Validation of a Machine Learning Model for Predicting Postmastectomy Radiotherapy Recommendation Following Immediate...

Area of Science:

• Oncology
• Medical Imaging
• Machine Learning in Medicine

Background:

• Postmastectomy radiotherapy (PMRT) can cause long-term morbidity in patients undergoing immediate implant-based breast reconstruction (IIBBR).
• The potential need for PMRT influences preoperative decisions regarding the type and timing of breast reconstruction.
• Accurate prediction of PMRT is crucial for optimizing patient care and reconstructive strategies.

Purpose of the Study:

• To validate a machine learning (ML) model for predicting PMRT recommendations in patients undergoing IIBBR.
• To ensure the model's adherence to transparent reporting guidelines for prediction models.

Main Methods:

• A cohort of 224 breast cancer patients who underwent mastectomy with IIBBR was analyzed.
• Data included 12 preoperative patient characteristics from clinical history, physical examination, diagnostic imaging, and pathology reports.
• A validated ML model was used to predict PMRT probability.

Main Results:

• 37% of patients were recommended for PMRT.
• The ML model achieved high predictive performance with an area under the receiver operating characteristic curve of 0.80.
• Key predictors for PMRT included lymph node size, axillary lymph node biopsy results, tumor size, and initial ultrasound use.

Conclusions:

• A validated ML model can reliably predict PMRT recommendations for IIBBR patients.
• This predictive tool can assist clinicians in preoperative discussions about breast reconstruction options.
• The model supports informed decision-making for breast cancer patients considering reconstruction after mastectomy.