A clinical predictive model for the long-term survival of progressive fibrosis interstitial lung disease patients

Background: In patients with chronic fibrosing interstitial lung disease (ILD), some may develop a progressive fibrosing (PF) phenotype, which presents as rapid progression and often results in poor clinical outcomes. Objective: The objective of this study was to construct and test a model to identify independent predictors of mortality in PF-ILD and to trace the lung function trajectory of patients with PF-ILD. Design: This multicenter retrospective cohort study enrolled patients with PF-ILD from two distinct centers with 8-year follow-up to develop and validate a prognostic nomogram based on clinical factors and assess longitudinal lung function trajectories. Methods: We enrolled patients diagnosed with PF-ILD from China-Japan Friendship Hospital (training cohort) and Jiulongpo Hospital of Traditional Chinese Medicine (validation cohort). Survival status was recorded during the 8-year follow-up period. Clinical demographics, laboratory data, pulmonary function test (PFT) results, and high-resolution computed tomography results were collected for analysis. A training cohort of patients with PF-ILD was used to identify predictors of mortality, which were then validated in an external cohort. A nomogram was established based on multivariate factors. The predictive performance of the model was evaluated using receiver operating characteristic curves and calibration curves. Survival estimates were performed using the Kaplan–Meier method and compared using the log-rank test. The PFT trajectory was estimated using a linear mixed model. Results: A total of 1419 patients with PF-ILD from China-Japan Friendship Hospital (training cohort) and 282 patients with PF-ILD from Jiulongpo Hospital of Traditional Chinese Medicine (validation cohort) were enrolled. During the 8-year follow-up, 150 (10.57%) patients received lung transplantation, while 43.55% ( n = 618) of cases reached mortality, with a median survival time of 53 months in the BJ cohort. A predictive model was built based on ILD subgroups, baseline Forced Vital Capacity% pred (FVC%pred), baseline Diffusing Capacity of the Lung for Carbon Monoxide% pred (DLCO%pred), age at diagnosis, antifibrosis treatment, gastroesophageal reflux complication, C-Reactive Protein levels, and BMI. We also found AEs and progressive declines in FVC and DLCO, particularly after the third year post-diagnosis, were strongly associated with poor prognosis and may serve as important longitudinal biomarkers for risk stratification in PF-ILD. Conclusion: A predictive model incorporating multiple factors effectively predicted 8-year survival in patients with PF-ILD. In addition to these baseline predictors, AEs and progressive declines in FVC and DLCO were strongly associated with poor prognosis and may serve as valuable longitudinal biomarkers for ongoing risk stratification.