Construction And Validation Of A Risk Prediction Model For Cancer-related Fatigue In Breast Cancer Patients

ObjectiveTo identify the independent risk factors of cancer-related fatigue in breast cancer patients,construct and validate the risk prediction model of cancer-related fatigue in breast cancer patients,so as to provide a more scientific and convenient evaluation tool for the prediction of cancer-related fatigue in breast cancer patients.MethodsTo investigate the influencing factors of cancer-related fatigue in breast cancer patients and identify alternative predictors based on literature review.Based on alternative predictors,convenience sampling method was used to collect information about breast cancer patients who visited the Taihe Hospital in Shiyan City from December 2020 to December 2021.According to the order of inclusion,the data of the first 70%patients（model construction group）were used for the construction and internal evaluation of the risk prediction model,and the data of the last 30%patients（model validation group）were used for the external validation of the model.Using the latest（2018 version）definition of cancer-related fatigue from the National Comprehensive Cancer Network（NCCN）as the diagnostic criteria,breast cancer patients were divided into the cancer-related fatigue group and the non-cancer-related group.With the occurrence of cancer-related fatigue as the dependent variable and alternative predictors as the independent variable,univariate Logistic analysis and multivariate Logistic analysis were performed to identify the independent factors of cancer-related fatigue in breast cancer patients in the model construction group.Based on the partial regression coefficients of screened independent factors,a risk model of cancer-related fatigue in breast cancer patients was constructed and evaluated internally and externally that the differentiation of the model was evaluated by the area under the ROC curve,the accuracy of the model was analyzed by the Calibration curve,the clinical practicability of model was assessed by the Decision Curve Analysis.Results1 Basic information of cancer-related fatigue in breast cancer patientsA total of 240 breast cancer patients were enrolled,including 168 in the model construction group and 72 in the model validation group.Among the 240 breast cancer patients,149 patients developed cancer-related fatigue,including 110 patients（65.5%）in the model construction group and 39 patients（54.2%）in the validation group.The overall incidence of CRF in breast cancer patients in this study was 62.1%.2 Comparison of equilibrium between model construction group and model validation groupThe general demographic data,disease-related data and laboratory indicators of the model construction group and the model validation group were analyzed,and the P values were all greater than 0.05 which indicated that there was no statistically significant difference in the baseline data between the two groups,and the homogeneity was good and comparable,so the model could be verified with the data of the validation group.3 Construction of a prediction model for the risk of cancer-related fatigue in breast cancer patientsMultivariate Logistic regression analysis showed that anxiety and depression were independent risk factors for cancer fatigue in breast cancer patients.The model of cancer-related fatigue in breast cancer patients:Logit（P）=-2.08-0.54X₁+0.15X₂+0.20X₃=-2.08-0.54*medical insurance type+0.15*anxiety+0.20*depression.4 Internal evaluation and external validation of the risk prediction model for canc er-related fatigue in breast cancer patientsIn the model construction group,the area under ROC curve was 0.832（95%CI:0.72-0.92）,and the C-Index was 0.824（95%CI:0.7666-0.898）,both of which indicated that the prediction model could distinguish the patients with cancer-related fatigue well from the patients with non-cancer-related fatigue.The Calibration curve fitted well（U:P=1.00>0.05）and Hosmer-Lemeshow goodness of fit testχ²=2.592（P=0.274>0.05）,indicating that the predicted risk of cancer-related fatigue in breast cancer patients was consistent with the actual risk.Decision curve analysis showed that the net benefit of this model was better when the thresholds were 2%-9%,15%-95%and 97%-99%.The optimal critical value of the model was 0.595 when the Yuden index in the ROC curve was the largest（0.336）,and the sensitivity and specificity of the model were 0.741 and 0.595.In the external validation of the model,the area under ROC curve was 0.819（95%CI:0.721-0.918）,and the Calibration curve was consistent（U:P=0.471>0.05）.Decision curve analysis shows that the model has good clinical practicability when the threshold value was between 5%～90%and 93%～97%..ConclusionThe established risk prediction model for cancer-related fatigue in breast cancer patients has good differentiation,accuracy and clinical practicability,which may be used as an evaluation and prediction tool to help medical staff identify the high-risk population of breast cancer patients with cancer-related fatigue as soon as possible.