Abstract
PURPOSE: This study aimed to develop and validate a preoperative CT-based nomogram combined with clinical and radiological features for distinguishing invasive from non-invasive pulmonary adenocarcinoma. PATIENTS AND METHODS: A total of 167 patients with solitary pulmonary nodules and pathologically confirmed adenocarcinoma treated between January 2020 and December 2020 at Hebei General Hospital were retrospectively assessed. To evaluate the probability of invasive pulmonary adenocarcinoma, we developed three models, the multivariate logistic regression model, the stepwise logistic regression model, and the cross-validation model. The Akaike information criterion (AIC) was used to compare the relative strength of different models, and the area under the curve (AUC) was used to quantify the predictive accuracy. The best performing model was presented as a nomogram, calibrated and evaluated for clinical utility. RESULTS: The stepwise logistic regression model revealed highest and mean attenuations of non-enhanced CT images, and lobulation and vacuole presence were predictive factors of invasive pulmonary adenocarcinoma. This model (AIC = 67.528) with the lowest AIC value compared with that of the multivariate logistic regression model (AIC = 69.301) or the cross-validation model (AIC = 81.216) was identified as the best model, and its AUC value (0.9967; 95% CI, 0.9887-1) was higher than those of the other two models. The calibration curve showed optimal agreement in invasive pulmonary adenocarcinoma probability as predicted by the nomogram and the actual value. CONCLUSION: We developed and validated a nomogram that could estimate the preoperative probability of invasive pulmonary adenocarcinoma in patients with solitary pulmonary nodules, which may be useful in clinical decision-making associated with personalized surgical intervention and therapeutic regimen selection.