Association between CT features and tumor invasiveness of lung cancer associated with cystic airspace

Background: Lung cancer associated with cystic airspaces is a rare pattern in lung cancer morphology and invasive adenocarcinoma makes the most part of pathological types after resection, of which the prognosis is significantly worse than pre-invasive adenocarcinoma. We aimed to investigate the predictors of lung invasive adenocarcinoma associated with cystic airspaces. Methods: We identified 54 patients from January 2017 to January 2022 with cystic associated lung lesions who underwent thoracic surgery or bronchoscopic biopsy proved to be malignant or benign. All patients underwent thin-section computed tomography (CT) scan of lung and three-dimensional reconstruction. All the adenocarcinoma associated with cystic airspaces were categorized into two groups: (Ⅰ) non-invasive adenocarcinoma, include atypical adenomatous hyperplasia (AAH), adenocarcinoma in situ (AIS) and minimally invasive adenocarcinoma (MIA), (Ⅱ) invasive adenocarcinoma, thus invasive adenocarcinoma (IAC). The clinical data and radiological features were collected and analyzed. Logistic regression models were used to identify findings of association between non-invasive and invasive adenocarcinoma. Receiver operating curve (ROC) was performed to evaluate the diagnostic performance. Results: Of the 54 patients with cystic airspaces associated lung lesions there were 49 lung cancer and 5 benign diseases. The histological examination showed 1(1/49) AAH, 3(3/49) AIS, 9(9/49) MIA, 33(33/49) IAC, 2(2/49) squamous cell carcinoma (SC) and 1(1/49) adenosquamous carcinoma. Univariable analysis of features in lung adenocarcinoma associated with cystic airspaces revealed that the maximum axial diameter of the whole lesion, the maximum diameter of the cyst, volume, mean attenuation and mass of the whole lesion, mixed-ground glass component of the cyst wall, spiculated and/or lobulated margins of the lesions, air bronchogram and pleural tag as associated with invasive lung adenocarcinoma associated with cystic airspaces (all, P < 0.05). Multivariable analysis revealed the mean attenuation of the whole lesion and the maximum diameter of the cyst remained as predictors of invasiveness (P = 0.013 and P = 0.035, respectively). The performance of model based on these two variables were as follows: sensitivity of 90.9% and specificity of 100%, with an area under the curve of 0.974 (95% confidence interval: 0.937-1.000). Conclusions: The mean attenuation of the whole lesion and the diameter the cystic airspaces could be predictors of tumor invasiveness of lung adenocarcinoma associated with cystic airspaces.