BACKGROUND Minimally invasive hepatectomy (MIH) has become an important option for the treatment of various liver tumours. A major concern is the learning curve required. The aim of this study was to perform a systematic review and summarize current literature analysing the learning curve for MIH. METHODS A systematic review of the literature pertaining to learning curves in MIH to July 2019 was performed using PubMed and Scopus databases. All original full-text articles published in English relating to learning curves for both laparoscopic liver resection (LLR), robotic liver resection (RLR), or a combination of these, were included. To explore quantitatively the learning curve for MIH, a meta-regression analysis was performed. RESULTS Forty studies relating to learning curves in MIH were included. The median overall number of procedures required in studies utilizing cumulative summative (CUSUM) methodology for LLR was 50 (range 25-58) and for RLR was 25 (16-50). After adjustment for year of adoption of MIH, the CUSUM-derived caseload to surmount the learning curve for RLR was 47.1 (95 per cent c.i. 1.2 to 71.6) per cent; P = 0.046) less than that required for LLR. A year-on-year reduction in the number of procedures needed for MIH was observed, commencing at 48.3 cases in 1995 and decreasing to 23.8 cases in 2015. CONCLUSION The overall learning curve for MIH decreased steadily over time, and appeared less steep for RLR compared with LLR.

中文翻译：

---

微创肝切除术中的学习曲线：系统评价和Meta回归分析。

背景技术微创肝切除术（MIH）已经成为治疗各种肝肿瘤的重要选择。一个主要的问题是所需的学习曲线。这项研究的目的是进行系统的审查，并总结分析MIH学习曲线的最新文献。方法使用PubMed和Scopus数据库对截至2019年7月MIH学习曲线的文献进行系统回顾。包括所有以英文发表的与腹腔镜肝切除术（LLR），机器人肝切除术（RLR）或这些方法的组合有关的学习曲线的原始全文。为了定量探索MIH的学习曲线，进行了荟萃回归分析。结果包括40项有关MIH学习曲线的研究。使用累积累加（CUSUM）方法进行的研究中，LLR所需的平均过程总数为50（范围为25-58），而RLR的平均过程总数为25（16-50）。在对采用MIH的年份进行调整之后，CUSUM派生的案例量超过了RLR的学习曲线，为47.1％（95％ci 1.2至71.6）。P = 0.046）小于LLR所需的值。观察到MIH所需的程序数量逐年减少，从1995年的48.3例开始，到2015年减少到23.8例。结论MIH的总体学习曲线随时间稳步下降，而RLR的陡峭程度降低与LLR相比。CUSUM派生的案例量超过了RLR的学习曲线，为47.1％（95％从1.2到71.6）。P = 0.046）小于LLR所需的值。观察到MIH所需的程序数量逐年减少，从1995年的48.3例开始，到2015年减少到23.8例。结论MIH的总体学习曲线随着时间的推移稳步下降，而RLR的陡峭程度降低与LLR相比。CUSUM派生的案例量超过了RLR的学习曲线，为47.1％（95％从1.2到71.6）。P = 0.046）小于LLR所需的值。观察到MIH所需的程序数量逐年减少，从1995年的48.3例开始，到2015年减少到23.8例。结论MIH的总体学习曲线随时间稳步下降，而RLR的陡峭程度降低与LLR相比。