原发性肝癌干预前并发肺部感染风险预测模型的建立与验证

Establishment and validation of a risk prediction model for primary liver cancer complicated with pulmonary infection before intervention

目的分析原发性肝癌(PHC)患者干预前合并肺部感染的影响因素,建立列线图风险预测模型并进行验证。方法选取首次确诊并住院的PHC患者1635例,依据是否合并肺部感染分为感染组197例和非感染组1438例。收集患者的体质量指数(BMI)、慢性乙型肝炎(CHB)、慢性丙型肝炎(CHC)、巴塞罗那分期(BCLC)等一般资料,白细胞(WBC)、中性粒细胞(NEU)、血红蛋白(Hb)等血液常规指标,总蛋白(TP)、前白蛋白(PA)、天冬氨酸转氨酶(AST)、γ谷氨酰转移酶(GGT)、碱性磷脂酶(ALP)、异常凝血酶原(PIVKA-Ⅱ)、甲胎蛋白(AFP)、癌胚抗原(CEA)、降钙素原(PCT)、超敏C反应蛋白(hs-CRP)、胆碱酯酶(ChE)、总胆固醇(TC)等血液生化指标,CD3细胞计数(CD3^(+))、CD4细胞计数(CD4^(+))、CD4/CD8比值(CD4^(+)/CD8^(+))、CD19细胞计数(CD19^(+)),白细胞介素(IL)-2、干扰素(IFN)-α、肿瘤坏死因子(TNF)-α等细胞因子。采用单因素分析和Lasso回归筛选变量后使用二元Logistic回归分析确定PHC患者干预前发生肺部感染的危险因素,建立并评价列线图风险预测模型。结果与非感染组比较,感染组的年龄、吸烟比例、CHC、胸腔积液、胃肠道出血、Child-Pugh分级C级、BCLC A期/C期/D期比例、WBC、NEU、AST、GGT、ALP、PIVKA-Ⅱ、AFP、CEA、PCT、hs-CRP、IL-2、IL-5、IL-6、IL-8、IL-10、IL-12、IFN-γ、TNF-α水平较高,BMI水平、CHB比例、Hb、TP、PA、ChE、TC、CD3^(+)、CD4^(+)、CD4^(+)/CD8^(+)、CD19^(+)、IFN-α水平较低(P<0.05)。Lasso回归和二元Logistic回归分析显示胸腔积液、胃肠道出血,较高水平的年龄、WBC、Hb,较低水平的TP是PHC患者干预前并发肺部感染的独立危险因素,据此建立的列线图模型受试者工作特征(ROC)曲线下面积为0.700(95%CI:0.659~0.740),Hosmer-Lemeshow检验结果示模型拟合优度较好,自抽样法重复抽样1000次进行内部验证,模型的一致性较好。结论胸腔积液,胃肠道出血,年龄较大,WBC、Hb水平较高以及较低水平的TP是PHC患者干预前并发肺部感染的独立危险因素,建立的列线图预测模型可有效评估PHC患者干预前发生肺部感染的风险。

Objective To analyze the influencing factors of pulmonary infection in patients with primary liver cancer(PHC)before intervention,and establish a nomogram risk prediction model to verify it.Methods A total of 1635 patients with PHC diagnosed and hospitalized for the first time were selected and divided into the infected group(197 cases)and the non-infected group(1438 cases)according to whether they had pulmonary infection.General data such as body mass index(BMI),chronic hepatitis B(CHB),chronic hepatitis C(CHC),Barcelona stage(BCLC),white blood cells(WBC),neutrophils(NEU),hemoglobin(Hb)and other blood routine indicators were collected.Total protein(TP),prealbumin(PA),aspartate aminotransferase(AST),gamma glutamylaminotransferase(GGT),alkaline phospholipase(ALP),abnormal prothrombin(PIVKA-Ⅱ),alpha-fetoprotein(AFP),carcinoembryonic antigen(CEA),procalcitonin(PCT),hypersensitive C-reactive protein(hs-CRP),cholinesterase(ChE),total cholesterol(TC)and other blood biochemical indicators,CD3 cell count(CD3^(+)),CD4 cell count(CD4^(+)),CD4/CD8 ratio(CD4^(+)/CD8^(+)),CD19 cell count(CD19^(+)),interleukin(IL)-2,interferon(IFN)-α,tumor necrosis factor(TNF)-αand other cytokines were also collected.Univariate analysis and Lasso regression were used to screen variables,and binary Logistic regression analysis was used to determine risk factors for pulmonary infection in PHC patients before intervention.The risk prediction model was established and evaluated.Results Compared with the non-infected group,age,smoking rate,CHC,pleural effusion,gastrointestinal hemorrhage,Child-Pugh grade C,BCLC Phase A/C/D ratio,WBC,NEU,AST,GGT,ALP,PIVKA-Ⅱ,AFP,CEA,PCT,hs-CRP,IL-2,IL-5,IL-6,IL-8,IL-10,IL-12,IFN-γand TNF-αlevels were higher in the infected group,and levels of BMI,CHB ratio,Hb,TP,PA,ChE,TC,CD3^(+),CD4^(+),CD4^(+)/CD8^(+),CD19^(+)and IFN-αwere lower(P<0.05).Lasso regression and binary Logistic regression analysis showed that pleural effusion,gastrointestinal hemorrhage,higher level of age,WBC,Hb and lower level of TP were independent risk factors for pulmonary infection in patients with PHC before intervention.The area under receiver operating curve(ROC)of the established nomogram model was 0.700(95%CI:0.659-0.740),and Hosmer-Lemeshow test results showed good goodness-fit of the model.Self-sampling was repeated 1000 times for internal verification.The consistency of the model was good.Conclusion Pleural effusion,gastrointestinal hemorrhage,higher level of age,WBC,Hb and lower level of TP are independent risk factors for pulmonary infection in PHC patients before intervention.The established nomogram prediction model can effectively evaluate the risk of pulmonary infection in PHC patients before intervention.