A Survival Prognosis Prediction Model for Locally Advanced Laryngeal Cancer Based on Feature Selection Through...

Area of Science:

• Oncology
• Medical Statistics
• Surgical Oncology

Background:

• Locally advanced laryngeal cancer (LALC) presents complex survival challenges.
• Accurate prognostication is crucial for tailoring treatment strategies.
• Existing staging systems may not fully capture individual risk.

Purpose of the Study:

• To identify high-risk factors associated with survival in LALC patients.
• To develop and validate a robust prognostic prediction model for LALC.
• To aid in the selection of optimal, individualized treatment plans.

Main Methods:

• Utilized LASSO, XGBoost, and random forests for feature selection from 283 LALC patients.
• Developed a nomogram based on a COX regression model with internal validation (bootstrap).
• Evaluated model performance using ROC, AUC, C-index, and DCA; compared with AJCC 8th TNM stage.

Main Results:

• Identified seven significant predictors for LALC survival.
• The developed nomogram demonstrated excellent discrimination (AUCs 0.852-0.829) and calibration.
• The COX model outperformed AJCC 8th TNM staging for predicting 5-year survival.

Conclusions:

• A COX regression model incorporating Age, Treatment, Surgery, DAA, K+, LNR, and TCIS effectively predicts overall survival (OS) in LALC.
• The model successfully stratifies patients into low- and high-risk groups.
• Surgery, potentially with adjuvant radiotherapy, is recommended for high-risk LALC patients.