Applicability of Different Immunohistochemistry Algorithms to Assess Gene Expression Profile In Patients with Diffuse Large B-Cell Lymphoma

Abstract 4134

Gene expression profile (GEP) allows to distinguish two groups with different origin in patients with diffuse large B-cell lymphoma (DLBCL): germinal-center (GC) and activated (ABC), with the latter having a significantly poorer outcome. However, GEP is a technique not available in current clinical practice. For this reason, attempts to reproduce GEP data by immunophenotyping algorithms have been made. The aim of this study was to apply the most popular algorithms in a series of patients with DLBCL homogeneously treated with immunochemotherapy, in order to assess the correlation with GEP data and their usefulness to predict response and outcome of the patients. One hundred fifty seven patients (80M/77F; median age 65 years) diagnosed with DLBCL in 5 institutions of the Grup per l'Estudi dels Limfomes de Catalunya I Balears (GELCAB) during a 5-year period, treated with Rituximab-containing regimens (in most cases, R-CHOP), in whom histological material to construct a tissue microarrays (TMA) was available, constituted the subjects of the present study. Four algorithms were applied: Colomo (Blood 2003, 101:78) using CD10, bcl-6 and MUM1/IRF4; Hans (Blood 2004, 103:275) using CD10, bcl-6 and MUM1/IRF4; Muris (J Pathol 2006, 208:714) using CD10 and MUM1/IRF4, and Choi (Clin Cancer Res 2009, 15:5494), using CD10, bcl-6, GCET1, FOXP1 and MUM1/IRF4. The thresholds used were those previously described. GEP studies were performed in 62 patients in whom fresh frozen material was available. Main clinical and evolutive data were recorded and analyzed. The proportion of positive cases for the different single antigens was as follows: CD10 26%, bcl-6 64%, GCET1 46%, FOXP1 78% and MUM1/IRF4 28%. The distribution of cases (GC vs. non-GC) according to the algorithms is detailed in the table. In 88 of 110 patients (80%) with all the antigens available, the patients were allocated in the same group (either GC or non-GC). When the immunochemistry was compared with GEP data, the sensitivity in the GC group was 59%, 52%, 70% and 40% for Colomo, Hans, Muris and Choi algorithms, respectively. The sensitivity in the non-GC group was 81%, 85%, 62% and 84%, respectively. On the other hand, the positive predictive value (PPV) in the GC group was 81%, 83%, 72% and 77%, respectively. In non-GC subset the PPV for the different algorithms was 59%, 55%, 72% and 52%, respectively. We observed a higher percentage of misclassified cases in the GC-phenotype subset than in the non-GC subgroup. None of the immunohistochemical algorithms showed a significant superiority as surrogate of GEP information among the others. The ability of GEP groups as well as of groups defined by the algorithms to predict complete response (CR) rate, progression-free survival (PFS) and overall survival (OS) of the patients is showed in the table. Thus, whereas the GEP groups showed significant prognostic value for CR rate, PFS and OS, none of the immunohistochemical algorithms were able to predict the outcome. In conclusion, in a homogeneous series of DLBCL patients treated with immunochemotherapy, the different immunohistochemical algorithms were not able to mimic the GEP information. The prognostic impact of the groups defined by immunohistochemistry (GC vs. non-GC) was particularly low.