Journal of Hebei Medical University ›› 2024, Vol. 45 ›› Issue (3): 278-283.doi: 10.3969/j.issn.1007-3205.2024.03.005

Previous Articles     Next Articles

Establishment and validation of probability prediction model for benign and malignant pulmonary nodules

  

  1. 1.Department of Respiratory and Critical Care Medicine, the Affiliated Hospital of North Sichuan 
    Medical College, Nanchong 637000, China; 2.Department of Respiratory and 
    Critical Care Medicine, Guang′an People′s Hospital, 
    Sichuan Province, Guang′an 638001, China

  • Online:2024-03-25 Published:2024-04-07

Abstract: Objective To analyze the clinical characteristics and imaging manifestations of pulmonary nodules, to screen the factors that affect the malignancy of pulmonary nodules, and to establish and validate prediction models, thus providing a reference basis for the differentiation of benign and malignant pulmonary nodules. 
Methods A total of 1 160 patients with pulmonary nodules were selected as the research subjects, and then randomly divided into two groups at a ratio of 2〖DK〗∶1: a modeling group (n=773) and a validation group (n=387). Univariate analysis was conducted on the data of modeling group, and variables with statistically significant differences were included in binary logistic regression analysis to obtain independent predictive factors for benign and malignant pulmonary nodules, and to establish a probability prediction model for benign and malignant lesions. The predictive performance of this model with traditional classic models such as Mayo model, Brock model, VA model, and model of Peking University People′s Hospital was verified and compared. 
Results Age, gender, family history of lung cancer, nodule texture, nodule diameter, lobulation sign, vacuolar sign, vascular bundle sign, calcification sign, bronchiolitis sign, and pleural traction sign were independent predictors of benign and malignant pulmonary nodules (P<0.05). The area under the receiver operating characteristic (ROC) curve (AUC) of this model (0.856) was higher than that of the Mayo model (0.604), Brock model (0.447), VA model (0.569), and model of Peking University People′s Hospital (0.677), with a predictive sensitivity of 86.10% and a specificity of 73.70%. 
Conclusion The prediction model constructed in this study is with good diagnostic efficacy, which may be superior to the traditional models, which has greater reference value for clinicians to distinguish the benign and malignant pulmonary nodules. 


Key words: multiple pulmonary nodules, lung neoplasms, prediction model