Predicting Cell of Origin and Survival in R-CHOP Treated De Novo Diffuse Large B-Cell Lymphoma – A Comparison of Immunohistochemical Algorithms
Graham W Slack, King L Tan, David W Scott, Laurie H Sehn, Joseph M Connors, Randy D Gascoyne. BC Cancer Agency, Vancouver, BC, Canada
Background: Gene expression profiling (GEP) divides diffuse large B-cell lymphoma (DLBCL) into biologically and clinically distinct groups based on cell of origin (COO) gene signatures. Germinal center like lymphoma is associated with better overall (OS) and progression-free (PFS) compared to activated B-cell like and unclassifiable lymphomas. Several immunohistochemical algorithms have been developed that predict COO and survival in CHOP treated DLBCL; however, the usefulness of these algorithms in R-CHOP treated de novo DLBCL remains controversial.
Design: 184 cases of formalin-fixed paraffin-embedded R-CHOP treated de novo DLBCL in a tissue microarray were independently evaluated by two pathologists for expression of CD10, BCL6, MUM1, GCET1, FOXP1, LMO2 and BCL2 (Dako 124 and Epitomics E17). Cases were assigned a COO immunophenotype using the following algorithms: Choi, Hans, Muris, Natkunam, Nyman and Tally. Clinical data were available for all cases. GEP data were available for 50 cases.
Results: The Tally algorithm had the highest level of concordance with GEP. Choi and Hans also showed high concordance with GEP, which was maintained when BCL6 was removed from the algorithms. The Muris algorithm showed the lowest level of concordance and results were not influenced by use of either BCL2 antibody. Tally, Natkunam and Choi were the only algorithms significantly associated with OS and PFS; the Hans algorithm did not predict survival. Multivariate analysis adjusting for International Prognostic Index score showed the Tally, Natkunam and Choi algorithms remained significant independent predictors of PFS but not OS.
|Algorithm||Concordance with GEP (%)||Overall Survival (p-value)||Progression-Free Survival (p-value)|