Bronchoalveolar lavage fluid cytokine-based clustering identifies prognostically relevant cytokine phenotypes across interstitial lung diseases

BACKGROUND: Existing classification systems for interstitial lung diseases (ILDs) are limited and do not fully reflect the underlying biological drivers. We sought to identify cytokine-defined ILD subgroups based on bronchoalveolar lavage fluid (BALF) cytokine profiles and to examine their clinical relevance, particularly their association with prognosis. METHODS: We reanalyzed a BALF cytokine dataset from 84 patients with ILDs, including idiopathic pulmonary fibrosis (IPF), idiopathic nonspecific interstitial pneumonia (iNSIP), cryptogenic organizing pneumonia, sarcoidosis, idiopathic inflammatory myopathy-associated ILD, and RA-ILD. Cytokine concentrations were measured using a multiplex bead-based assay. We used hierarchical clustering, principal component, correspondence, cosine similarity, and discriminant analyses to identify cytokine-defined subgroups and examine their clinical associations. RESULTS: We identified three cytokine-level-defined clusters: Clusters 1, 2, and 3, characterized by globally high, intermediate, and low cytokine levels, respectively. Clusters 1 and 3 were associated with poor and favorable prognoses, respectively. Cytokine-defined clusters did not fully correspond with conventional disease categories, although Cluster 3 predominantly included patients with sarcoidosis. RA-ILD was mainly distributed in Clusters 1 and 2 together with IPF, iNSIP, COP, and IIM-ILD. However, RA-ILD was distinguished by relatively elevated interleukin-17 and granulocyte-macrophage colony-stimulating factor levels. CONCLUSIONS: BALF cytokine profiling can identify cytokine intensity-defined ILD subgroups that are associated with prognosis and are not fully captured by conventional disease classification. Although RA-ILD shares cytokine features with IPF and iNSIP, it also exhibits distinct immune characteristics. These findings suggest that intrapulmonary cytokine profiling could facilitate ILD stratification.