探讨炎性复合指标对慢性阻塞性肺疾病(chronic obstructive pulmonary disease, COPD)并发心房颤动风险的预测价值。

选取江阴市人民医院和南京市胸科医院2019年1月至2024年12月收治的860例COPD患者为研究对象，按7︰3随机分为训练集600例及验证集260例，训练集按是否发生心房颤动分为心房颤动组128例、非心房颤动组472例，比较两组中性粒细胞与淋巴细胞比值(neutrophil-to-lymphocyte ratio, NLR)、血小板计数/淋巴细胞计数比值(platelet to lymphocyte ratio, PLR)、单核细胞/高密度脂蛋白胆固醇比值(monocyte to high density lipoprotein cholesterol ratio, MHR)、C反应蛋白/白蛋白比值(C-reactive Protein to albumin ratio, CAR)；采用Lasso回归筛选预测变量，构建列线图；通过受试者工作特征(receiver operating characteristic, ROC)曲线判断区分度，Hosmer Lemeshow检验判断校准度，决策曲线分析(decision curve analysis, DCA)判断临床净获益。

训练集中心房颤动组NLR、PLR、MHR、CAR高于非心房颤动组[(2.18±0.42)比(1.58±0.39)，(142.07±42.13)比(112.81±26.69)，(0.42±0.12)比(0.34±0.10)，(1.28±0.24)比(0.76±0.20)，P<0.001]。Lasso回归筛选出4个变量NLR、PLR、MHR、CAR构建列线图，各变量对应分值分别为12.11、14.51、6.58、44.44分。验证集ROC分析显示，AUC为0.902(95%CI：0.852~0.952)，灵敏度83.70%，特异度82.10%；校准曲线C指数为0.909，HosmerLemeshow检验χ²＝7.0874，df＝8，P＝0.5272；DCA显示，阈值0~0.8，模型净获益稳定>0.4，临床获益性高。

基于NLR、PLR、MHR、CAR构建的列线图对COPD并发心房颤动区分能力高、校准度良好和临床净获益显著，可为临床早期识别高危患者提供参考。

To investigate the predictive value of inflammatory composite indicators for the risk of atrial fibrillation in patients with chronic obstructive pulmonary disease (COPD).

A total of 860 patients with COPD admitted to Jiangyin People′s Hospital and Nanjing Chest Hospital from January 2019 to December 2024 were enrolled and randomly divided into a training set 600 cases and a validation set 260 cases at a 7︰3 ratio. In the training set, patients were categorized into an AF group 128 cases and a non atrial fibrillation group 472 cases according to the presence of atrial fibrillation. The neutrophiltolymphocyte ratio (NLR), platelettolymphocyte ratio (PLR), monocytetohighdensity lipoprotein cholesterol ratio (MHR), and Creactive proteintoalbumin ratio (CAR) were compared between the two groups. Lasso regression was used to select predictive variables for constructing a nomogram. The discriminative ability was assessed by receiver operating characteristic (ROC) curves, calibration was evaluated by the HosmerLemeshow test, and clinical net benefit was determined by decision curve analysis (DCA).

In the training set, the atrial fibrillation group had significantly higher NLR, PLR, MHR, and CAR than the non atrial fibrillation group [(2.18±0.42) vs. (1.58±0.39), (142.07±42.13) vs. (112.81±26.69), (0.42±0.12) vs. (0.34±0.10), (1.28±0.24) vs. (0.76±0.20), all P<0.001]. Lasso regression identified the four variables (NLR, PLR, MHR, CAR) for nomogram construction, with corresponding points of 12.11, 14.51, 6.58, and 44.44, respectively. In the validation set, ROC analysis showed an AUC of 0.902 (95%CI: 0.852~0.952), with a sensitivity of 83.70% and specificity of 82.10%. The calibration curve had a Cindex of 0.909, and the HosmerLemeshow test gave χ²=7.0874, df=8, P=0.5272. DCA demonstrated that the model′s net benefit remained stable above 0.4 across threshold probabilities of 0~0.8, indicating high clinical utility.

The nomogram based on NLR, PLR, MHR, and CAR shows good discriminative ability, satisfactory calibration, and significant clinical net benefit for predicting atrial fibrillation in COPD patients, providing a useful tool for early identification of highrisk individuals in clinical practice.