Document resource
Introduction Radiomic analysis of chest high-resolution computed tomography (HRCT) has emerged as a novel tool for the objective assessment of systemic sclerosis-associated interstitial lung disease (SSc-ILD). An increase in the pulmonary vessel volume (PVV) quantified on HRCT has been linked to SSc-ILD severity and linked to mortality in idiopathic pulmonary fibrosis (IPF) and rheumatoid arthritis (RA)-associated ILD. Here, we evaluated the value of CT-derived radiomic features as predictors of ILD progression and mortality in SSc-ILD.Material and Methods We included SSc patients from the Zurich cohort with ILD confirmed on HRCT and follow-up information.HRCT scans underwent lung texture analysis using the LTA™ software (Imbio). The extent of PVV and lung pathologies (hyperlucency, ground-glass, reticular, honeycombing) were quantified for whole lungs and by lung zones (upper, middle, lower).Progression analysis focused on patients with repeated pulmonary function tests within 12±3 months. Progression was defined as an absolute FVC decline >= 5% or absolute DLCO decline >=10%. Radiomics predictors of progression were assessed with Generalized Estimating Equations (GEE), adjusted for confounders (age, sex, disease duration, NYHA class >II, baseline DLCO and FVC, diffuse skin involvement).Cox proportional hazards models were used to assess radiomic predictors of mortality, observing patients from the first ILD-positive CT until last contact (death or alive status), adjusted for the SCOpE score.Results Among 261 yearly follow-up visits from 160 SSc-ILD patients, 40% showed at least one episode of ILD progression, with a total of 99 (37.7%) occurrences of functional decline. Radiomic features were comparable among progressors and non-progressors. Among all radiomic features, the extent of honeycombing in the whole lung was identified as independent predictor of functional decline [adj OR 1.539, 95% CI 1.174-2.017].Over 42 (IQR 27-110) months follow-up, 65/267 (24%) SSc-ILD patients died. Patients who died were older, with shorter disease duration, more frequently smokers ever, with increased inflammatory markers and more severe functional impairment. At baseline, SSc-ILD patients who died also showed greater extent of both PVV% and parenchymal alterations across all lung zones. PVV% was an independent predictor of mortality [adj HR 1.217, 95% CI 1.094–1.354], particularly in the upper zones [adj HR 1.280, 95% CI 1.118–1.466].Conclusions The extent of honeycombing predicts ILD progression, while PVV% is a predictor of mortality, consistently with the data in IPF and RA-ILD, supporting the role of radiomic LTA analysis in the risk stratification and prognosis of SSc-ILD patients.