Lymphoma-associated pancreatitis: a multicenter retrospective cohort study with a prediction model

Purpose Pancreatitis in patients with hematologic malignancies, particularly lymphoma, is a challenging complication with heterogeneous etiologies. Large studies with systematic etiological classification and prognostic models are lacking. Methods This retrospective multicenter study (2020–2025) included patients with hematologic malignancies who developed acute pancreatitis. Etiologies were adjudicated using predefined criteria. A nomogram-based prediction model was developed with internal bootstrap validation. Results Among 1,705 patients, 31 (1.82%) developed pancreatitis: 19 lymphoma (61.3%), 9 leukemia (29.0%), and 3 plasma cell disorders (9.7%). In the lymphoma cohort (median age 59 years, 57.9% male), diffuse large B-cell lymphoma (47.4%) and peripheral T-cell lymphoma (PTCL) (21.1%) were most common. Etiology distribution: treatment-related (36.8%), direct pancreatic infiltration (31.6%), conventional (15.8%), multifactorial (10.5%), and infection-related (5.3%). Endoscopic ultrasound (EUS)-guided tissue acquisition had 83.3% diagnostic sensitivity for direct infiltration; positron emission tomography/computed tomography (PET-CT) showed a “tetrad sign” in 5/6 cases. Severe pancreatitis occurred in 21.1% of lymphoma-associated pancreatitis. The nomogram (incorporating time to pancreatitis ≤30 days, recent L-asparaginase (LAsp)/immune checkpoint inhibitors (ICIs) exposure, and absence of pancreatic mass) discriminated treatment-related from non-treatment-related etiologies with AUC 0.84 (95% CI 0.71–0.96). PTCL subtype (HR 3.12, 95% CI 1.28–7.61, p=0.012), multifactorial etiology (HR 2.56, 95% CI 1.02–6.42, p=0.045), and age ≥65 years (HR 2.01, 95% CI 0.95–4.25, p=0.07) predicted 1-year mortality (C-index 0.79). Summary Lymphoma-associated pancreatitis is rare (1.8% of hematologic malignancies). Treatment-related causes slightly outnumber direct infiltration. EUS-guided tissue acquisition and PET-CT are valuable for differential diagnosis. The nomogram may assist etiological classification and risk stratification, but prospective validation is required.