解剖性肺切除术后持续漏气预测模型构建

Establishment of A Clinical Prediction Model of Prolonged Air Leak after Anatomic Lung Resection

背景与目的解剖性肺切除术后持续漏气(prolonged air leak,PAL)是胸外科常见并发症,重在准确预测及时预防,但目前国内尚缺少有效的预测模型,本研究旨在建立解剖性肺切除术后PAL临床预测模型。方法回顾分析2016年1月-2016年10月安徽医科大学附属省立医院胸外科解剖性肺切除术患者的临床资料和术后漏气情况,其中A组病例359例,通过对患者的年龄(岁)、性别、身体质量指数(body mass index,BMI)、吸烟史、肺功能指数、手术方式(开放或腔镜,肺段、肺叶或其他,如支气管袖式或血管袖式)、手术切除肺叶位置、肺部病灶性质和胸腔粘连情况进行单因素及多因素分析,寻找解剖性肺切除术后PAL的独立预测因子,并建立临床预测模型。随后利用不同时期、不同治疗组完成的112例解剖肺切除患者作为B组,用于验证本模型的诊断效能,并绘制受试者工作特征(receiver operating characteristic curve,ROC)曲线。结果多因素Logistic回归分析筛选出BMI、性别、吸烟史、第一秒用力肺活量占用力肺活量的百分比(forced expiratory volume in one second,FEV1%)、胸腔粘连及是否上叶切除为解剖性肺切除患者术后PAL的独立预测因子。利用筛选出的预测因子建立的诊断模型ROC曲线下面积为0.886(95%CI:0.835-0.937),最佳临界值P=0.299,对应的诊断敏感性为78.5%,特异性为93.2%。结论本研究建立的预测模型能较准确的预测解剖性肺切除术后PAL的发生,对及时有效预防PAL发生有指导作用。

Background and objective Prolonged air leak （PAL） after anatomic lung resection is a common and challenging complication in thoracic surgery. No available clinical prediction model of PAL has been established in China. The aim of this study was to construct a model to identify patients at increased risk of PAL by using preoperative factors exclusively. Methods We retrospectively reviewed clinical data and PAL occurrence of patients after anatomic lung resection, in department of thoracic surgery, Anhui Provincial Hospital Affiliated to Anhui Medical University, from January 2016 to October 2016. 359 patients were in group A, clinical data including age, body mass index （BMI）, gender, smoking history, surgical methods, pulmonary function index, pleural adhesion, pathologic diagnosis, side and site of resected lung were analyzed. By using univariate and multivariate analysis, we found the independent predictors of PAL after anatomic lung resection and subsequently established a clinical prediction model. Then, another 112 patients （group B）, who underwent anatomic lung resection in different time by different team, were chosen to verify the accuracy of the prediction model. Receiver-operating characteristic （ROC） curve was constructed using the prediction model. Results Multivariate Logistic regression analysis was used to identify six clinical characteristics [BMI, gender, smoking history, forced expiratory volume in one second to forced vi- tal capacity ratio （FEV1%）, pleural adhesion, site of resection] as independent predictors of PAL after anatomic lung resection. The area under the ROC curve for our model was 0.886 （95%CI： 0.835-0.937）. The best predictive P value was 0.299 with sen- sitivityof78.S%andspecificityof93.2%.Conclusion Our prediction model could accurately identify occurrence risk of PAL in patients after anatomic lung resection, which might allow for more effective use ofintraoperative prophylactic strategies.