心肺运动试验在肺癌患者术后肺部并发症列线图预后预测模型建立中的应用

关键词：非小细胞肺癌  术后肺部并发症  心肺运动试验  列线图

Keywords:non-small cell lung cancer  postoperative pulmonary complications  cardiopulmonary exercise test  nomogram

摘要：

      摘要 目的：将心肺运动试验（CPET）与临床指标结合以个体化预测肺癌患者胸腔镜肺切除（VATS）术后肺部并发症（PPCs）发生的影响因素，并绘制列线图模型，使预测结果更加直观、可视化。 方法：回顾性分析我院胸外科行VATS的肺癌患者相关数据，通过单因素和多因素Logistic逐步分析法找出PPCs发生的独立影响因素，绘制列线图预测模型，内部验证后绘制校准曲线，作该模型的受试者工作特征曲线（ROC曲线），并计算曲线下面积（AUC）及95%置信区间（CI）。 结果：共纳入患者168例，其中并发症组45例（26.8%），非并发症组123例（73.2%）。单因素分析显示：PPCs的发生与年龄、肿瘤分期、吸烟史、术中出血量、冠心病、VO2peak、VE/VCO2 slope、AT、PetCO2显著相关（P<0.05），与其余因素无关（P>0.05），且并发症组术后住院时间明显超过非并发症组（P<0.05）。多因素分析示：年龄（OR=6.51，95%CI：1.89—22.45，P<0.05）、术中出血量（OR=5.16，95%CI：0.93—1.00，P<0.05）和VE/VCO2 slope（OR=0.96，95%CI：1.64—16.25，P<0.05）是PPCs发生的独立影响因素。构建预测PPCs发生的列线图模型后，再利用ROC曲线进行分析列线图预测模型的区分度，结果示曲线下面积为0.792（95%CI：0.71—0.87），对PPCs预测的灵敏度为76.4%，特异度为77.8%，区分能力较好。Hosmer-Lemeshow χ2=11.595，P=0.170，提示模型预测值与实际观测值之间的差异没有显著性，预测模型有较好的校准能力。 结论：CPET与临床指标相结合，绘制的列线图预测模型，具有良好的精准度，可以更加直观、客观和个体化评估肺癌患者PPCs发生的风险。

Abstract:

      Abstract Objective：To predict the influencing factors of postoperative pulmonary complications（PPCs） in patients with lung cancer after video-assisted thoracic surgery by using the cardiopulmonary exercise test（CPET） combined with clinical indicators，and to draw a nomogram model to make the prediction results more intuitive and visualized. Method：The relevant data in patients with lung cancer who underwent video-assisted thoracic surgery in the thoracic surgery department of our hospital was retrospectively collected. Single factor analysis and multivariate logistic stepwise analysis were used to find out the independent influencing factors of PPCs. The nomograph model of PPCs risk was drawn and internal verification was done to get the calibration curve. Finally， the area under the curve （AUC） and 95% confidence interval （CI） were calculated by the receiver operating characteristic （ROC） curve. Result：There were 168 patients in total， including 45 patients in the complication group （26.8%） and 123 patients in the non-complication group （73.2%）. Unifactor analysis results showed that the occurrence of PPCs was significantly correlated with age， stage， smoking history， intraoperative blood loss， coronary heart disease， VO2peak， VE/VCO2 slope， AT and PetCO2（P<0.05）， but not with other factors （P>0.05）， and the length of postoperative hospital stay in the complication group was significantly higher than that in the non-complication group （P<0.05）. Multivariate analysis results showed that age （OR=6.51， 95%CI： 1.89—22.45， P<0.05）， intraoperative blood loss （OR=5.16， 95%CI： 0.93—1.00， P<0.05） and VE/VCO2 slope （OR=0.96， 95%CI： 1.64—16.25， P<0.05） were independent influencing factors for PPCs occurrence in VATS pneumonectomy. After constructing the nomogram model to predict the occurrence of PPCs， the ROC curve was used to analyze the discrimination of the nomogram prediction model. The results of the nomogram model with the ROC curve showed that AUC was 0.792（95%CI：0.71—0.87）， and the sensitivity and specificity of PPCs prediction were 76.4% and 77.8%. Hosmer-lemesho χ2=11.595， P=0.170， indicating that the difference between the predicted value of the model and the actual observed value was not statistically significant， and the prediction model had good calibration ability. Conclusion：The CPET combined with clinical indicators to make the prediction of PPCs more comprehensive and reliable. The nomogram model can be more intuitive， objective and individual to assess the risk of PPCs in lung cancer patients.