Something doesnt add up here: lung cancer calculators and radiomics applied to anca-associated vasculitis

SESSION TITLE: Pulmonary Manifestations of Systemic Disease: Don't Forget About Breathing SESSION TYPE: Rapid Fire Original Inv PRESENTED ON: 10/09/2023 12:00 pm - 12:45 pm PURPOSE: Granulomatosis with polyangiitis (GPA) commonly presents with radiographically detected pulmonary nodules or masses caused by necrotizing granulomatous inflammation [1]. Prior to a definitive diagnosis of GPA, such lesions often raise concern for malignancy. Several non-invasive models exist to guide management of the indeterminant pulmonary nodules (lesions <30 mm in diameter), yet most models are trained and validated from specific cohorts limiting the external validity of these tools{C}{C} [2]. Clinical risk calculators are commonly utilized, and additional radiomics has yielded several models with novel applications to the management of pulmonary nodules [3]. We sought to apply common risk calculators and a validated radiomics model to patients with GPA and pulmonary nodules to identify potential limitations of these modalities. METHODS: We retrospectively analyzed patients at Mayo Clinic sites with GPA with pulmonary nodules at time of diagnosis between 1/1/2009 and 9/1/2022. Inclusion criteria were age over 18 years, meeting Chapel Hill and ACR/EULAR consensus definitions of GPA, high resolution CT chest at initial presentation showing solid nodules 7-30 mm. The Mayo Clinic model and Brock model were calculated for the dominant nodule. Subsequently, a semi-automated region-growing approach was used to perform nodule segmentation (ANALYZE Biomedical Imaging Resource; Mayo Clinic, Rochester, MN, USA) with manual editing to exclude adjacent structures. From this, the BRODERS classifier (Benign versus aggRessive nODule Evaluation using Radiomic Stratification) was used to predict probability of malignancy. RESULTS: 45 patients met eligibility criteria among all Mayo Clinic Sites. Patients were predominantly male (N=24), with a mean of age 55 ±15, and 48% (N= 22) were former smokers. The median nodule size was 18 mm (IQR 11-24), with a median number of 5 (IQR 3-14) nodules, predominantly distributed in the middle/lower lobes (N=24), and without spiculation (N=37). Using the Brock and Mayo models, the median probability of malignancy was 9.8% (IQR 3.4, 20.0) and 15.4% (IQR 9.8, 36.8), respectively. According to the BRODERS radiomics model, the median probability of malignancy was 0.935 (IQR 0.833-0.989), ultimately classifying 41 (82%) of the nodules as malignant based on a cutoff value identified in previous studies [3]. CONCLUSIONS: Pulmonary nodules present at time of diagnosis of GPA were predominantly classified as intermediate risk by Brock and Mayo calculators and high risk for malignancy using BRODERS classifiers. CLINICAL IMPLICATIONS: Applying clinical risk calculators and radiomics models to pulmonary nodules requires both appropriate clinical context and knowledge of limitations of these prediction models. Inappropriate application of such models may lead to unnecessary testing or may delay definitive diagnosis and treatment. References 1. Lohrmann, C., et al., Pulmonary manifestations of wegener granulomatosis: CT findings in 57 patients and a review of the literature. Eur J Radiol, 2005. 53(3): p. 471-7. 2. MacMahon, H., et al., Guidelines for Management of Incidental Pulmonary Nodules Detected on CT Images: From the Fleischner Society 2017. Radiology, 2017. 284(1): p. 228-243. 3. Maldonado, F., et al., Validation of the BRODERS classifier (Benign versus aggRessive nODule Evaluation using Radiomic Stratification), a novel HRCT-based radiomic classifier for indeterminate pulmonary nodules. European Respiratory Journal, 2021. 57(4): p. 2002485. DISCLOSURES: No relevant relationships by Brian Bartholmai No relevant relationships by Rodrigo Cartin-Ceba No relevant relationships by Samuel Falde No relevant relationships by Augustine Lee No relevant relationships by Tobias Peikert No disclosure on file for Srini Rajagopalan Advisory relationship with ChemoCentryx Please note: 2022 Added 10/31/2022 by Ulrich Specks, source=Web Response, value=Consulting Advisory Committee Member relationship with Astra Zeneca Please note: 2021-2023 Added 12/03/2022 by Ulrich Specks, source=Web Response, value=Consulting fee Advisory Committee Member relationship with Boehringer Ingelheim Please note: 2022 Added 12/03/2022 by Ulrich Specks, source=Web Response, value=Consulting fee Advisory Committee Member relationship with Amgen Please note: 2023 Added 03/24/2023 by Ulrich Specks, source=Web Response, value=Consulting fee