实性结节型早期肺癌相关因素分析及模型比较

Analysis of related factors and model comparison of early solid nodular lung cancer

目的:探讨实性结节型早期肺癌的相关因素,建立良恶性预测模型。方法:回顾性分析经手术病理确诊的241例实性肺结节患者的临床及CT影像学资料,其中恶性结节160例作为研究组,良性结节81例纳为对照组。通过单因素分析比较两组的临床资料(年龄、性别、吸烟史、恶性肿瘤家族史、肿瘤标记物)和影像学征象(位置、大小、CT值、边缘征象、内部特征),将组间差异有统计学意义的指标纳入Logistic回归分析,建立预测模型,绘制受试者操作特征(ROC)曲线检验其诊断效能,并与北大模型比较。结果:右肺、肿瘤标记物升高、锯齿征、分叶-毛刺-胸膜凹陷征、直径、CT值、钙化、大片坏死、结节周围粟粒征作为实性肺结节良恶性预测因子,建立恶性概率预测模型为P=e^(x)/(1+e^(x)),e为自然对数,X=-3.107+1.066×右肺+1.540×肿瘤标记物升高+1.593×锯齿征+1.383×分毛凹+0.096×直径+0.016×CT值-2.697×钙化-3.058×大片坏死-1.713×结节周围粟粒征,其预测实性结节恶性概率的敏感度为89.4%,特异度为85.2%,阳性预测值为85.8%,阴性预测值为88.9%,符合率为87.3%,ROC曲线下面积(AUC)为0.934(95%CI:0.902~0.966)。将本组数据代入北大模型预测实性结节恶性概率的敏感度为76.3%,特异度为65.4%,阳性预测值为68.8%,阴性预测值为73.4%,符合率为70.9%,AUC为0.770(95%CI:0.706~0.833)。结论:本研究模型诊断效能高于北大模型,有助于提高良恶性肺结节术前预测的准确率。

Objective:To investigate the predictors of early solid nodular lung cancer and establish a benign and malignant prediction model.Methods:Clinical data and CT imaging data of 241 patients with solid pulmonary nodules confirmed by surgery and pathology were retrospectively analyzed.160 patients with malignant nodules were classified as study group and 81 patients with benign nodules were classified as control group.Clinical data(age,sex,smoking history,family history of malignancy,tumor markers)and imaging characteristics(location,size,CT value,marginal signs,internal features)were compared between the two groups by univariate analysis.The indicators with statistically significant differences between groups were included in multivariate logistic regression analysis,the prediction model was established,and receiver operating characteristic(ROC)curve was drawn to test its diagnostic efficiency,and the model was compared with the Peking University model.Results:The benign and malignant predictors of solid pulmonary nodules were found in the right lung,elevated tumor markers,serration sign,lobulated-burr-pleural indentation,diameter,CT values,calcification,massive necrosis within nodules,and miliary satellite lesions around nodules.The malignant probability prediction model was set up as P=e^(x)/(1+e^(x))and e was the natural logarithm,X=-3.107+1.066×right lung+1.540×increased tumor marker+1.593×serrated sign+1.383×lobulation-spiculation-pleural indentation+0.096×diameter+0.016×CT value-2.697×calcification-3.058×massive necrosis-1.713×miliary lesions around nodules.The sensitivity,specificity,positive predictive value and negative predictive value were 89.4%,85.2%,85.8%,88.9%,and 87.3%respectively.The area under ROC curve(AUC)was 0.934(95%CI:0.902~0.966).This set of data was substituted into the Peking University model.The sensitivity,specificity,positive predictive value and negative predictive value were 76.3%,65.4%,68.8%,73.4%,and 70.9%respectively.The area under ROC curve(AUC)was 0.770(95%CI:0.706~0.833).Conclusion:The diagnostic efficacy of this model(0.934)was higher than that of the PKU model(0.770).It is helpful to improve the accuracy of preoperative prediction of benign and malignant solid pulmonary nodules.