Exploring the Potential Role of Upper Abdominal Peritonectomy in Advanced Ovarian Cancer Cytoreductive Surgery Using...

Area of Science:

• Oncology
• Surgical Oncology
• Machine Learning in Medicine

Background:

• The Surgical Complexity Score (SCS) is used for advanced epithelial ovarian cancer (EOC) cytoreduction but doesn't capture all procedures.
• The European Society for Gynaecological Oncology (ESGO) established quality indicators for complete cytoreduction (CC0), necessitating a defined weighting for contributing surgical sub-procedures.

Purpose of the Study:

• To identify and weight surgical sub-procedures predictive of achieving CC0 in advanced EOC.
• To develop a machine learning model for predicting CC0 and assess the impact of specific procedures on survival outcomes.

Main Methods:

• Analysis of prospectively collected data from 560 advanced EOC patients.
• Utilized eXtreme Gradient Boosting (XGBoost) to model surgical sub-procedures and Shapley Additive explanations (SHAP) for feature importance.
• Employed Cox regression and Kaplan-Meier curves for survival analysis.

Main Results:

• The XGBoost model predicted CC0 with an AUC of 0.70.
• Upper abdominal peritonectomy (UAP) was identified as the most significant predictor of CC0, correlating strongly with bladder peritonectomy and diaphragmatic stripping.
• UAP addition improved the predictive model's AUC to 0.80 and was associated with poorer progression-free survival (HR=1.76).

Conclusions:

• Machine learning effectively identifies critical surgical predictors for CC0 in advanced EOC.
• UAP is the most important procedural predictor of CC0, highlighting the significance of upper abdominal quadrant assessment.
• The developed classification model can serve as a digital surgical reference, potentially improving CC0 achievement in high-volume centers.