基于CT征象构建肺腺癌表皮生长因子受体基因突变状态的列线图

Construction of a Nomogram of Epidermal Growth Factor Receptor Gene Mutation Based on CT Features

目的基于CT征象构建肺腺癌表皮生长因子受体(EGFR)基因突变状态的列线图。方法回顾性分析本院经基因检测确定EGFR野生型肺腺癌患者30例,EGFR突变型患者37例的资料。由两名影像科医师评价肿瘤位置、位置分型、分叶征、毛刺征、胸膜凹陷征、空泡征、支气管充气征、坏死、钙化、肿瘤内血管、肿瘤周围血管侵犯、纵隔淋巴结肿大、胸膜腔积液、支气管截断、血管集束征、胸膜转移及远处转移17个CT征象。比较两组患者一般资料及上述CT征象的差异。采用单变量及多变量Logistic回归筛选预测EGFR突变型的影响因素,根据多变量分析结果绘制列线图。采用Bootstrap 1000次重复抽样对列线图进行内部验证,绘制受试者工作特征曲线(ROC)并计算曲线下面积(AUC),计算敏感度及特异度用于评价列线图的区分度。采用H-L拟合优度检验评价列线图的校准度。结果两组患者的性别、肿瘤位置、支气管充气征及淋巴结肿大的构成比差异均有统计学意义(P<0.05)。EGFR突变型组肿瘤大小(30.89±12.84)mm小于野生型组(40.07±22.18)mm,差异有统计学意义(t=2.118,P=0.038)。单变量Logistic回归分析结果表明患者性别、肿瘤大小、肿瘤位置、分叶征、支气管充气征、坏死、肿瘤内血管、纵隔淋巴结肿大及支气管截断是EGFR突变状态的影响因素,多变量Logistic分析结果表明患者性别、肿瘤位置、支气管充气征及肿瘤内血管是预测EGFR突变型的独立影响因素,比值比(OR)分别为0.27(95%CI:0.08~0.88)、0.15(95%CI:0.04~0.60)、6.18(95%CI:1.17~32.74)及0.20(95%CI:0.07~0.75)(P<0.05)。根据上述四个变量绘制的列线图经Bootstrap 1000重复抽样后计算AUC为0.805,敏感度为81.08%,特异度为70.00%。H-L拟合优度检验表明该列线图有较高校准度(χ2=3.549,P=0.616)。结论基于CT征象构建肺腺癌EGFR突变状态的列线图具有较高的区分度及校准度,具一定临床应用价值。

Objective To construct the nomogram of Epidermal growth factor receptor(EGFR)gene mutation based on CT features.Methods 30 cases of EGFR wild-type lung adenocarcinoma and 37 cases of EGFR mutation were analyzed retrospectively.17 CT features were evaluated by two radiologists,including tumor location,location classification,lobulation,spiculation,pleural retraction,bubble-like lucency,air bronchogram,necrosis,calcification,intratumoral vessels,peritumoral vascular involvement,mediastinal lymphadenopathy,pleural effusion,bronchial occlusion,vascular convergence,pleural metastasis and distant metastasis.To compare the general data and the difference of the above CT features of the two groups.Univariate and multivariate logistic regression were used to screen the influencing factors of EGFR mutation,and the nomogram was drawn according to the results of multivariate analysis.Bootstrap 1000 repeated sampling was used to verify the nomogram internally,ROC was drawn and AUC was calculated.Sensitivity and specificity were used to evaluate the differentiation of nomogram.H-L goodness of fit test was used to evaluate the calibration of nomogram.Results There were significant differences in gender,tumor location,air bronchogram and the proportion of lymphadenopathy between the two groups(all P<0.05).The tumor size of EGFR mutation group(30.89±12.84 mm)was smaller than that of wild type group(40.07±22.18 mm),the difference was statistically significant(t=2.118,P=0.038).The results of univariate logistic regression analysis showed that gender,tumor size,tumor location,lobulation,air bronchogram,necrosis,intratumoral vessels,mediastinal lymphadenopathy and bronchial occlusion,were the influencing factors of EGFR mutation.Multivariate logistic analysis showed that gender,tumor location,air bronchogram and intratumoral vessels were independent factors in predicting EGFR mutation.The Odds ratio was 0.27(95%CI:0.08-0.88),0.15(95%CI:0.04-0.60),6.18(95%CI:1.17-32.74)and 0.20(95%CI:0.07-0.75),respectively,all P<0.05.The AUC calculated by bootstrap 1000 was 0.805,the sensitivity was 81.08%,and the specificity was 70.00%.H-L goodness of fit test shows that the nomogram has a higher calibration(χ2=3.549,P=0.616).Conclusion The nomogram of EGFR mutation based on CT features has a high differentiation and calibration,and has a certain clinical application value.