Investigating Lung-RADS 0: A Retrospective Study Analyzing Clinical Markers and Lung Cancer Risk
Vedika Vyas
29/09/2026
This study investigated Lung-RADS 0 (Lung CT Screening Reporting and Data System) to address gaps in Lung-RADS misclassification and to inform improvements to Lung-RADS 0 (LR0). Clinical markers peripheral capillary oxygen saturation (SpO2), white blood cell (WBC) count, and serum albumin were examined to determine each’s statistical association with lung cancer and used to create a combined logistic regression model to predict binary outcomes for lung cancer. Using a sample of 122 patients from Indiana University, Fisher's exact tests were used to determine associations between each clinical marker using raw data and KNN-adjusted data. It was found in the raw data that having a serum albumin value under 3.5 g/dL was significantly associated with malignancy (p = 0.0159). After a secondary sensitivity analysis using KNN adjustment, WBC count became significant (p = 0.0406), though the raw data did not support this finding, and serum albumin remained statistically significant (p = 0.0110). The logistic regression model produced a cross-validated AUC of 0.53. At the optimal threshold, specificity was 93.9%, but sensitivity was extremely low at 37.5%, which was caused by the lack of positive cases available compared to the negative cases (8 vs 114 cases). Overall, SpO2, WBC count, and serum albumin alone and combined were not reliable diagnostic indicators of lung cancer in LR0 patients, though larger and more diverse training sets may improve the accuracy of multivariate models, such as logistic regression. Still, research efforts should be redirected towards finding more specific molecular markers or radiological patterns and their association with lung cancer.