Abstract
BACKGROUND: Lymph node metastasis (LNM) of lung cancer is relatively common in clinical practice, and accurate diagnosis of LNM remains challenging. This study aimed to clarify the correlation between clinical computed tomography (CT) imaging features and LNM in lung cancers ≤30 mm according to different solid proportions (SPs). METHODS: A total of 2,074 patients with lung cancer confirmed by surgical pathology and lymph node dissection were included. All lung cancers were categorized into three groups: ground-glass nodule (GGN), solid nodule (SN), and GGN + SN. Predictive models 1, 2, and 3 were constructed for the respective groups, and univariate and multivariate logistic regression analyses were used to determine the LNM risk in each group. Differences between models were compared with the Delong test. RESULTS: In model 1 (GGN), SP was the sole independent risk factor, and the area under the curve (AUC) and accuracy of this model were 0.929 [95% confidence interval (CI): 0.911-0.947] and 0.850, respectively. In model 2 (SN), high blood pressure, heterogeneous ventilation or perfusion, short diameter (SD), and CT mean density value (CTmean) were independent risk factors, with the AUC and accuracy of this model being 0.733 (95% CI: 0.678-0.788) and 0.735, respectively. In model 3, age, SP, spiculation, pleural tag, and rim sign were independent risk factors, with the AUC and accuracy of this model being 0.904 (95% CI: 0.888-0.920) and 0.751, respectively. The Delong test revealed significant differences between model 2 and both models 1 and 3 (P<0.05). CONCLUSIONS: Model 1 and model 3 demonstrated high diagnostic value for the preoperative prediction of LNM in patients with lung cancer.