Publikation
Primary Mediastinal B-Cell Lymphoma: Insights from a Multi-Registry Dataset
Sebastian Germer; Christiane Rudolph; Nina Wiegers; Soo-Zin Kim-Wanner; Heinz Handels; AI-CARE Working Group
DKR e.V. Jahrestagung mit Plattform § 65c 2026, Page 35, 6/2026.
Zusammenfassung
Introduction
Primary mediastinal B-cell lymphoma (PMBCL) is a rare, aggressive subtype of non-Hodgkin
lymphoma (NHL) that primarily affects young adults. Although PMBCL is considered a curable
disease with chemotherapy regimens, the prognosis and long-term survival outcomes for
patients may vary. This study aims to use cancer registry data to conduct a retrospective
analysis of patients diagnosed with PMBCL to identify prognostic factors associated with
outcomes and evaluate the effectiveness of different treatment strategies.
Methods
Cancer registry data for patients diagnosed with NHL were requested from 15 federal states.
The study cohort comprised all male and female individuals diagnosed with PMBCL (ICD 10:
C85.2) between 2016 and 2022. Patients older than 90 years at the time of diagnosis and those
with multiple lymphomas diagnosed were excluded. For all eligible cases, age at diagnosis
(AaD), sex, Ann Arbor stage (ANN), and administered treatments within two months after
diagnosis (Treat2m) were extracted. Descriptive statistical analyses were performed, and
multivariate Cox proportional hazard and survival tree analyses were conducted.
Results
Based on our criteria, a total of 694 patients, 395 females (AaD: mean: 43.71, median: 38) and
299 males (AaD: mean: 45.52, median: 40) were retrieved. For 456 patients a Treat2m and for
259 patients an ANN were reported. In the Cox model, AaD and administration of medicinal
therapy within two months were significantly associated with hazard outcomes. Similarly,
survival tree analyses identified AaD as the primary discriminating variable.
Discussion
This tumor entity represents a rare subtype, as reflected by the number of identified cases.
Moreover, because ANN was available for only 37.3% of the cohort, the statistical power to
detect significant stage related associations is reduced, limiting the interpretability of stage
specific analyses. A possible next step here would be the use of data imputation methods.