Pulmonary function tests for identifying pathological invasive stage I lung adenocarcinoma: A multiple-center retrospective cohort study

Abstract

Background Preoperative biopsy can hardly be used to diagnose lung cancer invasion; therefore, supplementary methods to estimate pathological tumor invasiveness are needed to identify candidates for limited resection. We aim to ascertain the risk factors and create and verify a model for predicting lung cancer invasion likelihood. Methods A nomogram was trained and validated on retrospectively collected data of patients with primary lung cancer whose pulmonary function was examined within 3 months before surgery. Least absolute shrinkage and selection operator logistic regression were used for important factor selection. The nomogram was established by combining preoperative pulmonary function tests (PFTs) and clinical factors. The area under the receiver operating characteristic curve and decision curve analysis (DCA) were used to evaluate the model’s predictive performance and clinical utility, respectively. Results Lung function impairment was detected in 508 patients (38.72%, 508/1312). The prediction model, which included age (odds ratio [OR] = 1.02), tumor size (OR = 1.31), lung function (OR = 3.10), basophils (OR = 0.51), and direct bilirubin levels (OR = 1.15), showed good performance in both sets. The areas under the curve for predicting lung cancer invasion were 0.820 (95% confidence interval [CI]: 0.781–0.858), 0.758 (95% CI: 0.659–0.858), and 0.838 (95% CI: 0.797–0.879) in the training, internal validation, and external validation sets, respectively, indicating good performance. In the multivariable analysis, patients with restrictive ventilation impairment (OR 2.86 [95% CI 1.43–5.69]) and diffusion capacity impairment (OR 4.23 [95% CI 1.00-17.84]) had high tumor invasion risks. Conclusions Lung function impairment could potentially serve as a biomarker for stage I lung adenocarcinoma invasion.