Distinct phenotypic clusters associated with progressive pulmonary fibrosis in anti-synthetase syndrome-associated interstitial lung disease.

Background: The clinical manifestations of anti-synthetase syndrome-associated interstitial lung disease (ASyS-ILD) are highly heterogeneous, and the mechanisms underlying its development and progression remain poorly understood. This study aimed to identify distinct clinical phenotypes of ASyS-ILD and clusters associated with progressive pulmonary fibrosis (PPF) using unsupervised cluster analysis. Methods: A total of 134 patients with ASyS-ILD were retrospectively enrolled. Unsupervised hierarchical clustering based on Gower distance and Ward's method was performed. The optimal number of clusters was determined by silhouette analysis. Clinical manifestations, laboratory findings, pulmonary function, and prognosis were compared among clusters. A simplified decision tree model based on classification and regression tree (CART) analysis was further developed for cluster assignment. Results: Three distinct clinical clusters were identified. Cluster 1 (n=51, 38.1%) represented a relatively stable phenotype, characterized by mild systemic inflammation, preserved lung function, and favorable outcomes. Cluster 2 (n=49, 36.6%) exhibited dermatomyositis-like manifestations, including prominent muscle weakness (57.1%), Gottron's sign (53.1%) and V sign rash (16.3%), and elevated muscle-associated enzymes. In addition, systemic inflammatory markers were significantly increased in cluster 2, including CRP (1.13 ng/mL), ESR (26.0 mm/h), ferritin (204.1 ng/mL). Cluster 3 (n=34, 25.4%) represented a PPF phenotype, characterized by significantly elevated fibrosis-related biomarkers, including CA125 (22.3 U/mL), CA15-3 (50.6 U/mL), and CYFRA21-1 (8.07 ng/mL), together with severe pulmonary involvement, markedly impaired DLCO (44.6%), and the highest prevalence of PPF (47.1%). Notably, Cluster 3 exhibited the highest all-cause mortality (23.5%), and Kaplan-Meier analysis further confirmed its worse prognosis among the three clusters (log-rank p=0.029). For rapid cluster classification, a simplified decision tree model incorporating CYFRA21-1, serum albumin, and neutrophil count was further developed and demonstrated good performance. Conclusions: ASyS-ILD comprises three distinct phenotypic subgroups with different clinical manifestations and outcomes. Cluster-based phenotyping may improve risk stratification and facilitate individualized management in patients with ASyS-ILD.