Latent class analysis of CT-based phenotypes in eosinophilic granulomatosis with polyangiitis and pulmonary involvement.

objectivesPulmonary manifestations of eosinophilic granulomatosis with polyangiitis (EGPA) are highly heterogeneous, and a standardized CT-based classification is currently lacking. This study aimed to evaluate whether latent class analysis (LCA) of chest CT findings can identify phenotypes with clinical and prognostic relevance in patients with EGPA. MATERIALS AND

methodsWe retrospectively enrolled 140 patients with EGPA and pulmonary involvement between 2014 and 2025. Chest CT findings were independently evaluated by two thoracic radiologists and analyzed using LCA to identify CT-based phenotypes. Reproducibility and clinical applicability were assessed by having radiologists reassign patients to phenotypes based on LCA-derived CT patterns while blinded to the original LCA assignments. Baseline characteristics and one-year clinical and radiological outcomes were compared across phenotypes. Modified Poisson regression was used to assess clinical relapse risk.

resultsLCA identified four distinct phenotypes: bronchial (32.1%; airway-predominant abnormalities), infiltrative (20.7%; parenchyma-predominant abnormalities), mixed (14.3%; combined bronchial and infiltrative features), and nonspecific (32.9%; without a dominant bronchial or infiltrative pattern). The bronchial and infiltrative phenotypes showed more favorable short-term outcomes, whereas the nonspecific phenotype mainly showed stable outcomes. The mixed phenotype remained associated with a higher risk of one-year clinical relapse after adjustment for age, sex, and baseline treatment. Radiologist reassignment showed excellent inter- and intra-observer agreement (Cohen's κ = 0.85 and 0.87).

conclusionLCA enables the identification of CT-based phenotypes with distinct clinical characteristics and short-term outcome patterns in EGPA, suggesting potential prognostic value and a practical role in clinical management. KEY POINTS: QuestionCan latent class analysis (LCA) of chest CT findings identify imaging phenotypes of pulmonary involvement in eosinophilic granulomatosis with polyangiitis (EGPA)? FindingsLCA identified four CT-based phenotypes (bronchial, infiltrative, mixed, and nonspecific), each characterized by distinct clinical and radiological features and short-term outcome patterns. Clinical relevanceCT-based phenotyping may help characterize pulmonary disease heterogeneity in EGPA and identify patients, particularly those with the mixed phenotype, who may require closer monitoring for clinical relapse.