A nomogram integrating the clinical and CT imaging characteristics for assessing spread through air spaces in clinical stage IA lung adenocarcinoma

Purpose: This study aimed to create a nomogram model to predict the spread through air spaces (STAS) in patients diagnosed with stage IA lung adenocarcinoma, utilizing a substantial sample size alongside a blend of clinical and imaging features. This model serves as a valuable reference for the preoperative planning process in these patients. Materials and methods: A total of 1244 individuals were included in the study. Individuals who received surgical intervention between January 2022 and May 2023 were categorized into a training cohort (n=950), whereas those treated from June 2023 to October 2023 were placed in a validation cohort (n=294). Data from clinical assessments and CT imaging were gathered from all participants. In the training cohort, analyses employing both multivariate and univariate logistic regression were performed to discern significant clinical and CT characteristics. The identified features were subsequently employed to develop a nomogram prediction model. The evaluation of the model's discrimination, calibration, and clinical utility was conducted in both cohorts. Results: In the training cohort, multivariate logistic regression analysis revealed several independent risk factors associated with invasive adenocarcinoma: maximum diameter (OR=2.459, 95%CI: 1.833-3.298), nodule type (OR=4.024, 95%CI: 2.909-5.567), pleura traction sign (OR=2.031, 95%CI: 1.394-2.961), vascular convergence sign (OR=3.700, 95%CI: 1.668-8.210), and CEA (OR=1.942, 95%CI: 1.302-2.899). A nomogram model was constructed utilizing these factors to forecast the occurrence of STAS in stage IA lung adenocarcinoma. The Area Under the Curve (AUC) measured 0.835 (95% CI: 0.808-0.862) in the training cohort and 0.830 (95% CI: 0.782-0.878) in the validation cohort. The internal validation conducted through the bootstrap method yielded an AUC of 0.846 (95% CI: 0.818-0.881), demonstrating a robust capacity for discrimination. The Hosmer-Lemeshow goodness-of-fit test confirmed a satisfactory model fit in both groups (P > 0.05). Additionally, the calibration curve and decision analysis curve demonstrated high calibration and clinical applicability of the model in both cohorts. Conclusion: By integrating clinical and CT imaging characteristics, a nomogram model was developed to predict the occurrence of STAS, demonstrating robust predictive performance and providing valuable support for decision-making in patients with stage IA lung adenocarcinoma.