摘要
Abstract
Objective:To construct a risk prediction model for perioperative moderate-to-severe cancer-related fatigue(CRF)in lung cancer surgery patients based on machine learning algorithms,and to validate its discriminative performance.Methods:A total of 280 patients with non-small cell lung cancer(NSCLC)undergoing surgery at hospital from June 2023 to June 2024 were enrolled as the development cohort(106 cases with moderate-to-severe CRF),and another 120 patients of the same type from July 2024 to June 2025 were selected as the external validation cohort.Based on perioperative clinical data,variable selection was performed using univariate analysis and least absolute shrinkage and selection operator(LASSO)regression.Subsequently,Logistic regression,decision classification and regression tree(DCRT),and back propagation neural network(BPNN)were constructed.The discriminative ability of each model was compared using receiver operating characteristic(ROC)curves and the area under the curve(AUC),and generalization performance was assessed using independent external data.Results:Univariate and LASSO regression analyses revealed that comorbid chronic obstructive pulmonary disease(COPD),higher TNM stage,higher preoperative anxiety and depression levels,poorer preoperative sleep quality,more severe preoperative pain,and higher levels of C-reactive protein(CRP)and interleukin-6(IL-6)were all characteristic factors for perioperative moderate-to-severe CRF in lung cancer surgery patients(all P<0.05).The AUCs of all three models exceeded 0.700(all P<0.05),with the BPNN model demonstrating the best predictive performance.Decision curve analysis based on the BPNN model showed that,within a risk threshold range of 0.10～0.95,its clinical net benefit was superior to both the"treat-all"and"treat-none"strategies.Conclusions:Comorbid COPD,TNM stage,preoperative anxiety and depression levels,preoperative sleep quality,preoperative pain intensity,CRP,and IL-6 can effectively predict perioperative moderate-to-severe CRF in lung cancer surgery patients.The BPNN model constructed on this basis exhibits optimal discriminative ability and holds potential for clinical translation.关键词
肺癌/围术期/癌因性疲乏/机器学习Key words
lung cancer/perioperative period/cancer-related fatigue,CRF/machine learning
