OBJECTIVETo investigate the relationship between the clinical features, pathogenesis, immunophenotype, different classification models and prognosis in Chinese patients with diffuse large B-cell lymphoma (DLBCL).

METHODSA total of 147 patients with DLBCL who were treated with CHOP-like or R-CHOP were subjected to analysis. Standard two-step Envision method of immunohistochemical staining was used to assess the expression of CD10, Bcl-6, MUM1, FOXP1, GCET1, CD5, Bcl-2, Ki-67, then according to Hans algorithm, Choi algorithm and Molecular markers, we compared the differences of their prognoses.

RESULTS(1) Kaplan-Meier univariate analysis of the clinical data of 147 DLBCL patients found that the 3-year overall survival (OS) rates were better in early stage (P=0.032), low IPI score (P=0.001), less than one extranodal involvement (P=0.014), and complete remission (P<0.01). The prognoses had no significant difference in terms of gender, age, LDH, B symptoms and treatment options (P value> 0.05). (2 )For Hans model, GCB group had 42 cases, the ABC group 85 cases; GCB were 47 cases, ABC 80 cases (according to Choi model). Choi model suggested GCB subtype showed much better prognosis than ABC subtype (P=0.047), while Hans model shed no statistically significant difference (P=0.285). (3) Ki-67 of 75% was found to significantly discriminate patients with good or bad prognosis. In R-CHOP group at the same time, low Ki-67 (P=0.017) and CD5-negative groups (P=0.012) were better. Cox proportional hazards regression model showed that IPI score (P=0.002) and Ki-67 (P=0.019) were independent adverse prognostic factors.

CONCLUSIONThe Ann Arbor stage, IPI score, extranodal involvement status and Ki-67 were significantly associated with prognosis .Compared to Hans algorithm, Choi had an advantage to predict the different prognosis between subtypes, and ABC group had poor outcome. Finally, both Ki-67 and IPI score were independent adverse prognostic factors.