"Identifying complication risk factors in reduction mammaplasty: a single-center analysis of 1021 patients applying machine learning methods".

Abstract

Various surgical approaches and pedicles have been described to ensure safe and satisfactory results in reduction mammaplasty. Although different breasts require different techniques, complications are common. This study aims to assess the incidence of complications following primary bilateral reduction mammaplasties across a diverse range of pedicle methods within one of the largest single-center cohorts to date, utilizing machine learning methodologies. A retrospective review of primary bilateral reduction mammaplasties at a single surgical center between January 2016 and March 2020 was performed. Patient medical records and surgical details were reviewed. Complications were compared among three different pedicles. Binary recursive partitioning (CART) machine learning was employed to identify risk factors. In total, 1021 patients (2142 breasts) met the inclusion criteria. The superomedial pedicle was the most frequently utilized (48.0%), with an overall complication rate of 21%. While pedicle-based subgroups demonstrated significant demographic variance, overall complication rates differed most between the inferior (24.9%) and the superomedial pedicle (17.7%). Statistical analysis identified resection weight as the sole significant independent risk factor (OR 1.001, p = 0.007). The machine learning model revealed that total resection weights exceeding 1700 g significantly increased the risk of overall complications, while a sternal notch to nipple (SNN)-distance > 36.5 cm correlated with complications involving the nipple-areola complex (NAC). Higher resection weights are associated with elevated complication rates. Preoperative assessment utilizing SNN-distance can aid in predicting NAC complications.