急性放射性肺炎早期鉴别诊断多参数模型建立  

Establishing an early differential diagnosis model of acute radiation pneumonitis based on multiple parameters

作　　者：王志武[1] 董量 吴琼[1] 王静怡[1] 付浩宇[1] 李玉凤[2] 杨俊泉[1] WANG Zhi-wu;DONG Liang;WU Qiong;WANG Jing-yi;FU Hao-yu;LI Yu-feng;YANG Jun-quan(Tangshan People's Hospital,Tangshan 063000,P.R.China)

机构地区：[1]唐山市人民医院放化疗科,河北唐山063000 [2]唐山市人民医院肿瘤研究所,河北唐山063000

出　　处：《中华肿瘤防治杂志》2020年第8期636-641,共6页Chinese Journal of Cancer Prevention and Treatment

基　　金：国家自然科学基金青年项目(81603475)。

摘　　要：目的急性放射性肺炎(acute radiation pneumonitis,ARP)与感染性肺炎(infectious pneumonia,IP)的早期鉴别诊断是临床难点问题,基于单因素的鉴别方法效果不佳,本研究尝试基于多参数建立ARP和IP的早期鉴别诊断模型。方法回顾性研究2014-01-01-2016-12-31唐山市人民医院放化疗科135例肺癌放疗后6个月内发生肺炎患者,经过病原学检查结果及其他多种方法明确最终诊断为ARP或IP。收集放射物理参数、影像学特征和血清学指标等多种信息,应用Logistics回归单因素分析,Spearman相关分析筛选变量,并应用Logistics回归多因素分析拟合回归模型,应用R语言实现模型的可视化并进行内验证。结果135例患者中,ARP 82例,IP 53例。单因素分析显示,病变位置与V20关系(positional relation between lesions and V20 area,PRLV)、降钙素原(procalcitonin,PCT)、C反应蛋白(C reactive protein,CRP)、平均肺受量(mean lung dose,MLD)和V20(lung volume receiving≥20 Gy)与ARP和IP有关联(P<0.05)。排除共线性指标后多个相关性指标纳入多因素分析,结果显示,PRLV(P<0.001)、PCT(P=0.017)、MLD(P=0.020)为独立的诊断因素,白细胞计数(white blood cell count,WBC)与ARP具有边缘相关性(P=0.051)。基于PRLV、PCT、MLD和WBC 4个指标拟合回归模型,得到可视化列线表图形。ROC曲线分析显示,模型的AUC=0.849,明显优于各个单一指标(PRLV:0.734;PCT:0.741;MLD:0.670;WBC:0.580),且模型显示出了较高的敏感性(91.46%)和特异性(66.0%),校准曲线显示模型诊断情况与实际诊断情况呈高度一致性。DCA曲线显示,模型据有满意的阳性净获益。结论PRLV、PCT、MLD和WBC成功构建了诊断模型并实现了可视化,该诊断模型对于ARP和IP的鉴别诊断具有一定的价值。OBJECTIVE The early differential diagnosis of acute radiation pneumonitis(ARP)and infection pneumonia(IP)is a difficult issue in clinical practice.The method based on single factor was not effective.The object of this study was to establish an early differential diagnosis model of ARP and IP based on multiple parameters.METHODS Lung cancer patients who admitted for pulmonary inflammatory disease at the Department of Radiotherapy and Chemotherapy in Tangshan People’s Hospital from January 1,2014 to December 31,2016 were retrospectively included.All the patients were diagnosed as ARP or IP by pathogen examination results and other methods.Radiological physical parameters,imaging features,serological indicators and other information were collected.Univariate logistic regression and spearman correlation analysis were used to select variables,and multivariate logistic regression was used to fit regression models.R language software was used to develop nomogram and perform internal verification.RESULTS A total of 135 patients with pulmonary inflammatory disease occurred within half a year after radiotherapy were included,including 82 cases of ARP and 53 cases of IP.Univariate analysis showed that the positional relation between lesion location and V20(PRLV),Procalcitonin(PCT),C reactive protein(CRP),Mean lung dose(MLD)and lung volume receiving≥20 Gy(V20)were associated with RP(P<0.05).After excluding the collinear factors,multiple factors were included in the multivariate analysis.The results showed that PRLV(P<0.001),PCT(P=0.017)and MLD(P=0.020)were independent diagnostic factors of ARP,and WBC was marginally related to RP(P=0.051).The regression model was fitted based on the four indexes of PRLV,PCT,MLD and WBC,and the nomogram of the modal was created.ROC curve analysis showed that the model had an AUC value of 0.849,which was significantly better than each single factor(PRLV:0.734;PCT:0.741;MLD:0.670;WBC:0.580),and the model showed higher sensitivity and specificity(91.46%and 66.0%),the calibration curve showed

关 键 词：放射治疗 放射性肺炎 感染性肺炎 鉴别诊断 诊断模型 

分 类 号：R730.55[医药卫生—肿瘤] R563.1[医药卫生—临床医学]