Risk assessment of malignancy in solitary pulmonary nodules in lung computed tomography: a multivariable predictive model study. Issue 14 (20th July 2021)

• Record Type: Journal Article
• Authors: Liu, Hai-Yang; Zhao, Xing-Ru; Chi, Meng; Cheng, Xiang-Song; Wang, Zi-Qi; Xu, Zhi-Wei; Li, Yong-Li; Yang, Rui; Wu, Yong-Jun; Zhang, Xiao-Ju
• Editors: Wei, Pei-Fang
• Is Part Of: Chinese medical journal. Volume 134:Issue 14(2021)
• Page Start: 1687
• Page End: 1694
• Publication Date: 2021-07-20
• Subjects: CT image -- Lung cancer -- Prediction model -- Pulmonary nodules -- Regression algorithm; Medicine -- Periodicals; Medicine, Oriental -- Periodicals; Medicine; Medicine, Oriental; Medicine; Medicine, East Asian Traditional; Periodicals; Electronic journals; 610.5
• DOI: 10.1097/CM9.0000000000001507
• Languages: English
• ISSNs: 0366-6999

Abstract

Abstract: Background: Computed tomography images are easy to misjudge because of their complexity, especially images of solitary pulmonary nodules, of which diagnosis as benign or malignant is extremely important in lung cancer treatment. Therefore, there is an urgent need for a more effective strategy in lung cancer diagnosis. In our study, we aimed to externally validate and revise the Mayo model, and a new model was established. Methods: A total of 1450 patients from three centers with solitary pulmonary nodules who underwent surgery were included in the study and were divided into training, internal validation, and external validation sets ( n = 849, 365, and 236, respectively). External verification and recalibration of the Mayo model and establishment of new logistic regression model were performed on the training set. Overall performance of each model was evaluated using area under receiver operating characteristic curve (AUC). Finally, the model validation was completed on the validation data set. Results: The AUC of the Mayo model on the training set was 0.653 (95% confidence interval [CI]: 0.613–0.694). After re-estimation of the coefficients of all covariates included in the original Mayo model, the revised Mayo model achieved an AUC of 0.671 (95% CI: 0.635–0.706). We then developed a new model that achieved a higher AUC of 0.891 (95% CI: 0.865–0.917). It had an AUC of 0.888 (95% CI: 0.842–0.934) on the internal validation set, which was significantly higher than that of the revised Mayo model (AUC: 0.577, 95% CI: 0.509–0.646) and the Mayo model (AUC: 0.609, 95% CI, 0.544–0.675) ( P < 0.001). The AUC of the new model was 0.876 (95% CI: 0.831–0.920) on the external verification set, which was higher than the corresponding value of the Mayo model (AUC: 0.705, 95% CI: 0.639–0.772) and revised Mayo model (AUC: 0.706, 95% CI: 0.640–0.772) ( P < 0.001). Then the prediction model was presented as a nomogram, which is easier to generalize. Conclusions: After external verification and recalibration of the Mayo model, the results show that they are not suitable for the prediction of malignant pulmonary nodules in the Chinese population. Therefore, a new model was established by a backward stepwise process. The new model was constructed to rapidly discriminate benign from malignant pulmonary nodules, which could achieve accurate diagnosis of potential patients with lung cancer.