关键词: 肺腺癌, 高级别模式, 预测模型, 高分辨率CT

Abstract:

Background and purpose: Studies have shown that when high-grade histological patterns (micropapillary and solid patterns) are present, patients with lung adenocarcinoma have a significantly poorer prognosis and often require more aggressive treatment modalities, and preoperative determination of the presence of any high-grade patterns (HGP) in invasive lung adenocarcinoma can help predict patient prognosis and determine treatment strategies. The aim of the study was to establish a logistic regression model based on high-resolution CT signs to predict the HGP of stage ⅠA lung adenocarcinoma. Methods: The clinical, pathological and imaging data of 443 patients (445 lesions) with stage ⅠA lung adenocarcinoma confirmed by pathology diagnosis from First People's Hospital of Xiaoshan District (Oct. 2018 to Mar. 2021) and Zhongshan Hospital of Fudan University (Jan. 2018 to Dec. 2020) were retrospectively analyzed. The 445 lesions were divided into two groups according to the presence or absence of HGP in pathological findings: HGP (n=88) and non-HGP (n-HGP) (n=357). The clinical and pathological data of the patients included age, gender, smoking history, tumor location, stage and pathological growth pattern. On CT imaging, the size, density, shape, burr sign, lobulation sign, vacuole sign, air bronchus sign and pleural depression sign were observed. Mann-Whitney U test was used to compare quantitative parameters between the two groups, and χ² test or Fisher's exact test was used for enumeration of data. The independent predictors were screened by univariate combined with multivariate logistic regression analysis, and the clinical model, CT model and clinical-CT model were constructed according to the results of multivariate logistic regression analysis. DeLong test was used to compare the diagnostic efficacy between models. Results: In the univariate analysis, there were significant differences in age, gender, smoking history, tumor size, density, shape, burr, lobulation sign and pleural traction between the HGP group and the n-HGP group (P<0.05). Multivariate logistic regression analysis showed tumor size (P=0.04; OR=1.063, 95% CI: 1.003-1.126), density (P<0.001; OR=8.249, 95% CI: 4.244-16.034), lobulation sign (P=0.001; OR=3.101, 95% CI: 1.598-6.021) were independent predictors of HGP, and the area under curve (AUC) values of clinical model, CT model and clinical-CT model for predicting HGP were 0.634, 0.838 and 0.834, respectively. Conclusion: Tumor size, density and lobulation sign are independent predictors of HGP in stage ⅠA lung adenocarcinoma. The logistic regression model based on high-resolution CT signs has good diagnostic performance and can provide a certain reference for clinical diagnosis and surgical treatment.

Key words: Lung adenocarcinoma, High-grade pattern, Prediction model, High-resolution CT