Defining the Learning Curve in Minimally Invasive Cardiac Surgery: A Systematic Review and Meta-Analysis.

Abstract

Minimally invasive cardiac surgery (MICS) has become a popular approach due to its potential benefits, such as improved cosmesis, faster recovery, shorter hospital stays, and cost-effectiveness, compared with traditional median sternotomy. However, there have been some concerns regarding procedural efficiency and surgical outcomes, especially in the early phase of the learning curve of these procedures. In March 2025, a systematic review was conducted using MEDLINE, Embase, the Cochrane Library and Google Scholar databases to identify potential studies that quantitively assessed the learning curve in MICS using predefined metrics based on surgical times and/or clinical outcomes. There were 28 studies involving 13,257 patients that met the inclusion criteria, most of which were retrospective, focusing on 3 types of MICS: minimally invasive mitral valve surgery, aortic valve replacement, and coronary artery bypass grafting. The learning curve was assessed using arbitrary (split-group) and nonarbitrary (cumulative sum) methods. Common perioperative metrics included operative, cardiopulmonary bypass, aortic cross-clamp times, and postoperative complications. The reported number of cases needed to overcome the learning curve varied widely, ranging from 23 to 125 cases (mean, 39 cases [for repair] and 78 [for replacement]) for minimally invasive valve surgery, 40 to 138 cases (mean, 93 cases) for minimally invasive aortic valve replacement, and 16 to 100 cases (mean, 40 cases) for minimally invasive coronary artery bypass grafting. Differences in surgical process and postoperative outcomes suggest a learning curve in MICS, although stable morbidity and mortality rates indicate the safe adoption of these procedures with appropriate training. Nonetheless, significant heterogeneity across studies prevents precise learning curve characterization, highlighting the need for standardized, multivariable assessment frameworks.