食管癌术后癌因性疲乏预测模型的建立与评估

Construction and evaluation of prediction model of cancer-related fatigue after esophageal cancer surgery

目的探讨食管癌术后癌因性疲乏(CRF)的危险因素,构建列线图模型。方法收集2018年3月至2021年2月福建省肿瘤医院门诊及病房复诊的292例食管癌术后患者作为建模组,并进行内部验证。采用LASSO回归、十折交叉验证获得CRF最佳风险预测因子子集,采用多因素logistic回归分析判断CRF的独立危险因素,通过R软件(R4.1.1)建立列线图模型,绘制受试者工作特征曲线(ROC)并计算曲线下面积(AUC)评估模型的区分度,绘制校准曲线评估模型的精准度。将2021年3月至12月福建省肿瘤医院符合同样标准的98例食管癌术后患者作为验证组,对模型进行外部验证。结果建模组中,食管癌术后CRF的发生率为39.7%,LASSO回归结果表明年龄、性别、术前睡眠障碍、肿瘤大小、术后放疗、术后化疗是预测CRF的自变量(P<0.05)。多因素logistic回归结果表明高龄(β=0.943,OR=2.567,95%CI:1.532~4.303)、女性(β=0.632,OR=1.882,95%CI:1.081~3.274)、术前睡眠障碍(β=0.788,OR=2.200,95%CI:1.302~3.716)、术后放疗(β=0.849,OR=2.337,95%CI:1.239~4.406)、术后化疗(β=0.907,OR=2.478,95%CI:1.390~4.416)是CRF的独立危险因素(P<0.05)。构建列线图预测模型,建模组和验证组的AUC分别为0.732(95%CI:0.673~0.791)和0.757(95%CI:0.687~0.827);两组的校准曲线显示预测结果与实际结果的一致性均良好。结论食管癌术后CRF的发生率高,构建的列线图模型的预测效能好,为制定更加合理、准确的个体化预防措施提供参考。

Objective To explore the risk factors of cancer related fatigue(CRF)after esophageal cancer surgery and construct a nomogram model.Methods A total of 292 patients with esophageal cancer who were re-examined in the outpatient and ward of Fujian Cancer Hospital from March 2018 to February 2021 were collected as the modeling group,and internal validation was carried out.The LASSO regression and ten fold cross validation were used to obtain the best subset of risk predictors for CRF.The multifactor logistic regression analysis was used to determine the independent risk factors for CRF.The nomogram model was established through R software(R4.1.1),the receiver operating characteristic curve(ROC)was drawn,and the discrimination of the area under the curve(AUC)evaluation model was calculated.The calibration curve was drawn to evaluate the accuracy of the model.98 patients with esophageal cancer who met the same criteria from March to December 2021 were taken as the validation group,and the model was externally validated.Results In the modeling group,the incidence of CRF after esophageal cancer surgery was 39.7%.LASSO regression results showed that age,sex,preoperative sleep disorder,tumor size,postoperative radiotherapy,postoperative chemotherapy were independent variables predicting CRF(P<0.05).Multi factor logistic regression results show that the elderly(β=0.943,OR=2.567,95%CI:1.532-4.303),female(β=0.632,OR=1.882,95%CI:1.081-3.274),preoperative sleep disorder(β=0.788,OR=2.200,95%CI:1.302-3.716),postoperative radiotherapy(β=0.849,OR=2.337,95%CI=1.239-4.406),postoperative chemotherapy(β=0.907,OR=2.478,95%CI:1.390-4.416)were independent risk factors of CRF.The AUC of modeling group and validation group were 0.732(95%CI:0.673-0.791)and 0.757(95%CI:0.687-0.827)respectively(P<0.05).The calibration curves of the two groups showed that the predicted results were in good agreement with the actual results.Conclusion The incidence of CRF after esophageal cancer surgery is high,and the nomogram model has a good predictive effect,which provides a reference for formulating more reasonable and accurate individual preventive measures.