A nomogram combining thoracic CT and tumor markers to predict the malignant grade of pulmonary nodules ≤3 cm in diameter

Abstract

BackgroundWith the popularity of computed tomography (CT) of the thorax, the rate of diagnosis for patients with early-stage lung cancer has increased. However, distinguishing high-risk pulmonary nodules (HRPNs) from low-risk pulmonary nodules (LRPNs) before surgery remains challenging.MethodsA retrospective analysis was performed on 1064 patients with pulmonary nodules (PNs) admitted to the Qilu Hospital of Shandong University from April to December 2021. Randomization of all eligible patients to either the training or validation cohort was performed in a 3:1 ratio. Eighty-three PNs patients who visited Qianfoshan Hospital in the Shandong Province from January through April of 2022 were included as an external validation. Univariable and multivariable logistic regression (forward stepwise regression) were used to identify independent risk factors, and a predictive model and dynamic web nomogram were constructed by integrating these risk factors.ResultsA total of 895 patients were included, with an incidence of HRPNs of 47.3% (423/895). Logistic regression analysis identified four independent risk factors: the size, consolidation tumor ratio, CT value of PNs, and carcinoembryonic antigen levels in blood. The area under the ROC curves was 0.895, 0.936, and 0.812 for the training, internal validation, and external validation cohorts, respectively. The Hosmer-Lemeshow test demonstrated excellent calibration capability, and the fit of the calibration curve was good. DCA has shown the nomogram to be clinically useful.ConclusionThe nomogram performed well in predicting the likelihood of HRPNs. In addition, it identified HRPNs in patients with PNs, achieved accurate treatment with HRPNs, and is expected to promote their rapid recovery.