A Prediction Model for Metachronous Peritoneal Carcinomatosis in Patients with Stage T4 Colon Cancer after Curative...

Area of Science:

• Oncology
• Cancer Research
• Clinical Prediction Modeling

Background:

• Colon cancer staging and risk stratification are critical for patient management.
• Accurate assessment of metastatic potential is essential for treatment planning in advanced colon cancer.

Purpose of the Study:

• To develop and validate a prediction model for assessing individual risk of distant metastasis (mPC) in patients with pT4 colon cancer.
• To identify key predictors and create a risk stratification tool for pT4 colon cancer patients.

Main Methods:

• Developed 2044 Cox prediction models using a training set of 2003 pT4 colon cancer patients undergoing R0 resection.
• Selected the optimal model based on maximal C-index and minimal prediction error.
• Validated the final model using a testing set, employing time-dependent AUC and Brier score, and developed a risk scoring system with CART analysis for risk stratification.

Main Results:

• Identified five key predictors: tumor location, preoperative carcinoembryonic antigen (CEA) level, histologic type, T stage, and nodal stage.
• The developed model demonstrated high predictive accuracy (prediction error ≤5%) and good discrimination ability (AUC >0.7).
• Patients were successfully stratified into low-risk and high-risk groups based on the developed scoring system.

Conclusions:

• The prediction model effectively quantifies individual risk for developing distant metastasis in pT4 colon cancer.
• This tool is feasible for identifying and selecting high-risk pT4 colon cancer patients who may benefit from closer monitoring or specific therapeutic strategies.