Germinal Center Phenotype Determined by Immunohistochemistry on Tissue Microarray Does Not Correlate with Outcome in Diffuse Large B-Cell Lymphoma Patients Treated with Immunochemotherapy in the Randomized Trial LNH98-5. A GELA Study.

Germinal center (GC) subtype defined by immunohistochemistry in diffuse large B-cell lymphoma (DLBCL) treated by anthracycline-containing regimens has been found as a good prognostic parameter by some but not all authors. More recently, studies reporting series of patients treated with rituximab plus chemotherapy showed a loss of prognostic value of GC profile. However, these series did not include patients from randomized trials. We undertook an immunohistochemical study on tissue microarray (TMA) among patients included in the GELA LNH98-5 trial that compared CHOP and rituximab-CHOP (R-CHOP). Among the 399 patients (age 60 to 80 years) enrolled in this study with DLBCL, duplicate TMAs were available for 172 patients, consisting for each TMA of two cores of 0.6mm per case. We analyzed the expression of CD10, bcl-6 and mum-1 and classified patients as GC or non-GC (n-GC) according to the Hans’ algorithm. The samples were scored positive for bcl-6 and mum-1 if 30% or more of tumor cells were stained and for CD10 if there was a significant number of tumor cells positive. Internal controls for each spot were required for interpretation as well as concordance between the duplicate TMAs. T-cell-rich B-cell lymphomas (TCRBCL) were excluded because too few tumor cells could be evaluated. Overall 101 patients were evaluable (excluding 39 cases due to absence of internal controls or spots detachment, 25 discordant cases between the duplicates for bcl-6 and mum-1, and 7 TCRBCL). Patient characteristics were not statistically different from the whole series regarding sex, age, stage, LDH level, PS, number of extranodal sites, aa-IPI, IPI, median follow-up (85 months) as well as allocated regimens (CHOP, 49 patients; R-CHOP, 52 patients). 44 patients were classified as GC and 57 as n-GC. Patient characteristics were well balanced between GC and n-GC subtypes. EFS was not different among the 101 patients between GC and n-GC (p=0.28). Multivariate analyses confirmed that only aa-IPI (p=0.005), and treatment arm (p=0.005) influence EFS but not the GC/n-GC profile (p=0.17). Although the EFS among CHOP patients showed a borderline but not significant better prognosis among GC patients (p=0.11), this trend was completely abrogated among patients treated with R-CHOP (p =0.92). Overall survival was not different between the GC and n-GC profile among the whole population (p=0.41), the patients treated by CHOP (p=0.31) or by R-CHOP (p=0.74). Overall, our study demonstrates that in this randomized trial, GC profile defined by immunohistochemistry on TMA does not represent a significant prognostic parameter among DLBCL patients treated by CHOP with or without rituximab.