Is it time to stop treating subsets of DLBCL with R-CHOP?

Comment on Winter et al, page 4207

Correlative data evaluating BCL6 protein expression and outcome in the US inter-group phase 3 randomized study of R-CHOP versus CHOP demonstrate that CHOP chemotherapy without rituximab is ineffective in BCL6-negative DLBCL.

Diffuse large B-cell lymphoma (DLBCL) is markedly heterogeneous morphologically, clinically, and genetically. With the application of gene expression profiling analyses, DLBCL has evolved into at least 2 entities: germinal center B (GCB)–cell and an activated peripheral blood B-cell (ABC) lymphoma.¹  This separation is important, since the GCB subtype has a survival advantage with cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) chemotherapy.

Based on the results of several phase 3 studies, rituximab plus CHOP (R-CHOP) is standard therapy for DLBCL, but not all patients benefit; prognostic models that can identify patients for either new or risk-adapted treatment are needed.²  Historically, the International Prognostic Index (IPI) has been very useful, but other than identifying a patient's risk it does not suggest alternative treatment. There is a long list of individual biomarkers reported to be significant in DLBCL, but, except for limited information with BCL2, there is a paucity of data in patients treated with rituximab-based chemotherapy.FIG1 

Approximately 50% of DLBCL tumors express the bcl-2 protein. A correlative paper reported by Mounier et al,³  on the randomized GELA LNH 98.5 data set, determined that expression of the BCL2 protein was unfavorable in the CHOP-treated patients. However, this finding was recently refined, with Iqbal et al⁴  showing that expression of the BCL2 protein is unfavorable only in patients without a t(14:18) translocation (ie, those with the ABC subtype).

Between 50% and 75% of DLBCL tumors express the bcl-6 protein. There are inconsistent data concerning failure-free survival (FFS) and the rearrangement of the BCL6 gene. However, more recently, mRNA or protein expression of BCL6 was deemed prognostically favorable in patients receiving CHOP chemotherapy.⁵ 

In this issue of Blood, Winter and colleagues analyze the outcome of patients who were randomly assigned to receive 1 of 4 treatments: R-CHOP, R-CHOP and maintenance rituximab, CHOP, or CHOP and maintenance rituximab by expression of BCL6 or BCL2 protein. Contrary to the GELA data, and using the same 50% cutoff for positivity, BCL2 protein status had no impact upon outcome in this clinical trial. On the other hand, any positive staining for the BCL6 protein predicted for a favorable FFS and OS. Forty-five of the 199 cases were negative for the BCL6 protein; 34 of these patients had chemosensitive disease. Those who received rituximab (77%) had an excellent FFS (75%-89%), similar to BCL6 protein–positive tumors. However, the FFS was only 13% in the small number of patients (only 8) who did not receive any rituximab (see figure).

If biomarkers, as assessed by immunohistochemistry, are to be useful, methodology must be standardized and predetermined cutoffs for positive and negative results have to be uniform. Not all patients with BCL6 protein–positive DLBCL have a favorable GCB subtype. For example, in a report by Hans et al,⁶  using a 30% cutoff for positivity, 18% of DLBCL cases coexpressed BCL6 and MUM1. By definition, these patients have the less favorable ABC subtype, and would receive more rather than less aggressive therapy at our center.⁶  Finally, in the study by Winter et al, although not statistically significant, the best results in the BCL6-positive subgroup were with R-CHOP and maintenance rituximab; an 82% FFS will be difficult to improve upon, or even study in a prospective fashion.

The data resulting from the ECOG 4494 and GELA 98.5 trials must be considered hypothesis-generating and should not be used to decide whether rituximab should be used in the management of DLBCL. Only future randomized trials designed to ask whether rituximab has a particular benefit for DLBCL tumors of the GBC or ABC phenotype can appropriately address this question. ▪