Журнал молекулярных биомаркеров и диагностики

A Novel Immunohistochemistry Method for Predicting Activated B Cell-Like (ABC) Subtype of Diffuse Large B-Cell Lymphoma (DLBCL)

Abstract

Sarah Hersey, Yixin Wang, Alberto Risueno Perez and Fadi Towfic

Background: The ROBUST trial (CC-5013-DLC-002), is a phase 3 randomized, double-blind, placebo controlled, multicenter study to compare the efficacy and safety of Lenalidomide (CC-5013) plus R-CHOP (rituximab plus cyclophosphamide, doxorubicin, vincristine and prednisone) chemotherapy (R²-CHOP) versus placebo plus R-CHOP chemotherapy in subjects with previously untreated activated B-cell (ABC) type Diffuse Large B-cell Lymphoma (DLBCL). The most commonly utilized immunohistochemistry (IHC) algorithm for DLBCL subtyping, developed by Hans, et. al has been shown to have approximately an overall percent agreement of 80% with the gene expression profiling (GEP) classification of DLBCL into the germinal center B-cell–like (GCB) and non-GCB subtypes. The non-GCB subtype includes both the ABC subtype as well as the unclassified or indeterminate subtype which is neither ABC or GCB as defined by gene expression. New antibodies and algorithms specific to ABC tumors have been proposed with an aim to improve the performance of the IHC algorithm for specifically detecting the ABC subtype. This article describes one such new IHC algorithm which could be deployed for ABC subtype determination in clinic.

Methods: We analyzed 100 cases of newly diagnosed DLBCL with CD20, CD10, Bcl-6, MUM1, FOXP1, Bcl-2, Ki-67 and CD5 IHC assays using laboratory developed tests (LDT) and compared different combinations of the IHC assay results to the GEP classification. Statistical analyses were applied to evaluate the possible effect of inter-laboratory and inter-observer variations for the IHC assays and instead of using a decision tree algorithm approach to determine GCB and ABC, a novel approach was taken by using a weighted composite score algorithm. A new IHC algorithm using CD10, Bcl-6, MUM1, and FOXP1 was derived to identify ABC versus non-ABC tumors that closely approximated the GEP classification in this training set. The algorithm was assessed independently and also in conjunction with the Hans IHC algorithm to enhance testing performance. A separate set of 100 independent newly diagnosed DLBCL cases were used to validate the algorithms using LDTs developed independently from two different laboratories using different antibodies, different instrument systems and a total of four pathologists. It should be noted, the LDTs selected for use in this study were not pre-evaluated for performance.

Results: Statistical analyses indicated that the IHC assays and the algorithms for subtyping of DLBCL were robust and reproducible within the range of inter-laboratory and inter-observer variations. For the validation data set, comparing the GEP classification results and the IHC results which was derived using two independently developed LDTs per IHC marker and four pathologists, the new IHC algorithm using CD10, Bcl-6, MUM1, and FOXP1 achieved 81-91% concordance in identifying ABC tumors and 79%-86% concordance in overall classification between the individual pathologists’ calls and the GEP classification. When used in conjunction with the Hans algorithm, the IHC results achieved 89-97% concordance in identifying ABC tumors and 84-89% concordance in overall classification to the GEP results in the validation data set, simulating the predictive power of the GEP classification.

Conclusion: The use of the new IHC algorithm alone and in combination with the Hans algorithm can accurately predict ABC tumors of DLBCL and facilitate subtyping of DLBCL using standard pathology materials and routinely validated IHC assays.