基于决策树与Logistic回归模型预测肝细胞癌患者TACE术后的疗效

Decision Tree and Logistic Regression Model⁃Based Prediction of Outcome after Transarterial Chemoembolization in Patients with Hepatocellular Carcinoma

目的探讨肝细胞癌(HCC)患者经肝动脉化疗栓塞术(TACE)术后影响疗效的相关因素。方法回顾性分析113例HCC患者,根据改良实体瘤疗效评价标准(mRECIST),将患者分为客观疗效组和无疗效组,比较组间肝血管造影(HAG)、CT图像、临床信息的差异。将单因素分析中P<0.05的征象纳入Logistic回归和决策树模型,筛选与HCC患者TACE治疗反应相关的独立预测因素。计算受试者工作特征曲线(ROC)曲线下面积(AUC),并以DeLong检验比较模型差异。结果单因素分析得出甲胎蛋白(AFP)、肿瘤大小、相对延迟期碘油CT值、碘油沉积类型、肿瘤位置、肿瘤染色差异有统计学意义(P均<0.05);决策树模型分析得出,TACE预测疗效具有重要意义的前3项决策指标分别为碘油沉积类型、肿瘤位置、肿瘤染色,AUC为0.902;Logistic回归模型显示,碘油沉积类型、相对延迟期碘油CT值、肿瘤位置、肿瘤染色是TACE治疗疗效的独立预测因素,AUC为0.950;Delong检验得出,两种模型差异无统计学意义(Z=1.913,P=0.0558)。结论肿瘤染色、肿瘤位置、碘油沉积类型对预测HCC患者TACE治疗效果具有重要意义,有助于更好地筛选从TACE中获益的HCC患者。

Objective To investigate the factors affecting patients with hepatocellular carcinoma(HCC)after transarte⁃rial chemoembolization(TACE).Methods 113 HCC patients,divided into objective efficacy group and no efficacy group according to the modified efficacy evaluation criteria for solid tumors(mRECIST),comparing the differences in hepatic an⁃giography(HAG),CT images,and clinical information between groups were retrospective analysed.Signs with P<0.05 in the univariate analysis were included in Logistic regression and decision tree models to screen for independent predictors associated with response to TACE treatment in HCC patients.The area under the subject operating characteristic curve(AUC)was calculated and the DeLong test was used to compare model differences.Results Univariate analysis showed statistically significant differences in alphafeto protein,tumor size,relative delayed period iodine CT value,iodine deposition type,tumor location,and tumor staining(all P<0.05);decision tree model analysis showed that the first three decision in⁃dicators with significant significance in predicting the efficacy of TACE were iodine deposition type,tumor location,and tumor staining,with an AUC of 0.902.Logistic regression model showed that the type of iodine deposition,relative delayed period iodine CT value,tumor location,and tumor staining were independent predictors of the efficacy of TACE treatment,with an AUC of 0.9501 the Delong test yielded no statistically significant difference between the two models(Z=1.913,P=0.0558).Conclusion Tumor staining,tumor location,and type of iodine oil deposition are important in predicting the effect of TACE treatment in HCC patients and help to better screen HCC patients who benefit from TACE.