本文已被:浏览 178次 下载 177次
Received:November 24, 2025 Published Online:October 23, 2026
Received:November 24, 2025 Published Online:October 23, 2026
中文摘要: 目的 探讨囊腔型肺癌(LCCA)的临床及CT 特征与无进展生存期(PFS)的相关性并构建预后预测模型。方法 选取2017 年2 月至2024 年12 月在扬州大学附属医院及宝应县中医医院确诊并治疗的LCCA 患者共58例进行回顾性分析。通过单因素Cox回归分析筛选与PFS相关的因素,并针对小样本数据采用Firth惩罚回归构建PFS预测模型,与多因素Cox回归模型的性能进行比较,通过Bootstrap进行内部验证。利用校准度、临床决策曲线、生存曲线及时间依赖性受试者工作特征(ROC)曲线评价模型的临床应用价值。结果 单因素Cox回归分析显示,患者年龄、病灶长径、病灶边界有分叶或尖角、胸膜牵拉征及手术治疗与PFS显著相关(P<0.05)。多因素Cox 回归及Firth 惩罚回归模型结果均显示,胸膜牵拉征是LCCA 患者PFS 的独立危险因素(HR=8.90,95%CI:2.99~26.51,P<0.01;HR=18.61,95%CI:4.39~114.62,P<0.01),手术治疗是独立保护因素(HR=0.28,95%CI:0.12~0.64,P<0.01;HR=0.23,95%CI:0.04~0.95,P<0.05)。Firth 模型原始C-index 为0.881,校正后C-index 为0.873,高于Cox 模型的0.793。校准曲线显示模型预测概率与实际观察概率高度一致。决策曲线显示,在广泛风险阈值范围内该模型的临床净获益均优于全部干预或全部不干预策略。时间依赖性ROC证实模型在24、36和60个月的曲线下面积分别为0.85、0.82 和0.95。结论 对本研究小样本数据建立的Firth 惩罚回归预后模型与多因素Cox回归模型均识别出胸膜牵拉征和手术治疗为LCCA患者PFS的显著预测因子,但Firth方法更优。对于CT表现伴有囊腔的病灶,应进行随访,若出现进展或CT提示有胸膜牵拉征,应考虑手术治疗。
Abstract:Objective To explore the correlation between the clinical and CT characteristics of lung cancer associated with cystic airspaces(LCCA)and progression-free survival(PFS),and to construct a prognostic prediction model.Methods A total of 58 patients with LCCA who were diagnosed and treated from February 2017 to December 2024 at the Affiliated Hospital of Yangzhou University and Baoying Hospital of Traditional Chinese Medicine were selected for a retrospective analysis. Univariate Cox regression analysis was used to screen factors related to PFS,and for small sample data,Firth’s penalized regression was used to construct a PFS prediction model,with performance compared to the multivariate Cox regression model,and internal validation was conducted through Bootstrap. The clinical application value of the model was evaluated by calibration,clinical decision,survival and time-dependent receiver operating characteristic(ROC)curves. Results Univariate Cox regression analysis showed that patients 'age,lesion long diameter,lobulated or spiculated margins,pleural traction sign,and surgical treatment were significantly associated with PFS(P<0.05). Both multivariate Cox and Firth's penalized regression model showed that pleural traction sign was an independent risk factor for PFS(HR=8.90,95%CI:2.99-26.51,P<0.01;HR=18.61,95%CI:4.39-114.62,P<0.01),while surgical treatment was an independent protective factor(HR=0.28,95%CI:0.12-0.64,P<0.01;HR=0.23,95%CI:0.04-0.95,P<0.05). The original C-index of the Firth's model was 0.881,and the corrected C-index was 0.873,which was higher than the 0.793 of the Cox model. The calibration curve showed that the model's predicted probabilities were highly consistent with the actual observed probabilities. Decision curve analysis demonstrated a superior net clinical benefit over a wide threshold range compared to treat-all or treat-none strategies. Time-dependent ROC analysis further confirmed the model's high predictive accuracy at various time points,with area under curve(AUC)of 0.85 at 24 months,0.82 at 36 months,and 0.95 at 60 months. Conclusion Both the Firth's penalized regression prognostic model and the multivariate Cox model,established on the small sample data of this study,identify pleural traction sign and surgical treatment as significant predictors of PFS in LCCA patients,but the Firth’s method proves superior. For lesions with cystic airspaces on CT,follow-up should be conducted. If there is progression or pleural traction sign indicated by CT,surgical treatment should be considered.
keywords: Cystic airspaces Lung cancer CT features Progression-free survival Predictive model Firth's penalized regression
文章编号: 中图分类号:R734.2 R445.3 文献标志码:A
基金项目:
引用文本:
