Document resource
Introduction The presence and severity of systemic sclerosis-associated interstitial lung disease (SSc-ILD) is associated with increased pulmonary vascular volume (PVV), quantified through radiomic analysis of computed tomography images. Recent studies have identified clinical risk factors for new-onset of ILD starting from a previous negative CT, still not optimally identifying patients at higher risk. Here, we explored the potential of radiomics in predicting ILD onset.Material and Methods We included ILD-negative SSc patients from the Zurich cohort and at least one clinical-radiologic follow-up. Experienced thoracic radiologists evaluated ILD presence at each timepoint. Lung texture analysis™ (Imbio) quantified PVV and lung pathologies (hyperlucency, ground-glass, reticular, honeycombing), for whole lungs and by lung zones (upper, middle, lower). We tested radiomic predictors of ILD onset in univariable and multivariable models, using Cox regression for the long-term observation and generalized estimating equation for 1-year risk, the latter accounting for repeated measures. Both models were adjusted for clinical risk factors (age, sex, autoantibodies, DLCO%, increased inflammatory biomarkers, digital ulcers, NYHA class >= II and haemoglobin, combined into a single weighted variable).Results Among 248 eligible SSc patients without ILD, 54 (22%) had new-onset ILD over 39 (24-72) months median follow-up. These patients were more frequently male, diffuse SSc, anti-Topoisomerase-I positive, with shorter disease duration. At the time of negative HRCT, patients with new-onset ILD already showed higher whole-lung PVV, compared to those who remained ILD-negative. Conversely, all other radiomic findings were comparable. In multivariable Cox regression, whole-lung PVV independently predicted ILD onset [HR 1.054 (1.030-1.078)], and remained significant after adjustment for clinical risk factors [HR 1.031 (1.010-1.052)].A secondary exploratory analysis confirmed PVV from each lung zone as predictive of ILD onset; in particular, lower-zone PVV retained an independent association with ILD onset [HR 1.323 (1.221-1.560)] when adjusted for clinical risk factors.Focusing on ILD onset in 1 year, we analysed 279 visits from 193 patients and identified 22 new ILD cases (11.4%). Univariable analysis showed an association between whole-lung PVV and ILD onset, particularly in the middle and upper lung zones. After adjusting for clinical factors, whole-lung PVV was no longer predictive. However, exploratory analysis revealed a significant association between upper-zone PVV and ILD onset [OR 4.127 (1.016–16.762)].Conclusions PVV in SSc patients without ILD associates with new-onset ILD, supporting radiomics as a cohort enrichment tool for preventive trials. The localization of increased PVV may reflect underlying pathophysiological mechanisms of ILD development and warrants further studies.