肺癌脑转移患者预后不良的影响因素及预测模型建立

Influencing factors of poor prognosis of patients with brain metastasis from lung cancer and prediction model construction

目的探讨肺癌脑转移患者治疗后1年内预后不良的影响因素,建立预测肺癌脑转移患者治疗后1年内预后不良的列线图模型。方法肺癌脑转移患者387例,其中单纯放疗127例,单纯靶向治疗136例,单纯化疗84例,放疗+靶向治疗+化疗40例。患者治疗后随访1年,根据治疗后1年内生存情况分为死亡组290例和生存组97例;比较2组患者年龄、递归分区分析(recursive partitioning analysis,RPA)分级、脑脊液循环肿瘤DNA(circulating tumor DNA,ctDNA)水平等,采用多因素Cox回归分析肺癌脑转移患者治疗后1年内预后不良的影响因素,根据影响因素采用R软件建立预测肺癌脑转移患者治疗后1年内预后不良的列线图模型,采用C-指数评估列线图模型的预测效能;绘制校准曲线,进行内部验证,评估模型的校准度;绘制ROC曲线,评估列线图模型预测肺癌脑转移患者治疗后1年内预后不良的效能。结果死亡组年龄[(64.95±7.46)岁]、脑脊液ctDNA水平[(65.21±14.89)μg/L]均高于生存组[(62.71±6.85)岁、(47.38±10.32)μg/L](t=2.612,P=0.009;t=10.943,P<0.001),RPA分级Ⅰ级比率(26.55%)、放疗+靶向治疗+化疗比率(3.79%)均低于生存组(41.24%、29.90%)(P<0.05),性别比例、吸烟史和肺癌家族史比率、组织病理类型、KPS评分与生存组比较差异均无统计学意义(P>0.05)。年龄(OR=1.062,95%CI:1.008~1.120,P=0.025)、脑脊液ctDNA(OR=1.121,95%CI:1.088~1.156,P<0.001)、单纯放疗(OR=16.160,95%CI:2.930~89.124,P=0.001)、单纯靶向治疗(OR=12.793,95%CI:2.562~63.894,P=0.002)、单纯化疗(OR=10.848,95%CI:2.728~43.138,P=0.001)、RPA分级Ⅱ~Ⅲ级(OR=4.132,95%CI:1.104~15.625,P=0.035)是肺癌脑转移患者治疗后1年内预后不良的影响因素。以年龄、脑脊液ctDNA、RPA分级、治疗方案建立预测肺癌脑转移患者治疗后1年内预后不良的列线图模型,模型C-指数为0.913(95%CI:0.892~0.941),有较好预测效能。模型校准曲线与理想模型接近,内部验证C-指数为0.911。列线图模型预测肺癌脑转移患者治疗后1年内预后不良AUC为0.873(95%CI:0.836~0.904,P<0.001)。结论高龄、脑脊液ctDNA水平增高、单纯放疗、单纯靶向治疗、单纯化疗及RPA分级Ⅱ~Ⅲ级是肺癌脑转移患者治疗后1年内预后不良的危险因素;以年龄、脑脊液ctDNA、RPA分级、治疗方案建立的列线图模型可有效预测肺癌脑转移患者治疗后1年内预后不良的风险。

Objective To investigate the influencing factors of poor prognosis in 1 year after treatment in patients with brain metastasis from lung cancer,and to construct a nomogram model to predict poor prognosis.Methods In 387 patients with brain metastasis from lung cancer,127 patients received radiotherapy,136 patients received targeted therapy,84 patients received chemotherapy,and 40 patients received radiotherapy+targeted therapy+chemotherapy.The patients were followed up for 1 year after treatment,and were divided into death group(n=290)and survival group(n=97).The age,recursive partitioning analysis(RPA)grade and circulating tumor DNA(ctDNA)level of cerebrospinal fluid were compared between two groups.Multivariate Cox regression was used to analyze the influencing factors of poor prognosis in 1year after treatment in patients with brain metastasis from lung cancer.According to the influencing factors,R software was used to construct a nomogram model to predict the poor prognosis,and C-index was used to evaluate the prediction efficiency of the model.The calibration curve was drawn to verify the calibration degree of the evaluation model.ROC curve was drawn to evaluate the efficiency of the nomogram model on predicting poor prognosis.Results The patients were older in death group[(64.95±7.46)years]than survival group[(62.71±6.85)years](t=2.612,P=0.009),the ctDNA level was higher in death group[(65.21±14.89)μg/L]than that in survival group[(47.38±10.32)μg/L](t=10.943,P<0.001),the percentages of patients with RPA gradeⅠand patients receiving radiotherapy+targeted therapy+chemotherapy were lower in death group(26.55%,3.79%)than those in survival group(41.24%,29.90%)(P<0.05),and there were no significant differences in the gender ratio,percentages of patients with smoking history,family history of lung cancer,pathological type and KPS score between two groups(P>0.05).The age(OR=1.062,95%CI:1.008-1.120,P=0.025),ctDNA level(OR=1.121,95%CI:1.088-1.156,P<0.001),radiotherapy(OR=16.160,95%CI:2.930-89.124,P=0.001),targeted therapy(OR=12.793,95%CI:2.562-63.894,P=0.002),chemotherapy(OR=10.848,95%CI:2.728-43.138,P=0.001)and RPAⅡ-Ⅲgrade(OR=4.132,95%CI:1.104-15.625,P=0.035)were the influencing factors of poor prognosis in 1year after treatment in patients with brain metastasis from lung cancer.A nomogram model was constructed to predict the poor prognosis based on the age,ctDNA level,RPA grade and treatment regime.The C-index of the model was 0.913(95%CI:0.892-0.941),indicating agood predictive efficiency.The calibration curve of the model was close to the ideal model,and the internal validation C-index was 0.911.The AUCof nomogram model for predicting poor prognosis was0.873(95%CI:0.836-0.904,P<0.001).Conclusions The old age,increased ctDNA level,radiotherapy,targeted therapy,chemotherapy and RPA gradeⅡ-Ⅲare the risk factors of poor prognosis in 1year after treatment in patients with brain metastasis from lung cancer.The nomogram model by age,ctRNA,RPA grade and treatment regime can effectively predict the risk of poor prognosis.