Journal of thoracic disease · 2026
Development and internal validation of a prognostic model for esophageal cancer liver metastases using the Surveillance, Epidemiology, and End Results database.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Jiang Jingyuan, Zeng Yukai, Wan Jun, Zhang Lening, Zhang Zhe, Bai Rong, Wang Chao, Lin Fengwu
Аннотация
BACKGROUND: Accurate survival prediction for patients with advanced esophageal cancer (EC) and liver metastases is critical for clinical decision-making. However, existing nomograms often lack sufficient precision. To address this limitation, this study aimed to develop and validate a machine learning (ML) model to improve predictive accuracy.
METHODS: Data were extracted from the Surveillance, Epidemiology, and End Results (SEER) database for patients diagnosed with EC and liver metastases between 2010 and 2020. Patients were randomly assigned to training and internal test sets in a 7:3 ratio. Six ML models-Logistic Regression (LR), Random Forest (RF), XGBoost (XGB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Iterative Dichotomiser 3 (ID3)-were trained using clinical variables from the training set. Model selection was performed via the DeLong test. Performance was rigorously evaluated in the internal test set using receiver operating characteristic (ROC) curves, area under the curve (AUC), calibration plots, confusion matrices, and decision curve analysis (DCA). Model interpretability was achieved using the SHapley Additive exPlanations (SHAP) method.
RESULTS: A total of 1,760 patients with EC and liver metastases were included. Among the 6 models, the LR model demonstrated the best performance. In the internal test set, the model achieved an AUC of 0.764 [95% confidence interval (CI): 0.735-0.791] for predicting 6-month survival status, with a sensitivity of 0.641 (95% CI: 0.600-0.680) and a specificity of 0.788 (95% CI: 0.749-0.820). For predicting 1-year survival, the AUC was 0.792 (95% CI: 0.753-0.830), with a sensitivity of 0.784 (95% CI: 0.736-0.831) and a specificity of 0.675 (95% CI: 0.616-0.736). The calibration curve indicated good agreement (the Brier score of 6-month was 0.240, 95% CI: 0.238-0.241; the Brier score of 1-year was 0.158, 95% CI: 0.143-0.175). DCA confirmed the model's net clinical benefit within a reasonable threshold range. SHAP analysis revealed that receipt of chemotherapy was the most important variable influencing prognosis prediction.
CONCLUSIONS: The LR-based model demonstrates favorable predictive performance and may serve as a reference for short-term survival assessment in patients with EC and liver metastases, with receipt of chemotherapy identified as the key predictive variable.
METHODS: Data were extracted from the Surveillance, Epidemiology, and End Results (SEER) database for patients diagnosed with EC and liver metastases between 2010 and 2020. Patients were randomly assigned to training and internal test sets in a 7:3 ratio. Six ML models-Logistic Regression (LR), Random Forest (RF), XGBoost (XGB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Iterative Dichotomiser 3 (ID3)-were trained using clinical variables from the training set. Model selection was performed via the DeLong test. Performance was rigorously evaluated in the internal test set using receiver operating characteristic (ROC) curves, area under the curve (AUC), calibration plots, confusion matrices, and decision curve analysis (DCA). Model interpretability was achieved using the SHapley Additive exPlanations (SHAP) method.
RESULTS: A total of 1,760 patients with EC and liver metastases were included. Among the 6 models, the LR model demonstrated the best performance. In the internal test set, the model achieved an AUC of 0.764 [95% confidence interval (CI): 0.735-0.791] for predicting 6-month survival status, with a sensitivity of 0.641 (95% CI: 0.600-0.680) and a specificity of 0.788 (95% CI: 0.749-0.820). For predicting 1-year survival, the AUC was 0.792 (95% CI: 0.753-0.830), with a sensitivity of 0.784 (95% CI: 0.736-0.831) and a specificity of 0.675 (95% CI: 0.616-0.736). The calibration curve indicated good agreement (the Brier score of 6-month was 0.240, 95% CI: 0.238-0.241; the Brier score of 1-year was 0.158, 95% CI: 0.143-0.175). DCA confirmed the model's net clinical benefit within a reasonable threshold range. SHAP analysis revealed that receipt of chemotherapy was the most important variable influencing prognosis prediction.
CONCLUSIONS: The LR-based model demonstrates favorable predictive performance and may serve as a reference for short-term survival assessment in patients with EC and liver metastases, with receipt of chemotherapy identified as the key predictive variable.