文献速递慢阻肺肺结节与肺癌

整合定量CT生物标志物提升肺癌筛查中慢阻肺检出

Integrating Quantitative CT Scan Biomarkers to Enhance COPD Detection in the Holistic Implementation Study Assessing a Northern German Interdisciplinary Lung Cancer Screening Effort Lung Cancer Screening Program.
PubMed 文献 · Chest · 2026-08-14

在5014例接受肺癌筛查且有肺功能数据的成人中,既往未诊断慢阻肺患病率为22.2%。基于低剂量CT的肺气肿、气道壁厚度、气道分支数及临床特征构建的机器学习模型可显著提升慢阻肺检出效率,并减少需确认性肺功能检查的比例。

要点

  • 肺癌筛查人群既往未诊断慢阻肺患病率达22.2%。
  • 单纯肺气肿阈值5.1%检出慢阻肺的AUC为0.69。
  • 整合模型AUC提升至0.83,需确认性肺功能比例降至34%。

核心结论

  • 整合CT气道生物标志物与临床特征可更高效、有针对性地在肺癌筛查中检出慢阻肺。

研究设计与样本

  • 在成人肺癌筛查者中进行肺功能检查以识别既往未诊断慢阻肺,定义为目前吸烟者及吸烟≥10包年的既往吸烟者中FEV1/FVC<0.70。
  • 使用人工智能软件量化低剂量CT肺气肿范围、标准化气道壁厚度(Pi10的平方根壁面积)和气道分支数。
  • 结合临床特征(主要为吸烟史和呼吸困难)构建极限梯度提升(XGBoost)集成树机器学习模型。
  • 另采用FEV1/FVC正常下限定义进行敏感性分析。

主要发现

  • 5014例有肺功能数据的筛查患者中,1115例既往未诊断慢阻肺,患病率22.2%。
  • 在无已知气道疾病患者中,肺气肿单独在优化阈值5.1%时检出慢阻肺的AUC为0.69(95%CI 0.67-0.72),准确率66%,阳性预测值44%。
  • 41%的患者超过该阈值并符合确认性肺功能检查标准。
  • 整合模型(肺气肿、标准化Pi10平方根壁面积、气道分支数及临床特征)显著提升检出:AUC 0.83(95%CI 0.80-0.86),准确率78%,阳性预测值59%,需确认性肺功能检查比例降至34%。
  • 采用正常下限定义敏感性分析时,最佳模型AUC 0.86(95%CI 0.83-0.89),准确率84%,阳性预测值53%,转诊确认性肺功能检查比例22%。

关键数值与推荐

  • 慢阻肺定义:FEV1/FVC<0.70(目前吸烟者及吸烟≥10包年的既往吸烟者)。
  • 肺气肿优化阈值为5.1%。
  • 单纯肺气肿模型AUC 0.69(95%CI 0.67-0.72),准确率66%,阳性预测值44%。
  • 整合模型AUC 0.83(95%CI 0.80-0.86),准确率78%,阳性预测值59%。
  • 正常下限定义敏感性分析最佳模型AUC 0.86(95%CI 0.83-0.89),准确率84%,阳性预测值53%。

临床怎么用

  • 在肺癌筛查中可利用低剂量CT定量肺气肿、气道壁厚度和分支数辅助识别慢阻肺高危者。
  • 整合模型可减少不必要的确认性肺功能检查转诊比例。
  • 对超过肺气肿阈值或模型高概率者应行肺功能检查确认。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

BACKGROUND: Lung cancer screening offers an opportunity to enhance COPD detection among adults exposed to tobacco smoke; however, guidance on CT scan-based referral for spirometry is limited. RESEARCH QUESTION: Which low-dose CT scan emphysema threshold best identifies previously undiagnosed COPD in lung cancer screening patients, and does the addition of quantitative airway biomarkers enhance detection? STUDY DESIGN AND METHODS: In adults undergoing lung cancer screening, we performed spirometry to identify previously undiagnosed COPD, defined by airflow obstruction (FEV1/FVC < 0.70) in patients who currently smoke and patients who formerly smoked with ≥ 10 pack-years. We quantified emphysema extent, airway wall thickness (standardized square root wall area of airways with a theoretical internal perimeter of 10 mm), and airway branch count on low-dose CT scan using artificial intelligence-based software and combined these measures with clinical characteristics, mainly smoking history and dyspnea, to develop an ensemble tree-based machine-learning model (Extreme Gradient Boosting) for COPD detection. A sensitivity analysis using the lower limit of normal definition for FEV1/FVC was additionally performed. RESULTS: Among 5,014 screening patients with available spirometry, 1,115 had previously undiagnosed COPD, corresponding to a prevalence of 22.2%. In patients without known airway disease, emphysema alone at an optimized threshold of 5.1% showed moderate performance for COPD detection (area under the receiver operating characteristic curve [AUC], 0.69; 95% CI, 0.67-0.72; accuracy, 66%; positive predictive value [PPV], 44%), where 41% of patients exceeded this threshold and met criteria for confirmatory spirometry. An integrated model combining emphysema, standardized square root wall area of airways with a theoretical internal perimeter of 10 mm, airway branch count, and clinical characteristics significantly improved detection (AUC, 0.83; 95% CI, 0.80-0.86; accuracy, 78%; PPV, 59%) while reducing the proportion requiring confirmatory spirometry to 34%. In a sensitivity analysis using the lower limit of normal definition of COPD, the best-performing model achieved an AUC of 0.86 (95% CI, 0.83-0.89), accuracy of 84%, and PPV of 53%, while referring 22% for confirmatory spirometry. INTERPRETATION: Our results show that an integrated approach combining CT scan-derived airway biomarkers and clinical characteristics enables more efficient and targeted COPD detection within lung cancer screening programs. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov; No.: NCT04913155; URL: www. CLINICALTRIALS: gov.

来源:Chest,查看原文。版权归原作者所有。

引用

  1. Abdo M, Pott H, Reck M, et al. Integrating Quantitative CT Scan Biomarkers to Enhance COPD Detection in the Holistic Implementation Study Assessing a Northern German Interdisciplinary Lung Cancer Screening Effort Lung Cancer Screening Program. Chest. 2026. PMID: 42600771. doi:10.1016/j.chest.2026.07.5237

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。