Multiple Correspondence Analysis reveals two different phenotypes of idiopathic chronic pancreatitis

作者：Lucía Guilabert, Patrick Maisonneuve, Chiara Deli, Daniele Bellafante, Laura Apadula, Ilaria Marengon, Giuseppe Vanella, Matteo Tacelli, Giuseppe Dell’Anna, Maria Chiara Petrone, Alberto Mariani, Enrique de‐Madaria, Paolo Giorgio Arcidiacono, Gabriele Capurso · 发表于：Digestive and Liver Disease · 年份：2026 · DOI：10.1016/j.dld.2026.05.012 · 研究领域：Pancreatitis Pathology and Treatment、IgG4-Related and Inflammatory Diseases、Dermatological and COVID-19 studies

BACKGROUND: Chronic pancreatitis (CP) is a heterogeneous disease in which etiology influences presentation; however, idiopathic forms remain poorly characterized, and conventional approaches may not capture their complexity. AIMS: This study aimed to compare clinical features at diagnosis across etiologies, using unsupervised Multiple Correspondence Analysis (MCA)-based clustering to define distinct groups, focusing on idiopathic subtype. METHODS: A retrospective analysis of a prospectively maintained database of CP patients was performed. Demographics, clinical characteristics, and imaging features categorized based on the TIGAR-O classification were compared. MCA was performed to explore the underlying structure of a series of categorical variables related to CP. RESULTS: 282 patients were included. Toxic was the leading etiology (47.2%), followed by idiopathic (18.8%). Genetic patients were younger and leaner, whereas the toxic group had more men, pulmonary disease and calcifications; idiopathic cases were more often painless and incidentally diagnosed. MCA identified six clusters. Autoimmune and obstructive etiologies formed distinct clusters, while early-onset idiopathic cases (≤ 35 years) mapped near the genetic group, late-onset cases (> 35 years) were positioned between the genetic and the toxic and were characterized by painless incidental presentation. CONCLUSION: Early-onset idiopathic CP clustered near genetic CP, suggesting unidentified genetic components, wherea...