Optimizing Surgery Strategies in Stage IB Lung Squamous Cell Carcinoma: Insights from Interpretable Machine Learning

Area of Science:

• Oncology
• Machine Learning in Medicine
• Biostatistics

Background:

• Stage IB lung squamous-cell carcinoma (LSCC) lacks individualized survival prediction.
• The benefit of adjuvant chemotherapy in stage IB LSCC remains debated.
• Need for improved prognostic tools and treatment strategies in early-stage LSCC.

Purpose of the Study:

• Develop an interpretable machine-learning model for predicting survival in stage IB LSCC.
• Evaluate the added survival benefit of postoperative chemotherapy in stage IB LSCC.
• Provide tools for individualized survival prediction and risk-adapted treatment decisions.

Main Methods:

• Utilized SEER database data from 6445 stage IB LSCC patients (2000-2015).
• Trained six machine-learning algorithms for 1-, 3-, and 5-year overall survival (OS) prediction, with external validation.
• Employed SHAP analysis for model interpretability and propensity-score matching to assess chemotherapy benefit.

Main Results:

• The LightGBM model demonstrated strong predictive performance (AUCs ~0.80-0.83) and generalizability.
• Treatment modality was the key predictor; both surgery alone and surgery plus chemotherapy improved survival.
• No significant OS difference was found between surgery alone and surgery plus chemotherapy, even in high-risk subgroups.

Conclusions:

• The developed model provides individualized survival estimates for stage IB LSCC patients.
• Findings support a risk-adapted, conservative adjuvant treatment strategy.
• Results aid in integrating precision medicine and shared decision-making for early-stage LSCC management.