择

目的：二尖瓣成形术（MVr）伴随着成型失败和二尖瓣返流（MR）复发的风险，针对二次瓣膜修复的术中超声、瓣膜病变及初次手术技术分析可为减少二尖瓣成形失败提供宝贵的经验。本研究旨在通过回顾本中心进行的二尖瓣二次修复手术患者的围术期与远期随访结果，为减少MVr失败以及二次修复策略的选择提供经验和依据。

方法：回顾性连续纳入本中心自2009年1月至2022年1月于进行二次修复手术的114名退行性二尖瓣返流（DMR）患者，分析患者的基线信息、手术资料和随访数据，生存数据的分析通过 Kaplan-Meier 方法绘制生存曲线，远期预后的影响因素采用 Cox 比例风险回归模型进行评估。

结果：再手术时患者平均年龄49岁，女性24例（21%）。二次修复时病变累瓣叶部位包括前叶病变21例（18%），后叶病变25例（22%），双叶病变68例（60%）。二次手术前MVr组患者的左室射血分数（LVEF）高于MVR组（64 ± 5 % vs. 59 ± 14%），差异有统计学意义（P = 0.018）。二次手术指征为二尖瓣中度及以上反流的患者81例（71.0%），其中4例患者合并中度二尖瓣狭窄（MS）；手术指征为中度及以上MS的患者21例（18.4%），其中重度MS患者15例；其他手术指征还包括溶血15例（13.2%）、感染性心内膜炎4例（3.5%）、瓣叶穿孔2例（1.8%）、腱索断裂1例（0.9%）。本研究中61例（54.0%）与技术相关，48例（41.7%）与瓣膜相关，5例（4.4%）患者修复失败原因不明。技术原因组患者的中位时间间隔为3.4（1.0 - 9.1）年，瓣膜原因组患者的中位时间间隔为8.6（5.2 - 13.5）年，两组患者的二次修复手术的时间间隔有统计学差异（P = 0.002）。在评估术后死亡及二尖瓣返流复发时，二次修复时MVr和MVR两组患者的整体预后情况无统计学差异。

结论：技术原因是导致退行性二尖瓣病变患者行MVr后二尖瓣返流复发的主要原因。在细致评估患者瓣膜质量和瓣环形态、分析病因、确定二尖瓣成形技术的可行性及其二次成形后瓣膜的耐久性的情况下，可再次选择MVr。

关键词：二次修复，成形失败，预后，成形策略，病因

第三部分 经导管二尖瓣修复手术治疗二尖瓣成形术后失败：系统回顾与Meta分析

目的：传统外科二次修复与介入治疗的安全性和预后尚不清楚，本研究通过系统性文献回顾和荟萃分析，旨在评估经导管二尖瓣修复（TMVr）在初次二尖瓣成形（MVr）失败术后需二次修复的患者中的安全性及预后。

方法：检索 Pubmed、Embase 和 Cochrane Library 数据库中有关经初次MVr失败后接受TMVr治疗结果的研究。经过文献检索、文献筛选、数据提取、质量评价后纳入8项研究，采用I²检验来评估统计异质性，生成森林图以展示Meta分析结果。统计分析采用 R 及R包进行，P < 0.05 认为具有统计学意义。

结果：本研究最终纳入8项研究共计212例患者，其中197名患者接受了MitraClip手术，15名患者接受了NeoChord手术，术后即刻患者MR程度降低≥1级的汇总比例为96%（95% CI：86%~100%；I² = 38%；7项研究，149例患者），术后即刻患者残余MR少于轻度的汇总比例为76%（95% CI：67%~84%；I² = 0；7项研究，199例患者），术后即刻患者残余MR少于中度的汇总比例为91%（95% CI：84%~97%；I² = 34%；8项研究，206例患者），5%的患者术后即刻出现二尖瓣重度狭窄（95% CI：84%~97%；I² = 34%；8项研究，206例患者）。所有研究均报告了患者围手术期死亡率，汇总死亡率为 0%（95% CI：0%~1%；I² = 0%；8项研究，212例患者）。所有研究均报告术后近期随访结果，平均随访时间为1个月至15.9个月，68%的患者复发MR ≤ 轻度（95% CI：52%~82%；I² = 57%；6项研究，147例患者），90%的患者复发MR ≤ 中度（95% CI：78%~98%；I² = 58%；6项研究，147例患者）；汇总生存率为 94%（95% CI：88%~98%；I² = 0%；7项研究，196名患者），83%的患者（95% CI：75%~89%；I² = 47%；6 项研究，148 名患者）在随访期间的 NYHA 分级为I级或II级。

结论：TMVr 用于MVr的二次修复在评估近期疗效的情况下是安全且有效的，适合于二次成形难度大或再次手术风险较高的患者，该技术创伤小、可有效减轻患者的MR程度并改善患者心功能状态。

关键词：经导管缘对缘修复，二尖瓣成形术，二次修复，退行性二尖瓣返流，介入治疗

第四部分 基于机器学习的退行性二尖瓣返流患者行二尖瓣成形术的不良预后预测模型

目的：本研究分析了因退行性二尖瓣返流（DMR）接受二尖瓣成形术（MVr）的患者的经食道超声（TEE）数据，探讨MVr失败的发生率和相关危险因素；基于前瞻性队列研究详细收集患者基线资料、手术方式和二尖瓣瓣叶、瓣环和瓣下结构的解剖学参数，建立一个多种机器学习算法的MVr难度评估模型，旨在预测术前手术失败风险。

方法：基于ClinicalTrial.gov注册的前瞻性TEEMR队列研究（NCT05595226）数据库，连续纳入自2021年3月至2023年3月期间在阜外医院接受MVr手术且符合纳入和排除标准的所有成年DMR患者。最终纳入分析的患者队列包括159例男性患者和72例女性患者，中位年龄为 56（50, 64）岁，按照手术时间和8：2的比例将该队列人群拆分为训练队列及验证队列，通过数据预处理、特征筛选、模型构建、模型验证来构建预测模型。

结果：本研究开发了一种基于机器学习的系统以精准评估MVr的复杂程度。在比较了14种不同的机器学习模型后，本研究发现线性支持向量分类（LSVC）模型在预测MVr复杂程度方面显示出了最佳的性能，该模型在训练集与验证集中的表现最为突出，其AUC值在训练集中为0.942（95% CI: 0.920 - 0.982），而在验证集中达到了0.819（95% CI: 0.684 - 0.902），优于其他模型。该模型F1为0.882，召回率为0.790，精确率为1.000，准确率为0.981。根据临床意义以及变量的特征系数权重，最终LSVC模型中纳入特征系数最高的14个变量，包含年龄、瓣环参数（二尖瓣环收缩末期前后径、MA DLR、MA SAP）、瓣叶参数（A2区脱垂、A2长度、P2长度）、病变程度（RW、VCW、反流程度、PASP）、心脏功能（LAV、LVEF、LVEDV）等多维度变量。

结论：LSVC模型可用于预测DMR患者行MVr的复杂程度，年龄、瓣膜病变部位、瓣叶长度、瓣环大小等超声参数可用于预测MVr的失败风险。

关键词：退行性二尖瓣返流，二尖瓣成形术，经食道超声检查，机器学习，预测模型

ABSTRACT

Part I. Risk factors for adverse prognostic events after mitral valve repair in patients with degenerative mitral regurgitation

Objective: This study aims to evaluate the long-term follow-up outcomes of patients with degenerative mitral regurgitation (DMR) undergoing mitral valve repair (MVr) at our center, to summarize the occurrence of major adverse cardiac events (MACE) and moderate or greater mitral regurgitation (MR), to explore the risk factors for adverse prognostic events after MVr, and to compare the long-term prognosis of different patient subgroups.

Methods: A retrospective consecutive inclusion of 1,354 DMR patients who underwent MVr at our center from November 1, 2017, to June 30, 2023, was conducted. Survival data were analyzed using the Kaplan-Meier method to draw survival curves, and the Log-rank test was used to assess differences in survival times. Factors influencing long-term prognosis were evaluated using the Cox proportional hazards regression model.

Results: The median age of patients in this study was 54 years (IQR 45-62), with 410 female patients (30%). Patients in the MACE group had higher BMI, more preoperative atrial fibrillation, and larger left ventricular end-diastolic and left atrial anteroposterior diameters (54.7 ± 9.6 mm vs. 57.5 ± 10.4 mm, 55 ± 29 mm vs. 73 ± 75 mm, P < 0.05). In the MACE group, 12 patients (27.9%) had a left ventricular ejection fraction (LVEF) below 60%. Survival analysis indicated that patients with preoperative atrial fibrillation had a higher risk of experiencing MACE events in the long term compared to those without atrial fibrillation (P = 0.0038). Patients in the LVEF < 60% group had a higher risk of long-term MACE events than those in the LVEF ≥ 60% group (P = 0.0017). Patients with anterior leaflet lesions had a higher risk of MACE events in the long term compared to those without anterior leaflet lesions (P = 0.0042). Patients with anterior leaflet lesions experienced an increased risk of moderate or greater MR recurrence over time (P = 0.022). There was no significant difference in long-term survival free from moderate or greater MR among the simple, moderate, and complex groups. Cox multivariate regression predictive model analysis showed that preoperative atrial fibrillation, LVEF < 60%, and anterior leaflet lesions were significant independent risk factors for the occurrence of MACE. After adjusting for age, gender, BMI, and surgical complexity, the HR for preoperative atrial fibrillation was 2.20 (95% CI: 1.17-4.14, P = 0.014), for LVEF < 60% was 2.76 (95% CI: 1.41-5.40, P = 0.003), and for anterior leaflet lesions was 2.24 (95% CI: 1.25-4.01, P = 0.007).

Conclusion: Mitral valve repair is an effective treatment for patients with degenerative mitral regurgitation, achieving satisfactory medium to long-term prognosis in experienced medical centers. Preoperative atrial fibrillation, anterior leaflet prolapse, and LVEF < 60% are independent risk factors for the recurrence of moderate or greater mitral regurgitation.

Keywords: mitral valve, degenerative mitral regurgitation, mitral valve repair, prognosis, risk factors

Part II. Re-operation for failed degenerative mitral valve regurgitation: long-term prognosis analysis

Objective: Mitral valve repair (MVr) is associated with the risks of failure and recurrence of mitral regurgitation (MR). Analysis of intraoperative echocardiography, valvular pathology, and initial surgical techniques for secondary valve repair can provide valuable experience to reduce the failure of MVr. This study aims to provide experience and evidence for reducing MVr failures and the selection of secondary repair strategies by reviewing the perioperative and long-term follow-up outcomes of patients undergoing re-MVr and re-MVR at our center.

Methods: A retrospective consecutive inclusion of 114 patients with degenerative mitral regurgitation (DMR) who underwent secondary repair surgery at our center from January 2009 to January 2022 was conducted. Patients' baseline information, surgical data, and follow-up data were analyzed. Survival data were depicted by Kaplan-Meier survival curves, and the differences in survival times were assessed using the Log-rank test. Factors affecting long-term prognosis were evaluated using the Cox proportional hazards regression model.

Results: At the time of reoperation, the average age of the patients was 49 years, with 24 females (21%). Lesions at the time of secondary repair involved the anterior leaflet in 21 cases (18%), the posterior leaflet in 25 cases (22%), and both leaflets in 68 cases (60%). Patients in the MVr group before the secondary surgery had a higher LVEF than those in the MVR group (64 ± 5% vs. 59 ± 14%), which was statistically significant (P = 0.018). The main indications for the secondary surgery included moderate to severe mitral regurgitation, mitral stenosis, hemolysis, infective endocarditis, leaflet perforation, and chordae tendineae rupture. 81 patients (71.0%) had indications for surgery due to moderate or more severe mitral regurgitation, including 4 patients with moderate mitral stenosis; 21 patients (18.4%) had indications for surgery due to moderate or more severe mitral stenosis, including 15 patients with severe mitral stenosis; other indications included hemolysis in 15 cases (13.2%), infective endocarditis in 4 cases (3.5%), leaflet perforation in 2 cases (1.8%), and chordae tendineae rupture in 1 case (0.9%). In this study, 61 cases (54.0%) were related to technique, 48 cases (41.7%) to valvular reasons, and 5 cases (4.4%) had unknown reasons for repair failure. The median time interval for the technical reasons group was 3.4 (1.0 - 9.1) years, and for the valvular reasons group, it was 8.6 (5.2 - 13.5) years, with a significant statistical difference in the time intervals between the two groups of patients undergoing secondary repair surgery (P = 0.002). When assessing postoperative death and recurrence of mitral regurgitation, there was no statistical difference in the overall prognosis between the re-MVr and re-MVR groups.

Conclusion: Technical reasons are the main causes of recurrence of mitral regurgitation in patients undergoing prior mitral valve repair for degenerative mitral valve disease. Mitral valve repair can be reconducted after assessing the quality of the patient's valve and annular morphology, analyzing the etiology, determining the feasibility of mitral valve plasty techniques, and the durability of re-repair.

Keywords: mitral valve, mitral re-repair, prognosis, mitral valve replacement, etiology

Part III. Transcatheter Mitral Valve Repair for Failed Surgical Mitral Valve Repair: A Systematic Review and Meta-analysis

Objectives: To assess the outcomes of transcatheter mitral valve repair (TMVr) for failed previous surgical mitral valve repair (MVr).

Methods: We searched Pubmed, Embase, and Cochrane Library databases for studies that reported the outcomes of TMVr for failed initial surgical MVr. Data were extracted by 2 independent investigators and subjected to meta-analysis. The 95% confidence interval (CI) was calculated for preoperative demographics, peri-operative outcomes, and follow-up outcomes using binary and continuous data from single-arm studies.

Results: Eight single-arm studies were included, with a total of 212 patients, and mean follow-up ranged from 1.0 to 15.9 months. The pooled rate of residual procedural mitral regurgitation ≤ mild was 76% (95% CI: 67% ~ 84%; I2 = 0%; 7 studies, 199 patients). During follow-up, mitral regurgitation ≤ mild was found in 68% patients (95% CI: 52% ~ 82%; I2 = 57%; 6 studies, 147 patients). Follow-up survival was 94% (95% CI: 88% ~ 98%; I2 = 0%; 7 studies, 196 patients). 83% patients (95% CI: 75% ~ 89%; I2 = 47%; 6 studies, 148 patients) was in NYHA class I or II.

Conclusion: TMVr for failed surgical MVr was safe and effective, which should be recommended in selected patients if technically feasible.

Keywords: mitral valve; failure; recurrence; transcatheter mitral valve repair; MitraClip; Neochord

Part IV. Predicitive model for failed repair in mitral valve regurgitaiton based on machine learning

Objective: This study analyzed transesophageal echocardiography (TEE) data from patients undergoing mitral valve repair (MVr) for chronic degenerative mitral regurgitation (DMR) to explore the incidence of and risk factors associated with repair failure. A comprehensive collection of baseline patient data, surgical techniques, and anatomical parameters of the mitral leaflets, annulus, and subvalvular apparatus was conducted based on a prospective cohort study. The aim was to establish a difficulty assessment model for mitral valve repair using various machine learning algorithms, designed to predict the risk of preoperative surgical failure.

Methods: Based on the prospective TEEMR cohort study database registered with ClinicalTrial.gov (NCT05595226), all adult DMR patients who underwent mitral valve repair at Fuwai Hospital between March 2021 and March 2023, and met the inclusion and exclusion criteria, were consecutively enrolled. The final patient cohort for analysis included 159 male and 72 female patients, with a median age of 56 (Interquartile Range: 50, 64) years. The cohort was divided into training and validation sets in a ratio of 8:2 according to the surgery dates. A predictive model was constructed through data preprocessing, feature selection, model building, and validation.

Results: The study developed a machine-learning-based system to accurately assess the complexity of mitral valve repair procedures. After comparing 14 different machine learning models, the Linear Support Vector Classification (LSVC) model demonstrated superior performance in predicting the complexity of mitral valve repair. This model stood out in both the training and validation sets, with an Area Under the Curve (AUC) of 0.942 (95% CI: 0.920 - 0.982) in the training set and 0.819 (95% CI: 0.684 - 0.902) in the validation set, outperforming other models. The model had an F1 score of 0.882, a recall of 0.790, precision of 1.000, and an accuracy of 0.981. Based on clinical significance and the feature coefficients' weights, the final LSVC model included the 14 variables with the highest feature coefficients, encompassing age, annular parameters (MA SLR, MA DLR, MA SAP), leaflet parameters (A2 prolapse, A2 length, P2 length), degree of pathology (RW, VCW, regurgitation degree, PASP), and cardiac function (LAV, LVEF, LVEDV) among others.

Conclusion: The LSVC model exhibited outstanding performance in predicting the complexity of mitral valve repair. Ultrasound parameters such as age, valve lesion location, leaflet length, and annulus size can be used to predict the risk of MVr failure.

Keywords: degenerative mitral regurgitation, mitral valve repair, transesophageal echocardiography, machine learning, prediction model