A Modified Model for Preoperatively Predicting Malignancy of Solitary Pulmonary Nodules: An Asia Cohort Study

Background. With the recent widespread use of computed tomography, interest in ground glass opacity pulmonary lesions has increased. We aimed to develop a model for predicting the probability of malignancy in solitary pulmonary nodules. Methods. We assessed 846 patients with newly discovered solitary pulmonary nodules referred to Fujian Medical University Union Hospital. Data on 18 clinical and 13 radiologic variables were collected. Two thirds of the patients were randomly selected to derive the prediction model (derivation set); the remaining one third provided a validation set. The lesions were divided according to proportion of ground glass opacity (less than 50% or 50% or greater). Univariate analysis of significant covariates for their relationship to the presence of malignancy was performed. An equation expressing the probability of malignancy was derived from these findings and tested on data from the validation group. Receiver-operating characteristic curves were constructed using the prediction model and the Mayo Clinic model. Results. In lesions with less than 50% ground glass opacity, three clinical characteristics (age, presence of symptoms, total protein) and three radiologic characteristics (diameter, lobulation, calcified nodes) were independent predictors of malignancy. In lesions with 50% or more ground glass opacity, two clinical characteristics (sex, percent of forced expiratory volume in 1 second accounting for expected value) and two radiologic characteristics (diameter, calcified nodes) were independent predictors of malignancy. Our prediction model was better than the Mayo Clinic model to distinguish between benign and malignant solitary pulmonary nodules (p < 0.05). Conclusions. Our prediction model could accurately identify malignancy in patients with solitary pulmonary nodules, especially in lesions with 50% or more ground glass opacity. (C) 2015 by The Society of Thoracic Surgeons