A random forest prediction study on the 3-year recurrence-free survival of early and middle stage esophageal cancer after total endoscopic resection
10.3969/j.issn.1005-6483.20240613
- VernacularTitle:早中期食管癌全腔镜切除术后3年无复发生存的随机森林预测研究
- Author:
Sanhu YANG
1
;
Yan LI
1
;
Lijun HUANG
1
;
Zhenke YAN
1
;
Xu LIU
1
;
Wanshan LI
1
;
Xiang JI
1
Author Information
1. 710038 西安,空军军医大学唐都医院胸腔外科
- Publication Type:Journal Article
- Keywords:
esophageal cancer;
early metaphase;
total endoscopic resection;
random forest method;
Cox regression analysis
- From:
Journal of Clinical Surgery
2025;33(5):486-492
- CountryChina
- Language:Chinese
-
Abstract:
Objective To construct a predictive model for the 3-year recurrence-free survival(RFS)after total endoscopic resection of early and mid-stage esophageal cancer,and to test it,in order to provide decision support for standardized management after total endoscopic resection of early and mid-stage esophageal cancer.Methods A retrospective study was conducted to include 306 patients with early-to-mid stage esophageal cancer who underwent total endoscopic resection in our hospital from January 2018 to December 2020.The patients were divided into a modeling set(n=204)and a validation set(n=102)according to a 2∶1 ratio.Univariate analysis and random forest algorithm were used to screen variables,and Cox regression analysis was used to analyze the factors affecting the 3-year RFS after total endoscopic resection for early-to-mid stage esophageal cancer.The R language was used to construct a nomogram prediction model for model validation,and the receiver operating characteristic curve(ROC curve)was drawn to calculate the area under the curve(AUC).The discrimination of the prediction model was evaluated,and the calibration curve and decision curve(DCA curve)were used to evaluate the predictive performance and clinical applicability of the prediction model.Results Among the 306 patients with early and mid-stage esophageal cancer who underwent total endoscopic resection,18 died 3 years after the operation,55 relapsed,233 achieved RFS,and the 3-year RFS rate was 76.14％.Through univariate and random forest algorithm screening,seven factors were identified as being associated with the RFS of patients three years after surgery.These factors were entered into a Cox regression analysis,and the results showed that positive abdominal lymph nodes,vascular cancer thrombus,clinical stage Ⅲ,gross type of erosion,age ≥ 65 years,and tumor diameter＞3 cm were risk factors for RFS three years after surgery(P＜0.05).Based on this,a nomogram prediction model for RFS three years after full endoscopic resection for early-to-mid stage esophageal cancer was constructed.Internal and external validation showed that the consistency index of the prediction model in the modeling set was 0.881,and the consistency index in the validation set was 0.867.The ROC curve validation showed that the AUC of the prediction model in the modeling set and validation set were 0.855(95％CI:0.778-0.932)and 0.826(95％CI:0.763-0.890),respectively.The DCA curve validation showed that the risk threshold of the modeling set and validation set were 0-0.95 and 0-0.77,respectively,when the model could achieve high net benefits.Conclusion The 3-year RFS after total endoscopic resection for early and middle stage esophageal cancer is related to multiple factors.The nomogram model based on clinical stage Ⅲ,positive abdominal lymph nodes,vascular tumor thrombus,and gross type of erosion has good clinical utility for predicting the 3-year RFS of patients after surgery,and is of guiding significance for medical staff in making decisions about the management of early and middle stage esophageal cancer after surgery.
