Liu Suli, Yang Yao, Wang Dongmei, Gao Lijuan, Qin Jiangyue, Wu Yanqiu, Li Diandian, Li Xiaohua, Chen Mei, Wang Hao, Shen Yongchun, Wen Fuqiang, Chen Fangying
Division of Pulmonary Diseases, State Key Laboratory of Biotherapy of China, Department of Respiratory and Critical Care Medicine, West China Hospital, Sichuan University, Chengdu, China.
State Key Laboratory of Respiratory Health and Multimorbidity, West China Hospital, Sichuan University, Chengdu, China.
Front Med (Lausanne). 2025 Jul 18;12:1589406. doi: 10.3389/fmed.2025.1589406. eCollection 2025.
Diagnosing tuberculous pleural effusion (TPE) is challenging. There is a lack of cross-sectional lateral comparisons among TPE prediction models.
We aimed to develop and validate a novel TPE prediction model and compare its diagnostic performance with that of existing models.
Patients with pleural effusion were included in the training, testing, and external validation sets. Variable selection strategies included LASSO and logistic regression. The discriminability, calibration, and clinical efficacy of the prediction model were estimated in the three sets. The performance of the model was compared with that of two existing prediction models.
Fever, tuberculosis interferon-gamma release assays, pleural adenosine deaminase, the pleural mononuclear cell ratio, the ratio of pleural lactate dehydrogenase to pleural adenosine deaminase, pleural carcinoembryonic antigen, and pleural cytokeratin 19 fragment were selected to establish the prediction model. The AUCs were 0.931 (0.903-0.958), 0.856 (0.753-0.959), and 0.925 (0.867-0.984) in the training, testing, and external validation sets, respectively. The AUCs of the two existing prediction models were 0.793 (0.737-0.850) and 0.854 (0.816-0.892). The calibration curves revealed that this model had good consistency. Decision curve analysis revealed the acceptable clinical benefit of this model.
Compared with the existing models, the TPE prediction model developed in this study demonstrated good diagnostic performance.
诊断结核性胸腔积液（TPE）具有挑战性。目前缺乏对TPE预测模型的横断面横向比较。
我们旨在开发并验证一种新型TPE预测模型，并将其诊断性能与现有模型进行比较。
将胸腔积液患者纳入训练集、测试集和外部验证集。变量选择策略包括LASSO和逻辑回归。在这三个数据集中评估预测模型的辨别力、校准度和临床疗效。将该模型的性能与两个现有预测模型的性能进行比较。
选择发热、结核干扰素-γ释放试验、胸腔腺苷脱氨酶、胸腔单核细胞比例、胸腔乳酸脱氢酶与胸腔腺苷脱氨酶的比值、胸腔癌胚抗原和胸腔细胞角蛋白19片段来建立预测模型。训练集、测试集和外部验证集的曲线下面积（AUC）分别为0.931（0.903 - 0.958）、0.856（0.753 - 0.959）和0.925（0.867 - 0.984）。两个现有预测模型的AUC分别为0.793（0.737 - 0.850）和0.854（0.816 - 0.892）。校准曲线显示该模型具有良好的一致性。决策曲线分析显示该模型具有可接受的临床效益。
与现有模型相比，本研究开发的TPE预测模型具有良好的诊断性能。
