Limitation of immunohistochemisty in predicting the outcome of patients treated for germinal center diffuse large B-cell lymphoma

Abstract

8604 Background: Although International Prognostic Index (IPI) actually represents the most useful prognostic score in patients (pts) with diffuse large B-cell Lymphoma (DLBCL), pts at same risk could show a different outcome. Germinal center (GC) subtype defined by tissue microarrays in DLBCL has been found as a good prognostic parameter compared to non-germinal center (n-GC) profile. These data, valuated with immunohistochemistry, were confirmed by some but not all authors. Methods: On 105 consecutive pts with DLBCL referred to our institution between 1990 and 2005 and treated with an antracycline based regimen with or without rituximab, we undertook a retrospective immunohistochemical study on 3 μm-thick section obtained from 73 well preserved paraffin block. We analyzed the expression of CD10, Bcl-6, mum1 and Bcl-2 and classified the pts as GC or n-GC according to the Hans’ algorithm. The sample was scored positive for Bcl-6, mum1, CD10 if 30% or more of tumor cells were stained and for Bcl-2 if there was at least 50% of tumor cells positive. Combining Bcl-2 with the GC/n-GC markers we also classified pts in 2 prognostics group as proposed by Muris’ algorithm. Results: We identified 30 GC type DLBCL pts (41%) and 43 n-GC type DLBCL pts (59%) according to the Hans’ algorithm, while 33 GC type DLBCL pts (45%) and 40 n-GC type DLBCL pts (59%) were found with the Muris’ system. Pts characteristics were well balanced between the two subtypes. With a median follow-up time of 29 months, PFS and OS was not different between GC an n-GC pts (p=0.4 and p=0.6 respectively); Same results occurred with the Muris’ algorithm. Multivariate analysis confirmed that only IPI score (p<0.001), bcl-6 (p=0.03) and treatment arm including rituximab (p=0.04) influence the OS and PFS. Although rituximab was delivered only in 20 pts no differences between GC and n-GC were observed dividing the patients according the treatment regimens. Conclusions: The immunohistochemical determination of the GC/n-GC immunophenotype, proposed as a less expensive technique, failed to predict the outcome of DLBCL pts in our study. Randomized trials, including the identification of new markers and the determination of the optimal cut-offs, are needed. No significant financial relationships to disclose.