Predictive Model for Identifying Fibrotic Interstitial Lung Abnormalities on Computed Tomography for Respiratory Specialists

This is the authors' abstract. We don't add key points for this paper.

BACKGROUND AND OBJECTIVE: Interstitial lung abnormalities (ILAs) are often unreported, and fibrotic ILAs carry a worse prognosis. In this study, we aimed to develop and validate a prediction model for fibrotic ILAs using non-radiological factors. METHODS: Using data from the Kumamoto ILA study (a multicenter prospective cohort in Japan), we developed a four-variable model (age, pack-years, fine crackles, elevated serum surfactant protein-D) in 164 patients (2022-2023, derivation cohort) and temporally validated it in 207 patients (2023-2024, validation cohort), with internal-external cross-validation for between-center heterogeneity. A simplified integer score was similarly assessed against 24-month functional decline and treatment initiation within 3 years. RESULTS: Fibrotic ILAs comprised 82% of cases (approximately 60% subsequently met ATS-defined ILD criteria), consistent across cohorts. The four-variable model achieved an AUC of 0.79 (derivation) and 0.76 (validation; pooled 0.76, 95% CI 0.68-0.82), with good calibration and minimal between-center heterogeneity. The integer score performed similarly (AUC 0.79/0.75) and correlated with functional decline (%FVC ≥ 10% or %DLco ≥ 15%) and treatment initiation. CONCLUSION: Age, smoking exposure, fine crackles, and elevated serum surfactant protein-D are key predictors of progressive fibrotic ILAs, supporting early risk stratification and clinical decision-making. TRIAL REGISTRATION: UMIN000045149/2021.12.1.