Long-term radiologic trajectories and corticosteroid strategies in COVID-19 patients with pre-existing ILD: a 24-month cohort study

Patients with pre-existing interstitial lung disease (ILD) are highly vulnerable to SARS-CoV-2 infection, yet the long-term radiologic and clinical consequences remain poorly defined. This study evaluated 24-month outcomes, radiologic trajectories, and the impact of corticosteroid strategies in hospitalized COVID-19 patients with ILD. We conducted a retrospective longitudinal cohort study including 82 hospitalized COVID-19 patients with pre-existing ILD (December 2022–April 2023). Clinical variables, treatments, and serial high-resolution CT (HRCT) scans at baseline, 6, 12, and 24 months were analyzed. HRCT data were quantified using AI-assisted automated whole-lung segmentation, and radiologic trajectory phenotypes were identified using latent class growth mixture modeling (LCGMM). Cox regression, ROC analysis, and LOESS models assessed the associations between corticosteroid exposure, tapering duration, radiologic progression, and mortality. Three distinct 24-month CT trajectories were identified: Improved (n = 22, 43.1%), Stable–Low (n = 10, 19.6%), and Unfavorable (n = 19, 37.3%). The Unfavorable group exhibited higher baseline inflammatory markers and lower PaO₂/FiO₂ ratios. In adjusted Cox analyses incorporating disease severity indicators, longer corticosteroid tapering duration was statistically associated with improved survival (HR = 0.929, 95% CI 0.866–0.996), whereas associations between maximal corticosteroid dose and mortality were attenuated after adjustment. A tapering duration of ≤ 35 days was identified as a threshold associated with increased mortality risk (AUC = 0.742). Longer tapering duration (> 35 days) was associated with greater radiologic resolution, whereas rapid tapering (≤ 35 days) was associated with persistent radiologic abnormalities. COVID-19 survivors with pre-existing ILD exhibit divergent 24-month radiologic trajectories, and a substantial proportion experience progressive parenchymal involvement. In this observational cohort study, the duration of corticosteroid dose reduction is associated with imaging changes and survival rates. AI-assisted HRCT quantification combined with trajectory modeling provides a promising framework for individualized monitoring and early antifibrotic consideration in this high-risk population.