P62.06 A Radiomics Nomogram for Preoperative Prediction of Occult Lymph Node Metastasis in Early-Stage Solid Lung Adenocarcinoma

smoking status (active 5.8[4.6-7.4],former < 15 years 7. 3[5.5-9.8],former > 15 years 7. 3[5.5-9.6] vs. never), number of nodules 1.13[1.07-1.20],largest nodule size 1.06[1.05-1.06],nodule location (ORs varied 0.50 to 1.17 with upper lobes at highest risk).Nodule edge characteristics and nodule density were not predictive of a cancer diagnosis, but had a high percentage of missing information.Our full multi-variable model included age, BMI, co-morbidities, family cancer history, COPD, asbestos exposure, pack-years of smoking, smoking status, nodule number, nodule size, nodule location, and cavitation.This model had an AUC of 0.84 for predicting cancer diagnosis in patients with incidentally detected lung nodules (Figure 1).Using the optimal cut-off from this predictive model to identify lung cancer yielded a sensitivity of 73% and specificity of 79%.Conclusion: Systematic management of incidental lung nodules is important for identifying persons with early stage lung cancer.Clinical characteristics, beyond those identified by radiology, are useful in identifying patients whose incidentally detected nodules are most likely to result in a lung cancer diagnosis in this real-world population.