Predicting complications in breast reconstruction: External validation of a machine learning model

Area of Science:

• Oncology
• Plastic Surgery
• Medical Informatics

Background:

• Nipple-sparing mastectomy (NSM) with immediate implant-based breast reconstruction offers aesthetic and psychosocial benefits.
• Nipple-areolar complex (NAC) necrosis is a significant risk following NSM.
• A machine learning (ML) model was previously developed to predict NAC necrosis.

Purpose of the Study:

• To externally validate a previously developed ML model for predicting NAC necrosis.
• To assess the model's performance in a new patient cohort undergoing NSM with immediate breast reconstruction.

Main Methods:

• Retrospective cohort study of 388 patients undergoing NSM with immediate breast reconstruction.
• Collected demographic, oncologic, and surgical data.
• Applied a validated random forest ML model to predict NAC necrosis and assessed performance using accuracy, AUC-ROC, sensitivity, and specificity.

Main Results:

• 19 patients (4.9%) experienced NAC necrosis.
• Risk factors included older age, higher BMI, active smoking, and larger mastectomy specimen weight.
• The ML model achieved 96% predictive accuracy and an AUC-ROC of 0.70, indicating moderate discriminative ability.

Conclusions:

• The externally validated ML model accurately predicted NAC necrosis in a distinct patient population.
• The model shows potential for personalized risk assessment in NSM candidates.
• Further validation in diverse populations is recommended.