A survival prognostic model for gastro-intestinal diffuse large B-cell lymphomas: A multicenter retrospective study.

Primary gastro-intestinal lymphomas (PGIL) are the most common extra-nodal lymphomas. Currently, existing prognostic models for gastrointestinal diffuse large B-cell lymphoma (GI-DLBCL) like the HLAMA model is lack of rigorous external validation. In this study, we conducted a multi-center retrospective investigation to establish a prognostic model for PGI-DLBCL. A total of 1 023 consecutive newly-diagnosed patients with PGIL were recruited from 9 medical centers in China. Among them, data of 462 patients with diffuse large B-cell lymphoma (DLBCL) who received R-CHOP treatment were analyzed, which was used as the training cohort. Additionally, data of 192 PGl-DLBCL patients from the other 3 medical centers were collected and used as the external validation cohort. A risk prediction model was constructed using the coefficients of weighted independent prognostic co-variates from the multi-variate analysis. It was found that patients with PGI-DLBCL receiving R-CHOP had better survival than those receiving CHOP (P < 0.001). In multi-variate analysis of the training cohort with PGI-DLBCL receiving R-CHOP, Age (P = 0.03), Lugano stage (P < 0.001), Hemoglobin (P = 0.04) and Lactic dehydrogenase (LDH) concentrations at diagnosis (P < 0.001) were significantly-associated with survival. Using these co-variates we constructed a survival prognosis nomogram model, resulting in C-statistics of 0.83 (0.79, 0.88) and 0.81 (0.73, 0.88) in the training cohort and the validation cohort, respectively. Our model, which includes 4 variables (Age, Lugano stage, LDH and Hemoglobin), is efficient in predicting survival of patients with PGI-DLBCL.