Diffuse large B-cell lymphoma (DLBCL) is the most common lymphoid neoplasm. Three subtypes have been identified by gene-expression studies; germinal center (GC) DLBCL, non-GC DLBCL, and mediastinal large B-cell lymphoma. GC and non-GC DLBCL differ with respect to the cell of origin, pathogenetic mechanisms and prognosis. Gene-expression studies require fresh biopsy material or biopsy material that is frozen immediately after excision. Immunohistology of formalin-fixed, paraffin-embedded biopsies using antibodies against CD10, BCL-6 and MUM-1 allows DLBCL subtypes to be categorized to the GC and non-GC subtype.1
Many studies have shown that the distinction between the cell origin of GC and non-GC subtypes of DLBCL is an independent prognostic factor for patients treated with cyclophosphamide, adriamycin, vincristine and prednisone (CHOP) chemotherapy. However, as the addition of rituximab to CHOP (R-CHOP) improved the outcome of all subgroups of patients with CD20+ DLBCL, a confirmation of the prognostic effect of the cell of origin in DLBCL in the rituximab era is warranted; differences between the subgroups can be assumed to become smaller, which calls in to question the value of the distinction between GC and non-GC DLBCL.
In 2007, a retrospective study in patients with de novo DLBCL, who were treated with CHOP or R-CHOP, showed that by assigning the cases to the GC and non-GC subgroups by immunohistochemistry, the classification of DLBCL according to the cell of origin continues to have prognostic importance in the rituximab era. This analysis in 104 Finnish patients confirmed the improved outcome (that is, improved overall survival and progression-free survival) of patients with immunohistochemically defined GC DLBCL compared with the non-GC phenotype when treated with CHOP only, but this difference was not observed in 90 patients treated with R-CHOP.2 These results are in contrast to a retrospective, immunohistochemistry-based study conducted in the US in 2008, in which 131 patients with de novo DLBCL were treated with R-CHOP, and 112 patients were treated with standard CHOP or CHOP-like therapy only.3 Addition of rituximab improved the 3-year overall survival (85% versus 52%, P <0.001), and event-free survival (67% versus 47%, P = 0.005) of patients with GC DLBCL. Similarly, patients with non-GC DLBCL had improved 3-year overall survival (69% versus 33%, P <0.001), and event-free survival (52% versus 29%, P <0.001) when rituximab was added to CHOP.3 The use of the Hans classifier for the phenotypic distinction between the GC and non-GC type of DLBCL showed a difference in 3-year overall survival, both for the groups treated with chemotherapy only (52% versus 33%), and also for rituximab-treated patients with GC DLBCL (85% versus 69%, P = 0.032).3
It is unclear why the Finnish2 and US3 studies yielded contrasting results. Both studies are limited by a relatively small number of patients, and both studies were retrospective and used historical controls. To date, there is only one study on patients treated within a randomized trial comparing CHOP with R-CHOP.4 In the study by Molina and colleagues, multivariate analysis confirmed that only the International Prognostic Index and treatment arm influence the outcome, but not the immunohistochemically defined GC or non-GC phenotype.4 Unfortunately, of the 399 patients included in the clinical study, immunohistochemistry with tissue microarrays was possible in only 101 patients (25%), which means huge differences in results would be required to reach significance. A small number of patients and historical comparison also limits the validity of a publication from 2008, which assessed 107 patients with GC DLBCL and 93 patients with non-GC DLBCL, where gene-expression profiling confirmed the prognostic difference of the two groups in response to R-CHOP.5
One has to keep in mind that the positive predictive value of the immunohistochemical approach is only 87% for the GC and 73% for the non-GC phenotypes, with a misclassification of 20%.1 Moreover, the Lunenburg Lymphoma Biomarker Consortium observed highly variable results and poor reproducibility for many markers assessed by immunohistochemistry, and the results did not correlate well for BCL-6, which is included in the Hans classifier.6
Even if the data were convincing, would it influence our therapeutic strategies?
In summary, none of the studies assessing the prognostic value of the immunohistological phenotype in the rituximab era has sufficient quality to allow for conclusions. Studies assessing tissue microarray data from a larger number of patients, such as the Lunenburg Lymphoma Biomarker Consortium, which includes patients from different prospective trials,6 and the German High-Grade Non-Hodgkin Lymphoma Study Group (RICOVER-60),7 both of which adhere to rigorous protocols for the immunohistochemical procedures, should hopefully address this issue. Even if the data were convincing, would it influence our therapeutic strategies? We doubt it, because firstly, there is only a small subgroup of patients with such an excellent prognosis that reduction of therapy is justified within prospective trials, such as the ongoing FLYER study of the German High Grade Non-Hodgkin Lymphoma Study Group. This excellent subgroup of young patients is defined clinically (age <60 years, no risk factor according to the age-adjusted International Prognostic Index, no bulky disease) and not by immunophenotype or gene-expression profiling. Secondly, the outcome of all other patients must be improved, but there are no differential improvement strategies for GC and non-GC DLBCL. With the exception of a subgroup of DLBCL with a stromal-2 signature—which emerged from gene-expression profiling of the stromal cells in DLBCL5 that might mirror an angiogenic switch, and calls for studies with antiangiogenic agents such as bevacizumab—no differential therapeutic strategies have emerged to date for DLBCL subgroups. Gene expression and immunohistochemical analyses should be included in randomized trials that are aimed at improving upon R-CHOP treatment, in order to develop differential strategies. Since the studies to date are limited by a small number of patients, the issue of the prognostic effect of the immunophenotypic subclassification of DLBCL remains unsettled and, currently, without therapeutic consequences. For the time being patients with CD20+ DLBCL who do not qualify for, or do not want to participate in such studies, can be treated with R-CHOP without the knowledge of their cell of origin.
Practice point
- Addition of rituximab to CHOP chemotherapy improves both germinal-center and nongerminal-center diffuse large B-cell lymphoma
- It is unclear whether the immunohistochemical distinction between these subtypes is of prognostic significance in the rituximab setting, and it is currently not relevant for the treatment of these patients
