Development and internal validation of predictive models for spread through air spaces in clinical stage IA lung adenocarcinoma.

Abstract

Objective: Spread through air spaces (STAS) in lung adenocarcinoma impacted prognosis and treatment decisions, but lacked reliable preoperative prediction. We aimed to construct an easy-to-use model for clinical stage IA adenocarcinoma patients.

Methods: This study analyzed 1212 patients with clinical stage IA lung adenocarcinoma undergoing lung resections from November 2020 to January 2022. Two logistic regression models were developed. Model 1 used demographic and computed tomography features, and Model 2 incorporated maximum standardized uptake values additionally. Internal validation used tenfold cross-validation. Model discrimination and calibration were described by the area under the curve (AUC) and Spiegelhalter z test, respectively.

Results: Prevalence of STAS was 10.6%. Model 1 consisted of maximum tumor diameter, smoking history, location, spiculation and lobulation, showing moderate discrimination (AUC = 0.700). Model 2 consisted of smoking history, the maximum standardized uptake value, spiculation and lobulation, receiving an AUC of 0.807 and good calibration. Model 2 has a sensitivity and a specificity of 0.857 and 0.652. A nomogram for Model 2 was also developed.

Conclusion: Our study developed and validated two predictive models for STAS for clinical stage IA lung adenocarcinoma. Model 2, integrating maximum standardized uptake value, outperformed Model 1 and offered a more comprehensive approach to predicting STAS. Surgeon could consider the results of Model 2 and intraoperative frozen sections sequentially to optimize surgical strategies. External validation remained warranted.