肺部恶性小结节危险因素分析及恶性预测模型建立

目的 分析肺部小结节恶性病变的危险因素,并建立恶性病变的数学预测模型,探讨有效检出恶性病变的理论依据.方法 回顾性分析2017年1月至2017年6月南京市胸科医院胸外科手术治疗的100例肺部小结节患者的临床资料.收集患者的性别、年龄、肿瘤指标、以及胸部CT影像学特征等资料并进行单因素分析.采用多因素logistic回归分析筛选肺部恶性小结节的危险因素,并建立相应的数学预测模型.结果 多因素logistic回归分析提示病灶位置(OR=4.218,P=0.042)、病灶类型为磨玻璃结节(GGN)(OR=24.625,P=0.000)、病灶类型为部分实性结节(PSN)(OR=6.228,P=0.052)、血管穿行征(OR=10.646,P=0.036)、分叶征(OR=18.162,P=0.027)及毛刺征(OR=8.054,P=0.018)为肺部恶性小结节的独立危险因素.恶性预测值(P)=ez/(1+ez),Z=-2.761+(3.204×GGN)+(1.829×PSN)+(1.439×位置)+(2.086×毛刺征)+(2.899×分叶征)+(2.365×血管穿行征).根据ROC曲线,选取预测概率0.64作为良恶性判断的临界值,模型预测的准确率为87.0%,敏感性为97.4%,特异性为54.2%,阳性预测值87.1%,阴性预测值为86.7%.结论 肺部病灶位于上叶,CT影像特征表现为GGN、PSN、血管穿行征、毛刺征和分叶征的小结节,恶性概率较高.通过logistic回归建立的数学预测模型预测肺部小结节恶性概率具有较高的准确性.

Analysis of risk factors of malignant pulmonary nodulesand establishment of malignancy prediction model

Objective To analyze the risk factors of malignant small pulmonary nodules and establish the mathematical prediction model of malignant lesions,so as to provide theoretical basis for effective detection of malignant lesions.Methods The clinical data of 100 patients with small pulmonary nodules,underwent thoracic operation in our hospital from Jan 2017 to Jun 2017,were retrospectively analyzed. Univariate analysis of such indexes as gender,age,tumor markers,and chest CT image features of all patients were performed.Multivariate logistic regression analysis was used to screen the risk factors of pulmonary malignant nodules, and with it the corresponding mathematical prediction model was established.Results Multivariatelogistic regression analysis showed that lesion location(OR=4.218;P=0.042), ground-glass nodule(GGN)of lesion type(OR=24.625;P=0.000), partial solid nodule(PSN)of lesion type(OR=6.228;P=0. 052),vascular breakthrough sign(OR=10.646;P=0.036), lobulation sign(OR=18.162;P=0.027)and spiculation sign(OR=8. 054;P=0.018)were independent risk factors formalignant pulmonary nodules.And their malignancy predicted value(P)was ez/(1+ez), whereas×Z×equaled to-2.761+(3.204×GGN)+(1.829×PSN)+(1.439×position)+(2.086×spiculation sign)+(2.899×lobu-lation sign)+(2.365×vascular breakthrough sign).Based on the ROC curve,predictive probability of 0.64 was chosen as the critical value of benign neoplasm and malignancy judgment,with accuracy rate of model prediction 87%,sensitivity of 97.4%,specificity of 54.2%,and positive predicted value of 87.1%and negative predicted value of 86.7%.Conclusion If those small pulmonary lesions locate in the upper lobe and the CT images are characterized by GGN,PSN,and nodules with vascular breakthrough sign, spiculation and lobulation signs,the probability of malignancy would be higher.The mathematical prediction model established by logistic regression may have higher accuracy in the prediction of malignant probability of small pulmonary nodules.