Risk factors for pathological invasiveness of non-mucinous invasive pulmonary adenocarcinoma with a maximum diameter of less than or equal to 3 cm in thin-section CT: a retrospective study.

To explore the risk factors for pathological invasiveness and lymph node metastasis of non-mucinous invasive lung adenocarcinoma with a maximum diameter of less than or equal to 3 cm in thin slice CT. The aim of this study was to develop some models for predicting pathological invasiveness and lymph node metastasis in patients with early nonmucinous invasive lung adenocarcinoma. Our study analyzed the clinical, imaging, and pathological characteristics of 700 cases of non mucinous invasive lung adenocarcinoma, and used multiple learning methods such as univariate, multivariate logistic regression, and ROC working curve to establish risk factors for pathological invasion and lymph node metastasis. The results were as follows: 1. In the mGNN group, multivariate logistic regression identified two independent risk factors: CTR (p < 0.001) and the presence of micropapillary component (p = 0.005).With an area under the ROC curve of 0.881, sensitivity of 79.6%, and specificity of 89.3%0.2. In the SN group, multivariate logistic regression identified three independent risk factors: maximum lesion diameter (p = 0.045), IASLC grading (p < 0.05), and the presence of micropapillary component (p = 0.012). With an area under the ROC curve of 0.756, sensitivity of 71.9%, and specificity of 68.3%0.3. In the SN group, multivariate logistic regression identified two independent risk factors for lymph node metastasis: maximum lesion diameter (p = 0.007), vascular invasion (p = 0.004). With an area under the ROC curve of 0.782, sensitivity of 96.6%, and specificity of 45.3%. The models showed excellent discrimination to predict pathological invasiveness and lymph node metastasis in patients with non-mucinous invasive lung adenocarcinoma with a maximum diameter of less than or equal to 3 cm in thin slice CT. In addition, the predictive model had predictive potential before the end of surgery and could inform clinical decision making.