【摘要】 目的 基于最小绝对收缩和选择算子(LASSO)回归分析构建并验证微波消融术(MWA)术后肺癌患者气胸发生风险的预测模型，为其临床诊治提供参考。方法 回顾性分析2017-01-07-2024-12-28滕州市中心人民医院接受CT引导下MWA治疗的230例肺癌患者临床资料，按7∶3比例随机分为训练集(161例)和测试集(69例)。筛选出19个临床变量，进行Spearman相关分析发现部分变量之间的相关系数超过0.6。为解决多重共线性问题，采用LASSO回归在训练集中筛选出关键预测因素，并通过多因素logistic回归构建预测模型，绘制列线图。模型的预测性能及临床适用性通过受试者工作特征(ROC)曲线、校准曲线、决策曲线分析和临床影响曲线进行评估。结果 基于LASSO回归分析，最终筛选出肺气肿病史(OR=2.762,95%CI:1.079～7.342,P=0.037)、肿瘤大小(OR=1.518,95%CI:0.985～2.407,P=0.046)、消融同步活检(OR=5.147,95%CI:2.061～13.822,P=0.001)、穿刺过胸膜次数(OR=2.721,95%CI:1.385～5.641,P=0.005)及靶胸膜距离(OR=0.448,95%CI:0.294～0.657,P<0.001)5项预测因子，并据此构建气胸风险预测列线图模型。模型在训练集和测试集中曲线下面积分别为0.824(95%CI:0.762～0.887)和0.825(95%CI:0.730～0.920)。校准曲线分析显示，模型预测的气胸发生率与实际发生率高度一致。Hosmer-Lemeshow拟合优度检验结果显示，模型预测值与实际观测值之间差异无统计学意义，χ~2=4.671,P=0.792。决策曲线分析和临床影响曲线进一步验证了模型在不同风险阈值下的临床应用价值。结论 合并肺气肿、消融同步活检、肿瘤直径大、穿刺过胸膜次数多及靶病灶距胸膜近均为肺癌MWA术后气胸的独立危险因素；基于这5个预测因素构建的气胸风险预测模型具有良好的预测准确性、稳定性及泛化能力。更多还原

【Abstract】 Objective To develop and validate a least absolute shrinkage and selection operator(LASSO) regression-based nomogram for predicting pneumothorax following microwave ablation(MWA),to provide reference for its clinical diagnosis and treatment.Methods A retrospective analysis was performed on the clinical data of 230 lung cancer patients who underwent CT-guided MWA at Tengzhou Central People’s Hospital between January 7,2017 and December 28,2024.These patients were randomly assigned in a 7:3 ratio to a training cohort(n=161) and a test cohort(n=69).Nineteen clinical variables were identified,and Spearman correlation analysis revealed correlation coefficients exceeding 0.6 among several variables.To address multicollinearity,LASSO regression was used to select key predictive factors,followed by multivariable logistic regression to construct a predictive model visualized as a nomogram.The model’s performance and clinical utility were evaluated by using receiver operating characteristic curves,calibration curves,decision curve analysis,and clinical impact curves.Results Five key predictive factors were identified by using the LASSO regression model:emphysema(OR=2.762,95% CI:1.079-7.342,P=0.037),tumor size(OR=1.518,95 % CI:0.985-2.407,P=0.046),synchronous biopsy during ablation(OR=5.147,95% CI:2.061-13.822,P=0.001),number of pleural needle passes(OR=2.721,95%CI:1.385-5.641,P=0.005),and tumor-pleura distance(OR=0.448,95% CI:0.294-0.657,P<0.001).A nomogram was then constructed based on these five risk factors.The area under the curve of the model was 0.824(95 % CI:0.762-0.887) in the training cohort and 0.825(95 % CI:0.730-0.920) in the test cohort,demonstrating stable and consistent predictive performance across both groups.Calibration curve analysis revealed high concordance between the predicted and actual incidence of pneumothorax.The Hosmer-Lemeshow goodnessof-fit test yielded a χ~2 value of 4.671(P=0.792),indicating no significant difference between the predicted and observed values.Decision curve analysis and clinical impact curve analysis further confirmed the clinical utility of the model across various risk thresholds.Conclusion Emphysema,synchronous biopsy during ablation,larger tumor size,a greater number of pleural passes and a shorter tumor-pleura distance emerged as independent risk factors for pneumothorax after microwave ablation of lung cancer;the LASSO regression model built on these variables displayed excellent discriminatory power,robustness and potential for generalization.更多还原