Cell‐of‐origin in diffuse large B‐cell lymphoma: findings from the UK's population‐based Haematological Malignancy Research Network

Diffuse large B-cell lymphoma (DLBCL) is the commonest haematological malignancy, accounting for approximately half of all aggressive B-cell lymphomas. Around 80% of patients present with DLBCL not otherwise specified (NOS), which, although potentially curable with combination therapy (R-CHOP; rituximab, cyclophosphamide, doxorubicin, vincristine, prednisolone), comprises a biologically heterogeneous group that varies widely in terms of clinical characteristics and prognostic factors. The classification of DLBCL NOS into germinal centre B cell (GCB) and activated B cell (ABC) using gene expression profiling (GEP) provided a milestone in the understanding of DLBCL pathogenesis; cellof-origin (COO) is now incorporated into the latest World Health Organization classification, and is a requirement for entry into most contemporary clinical trials (Swerdlow et al, 2017). More recently, in pursuit of molecular-based approaches to the differentiation of Burkitt lymphoma from DLBCL, further subdivisions that include ‘Burkitt-like’ or ‘high-grade’ gene expression profiles have emerged (Dave et al, 2006; Hummel et al, 2006; Sha et al, 2015). Set within the UK’s population-based Haematological Malignancy Research Network (www.hmrn.org), and utilizing both established and potentially extended classifications, the findings reported here are from the largest real-world DLBCL GEP series assembled to date. Full details of HMRN’s methods can be found elsewhere (Smith et al, 2015, 2018). Importantly, initiated in September 2004, and tracking all patients newly diagnosed with a haematological malignancy until death, all diagnoses across HMRN’s 14 hospitals (catchment population ~4 million) are made by specialist haematopathologists at a single integrated haematopathology laboratory – the Haematological Malignancy Diagnostic Service (www.hmds.info). The present report includes data on 2197 patients (aged ≥18 years) newly diagnosed with de novo DLBCL-NOS (International Classification of Diseases for Oncology, third edition, 9680; excluding primary CNS) between 1 September 2004 and 31 August 2016, all of whom were treated with curative intent and were followed-up for mortality through UK-wide national systems until 31 March 2018. Of these, 706 (32 1%) had suitable material available for GEP; which was carried out at HMDS on RNA extracted from formalinfixed paraffin-embedded (FFPE) pre-treatment biopsies using the WG-DASL platform (Illumina, Cambridge, UK) and the ‘DLBCL automatic classifier’ (DAC) to classify COO (Care et al, 2013). The same methods (Barrans et al, 2012; Care et al, 2013) were applied in the recent REMoDL-B Phase III trial, ISRCTN51837425) (Davies et al, 2015). A transcriptomic classifier, originally developed to identify Burkitt lymphoma-like gene expression signatures (Sha et al, 2015), was then employed to further subdivide cases to include a molecular high-grade (MHG) class. The demographic and clinical characteristics of the 706 patients with GEP data are distributed by COO group in Table I; data on the source cohort (n = 2197) are presented on the left. Albeit younger (median age 66 8 years vs. 68 5 years, P < 0 05), the presenting characteristics of patients in the COO study group are broadly similar to those of the cohort as a whole. Furthermore, in both groups around 89% of patients were treated with R-CHOP, and 2– 3% with CODOX-M (cyclophosphamide, vincristine, doxorubicin, methotrexate) based chemotherapies. Survival of patients in the COO study group was, however, significantly better than in the cohort as a whole; the 5-year overall survival (OS) being 66 8% and 61 2% (P < 0 05) respectively, and relative survival (RS), which takes into account the underlying age-specific and sex-specific mortality in the population as a whole, were 76 0% vs. 71 1% (P < 0 05). The standard 3-group classifier assigned 384 (54 4%) patients to GCB, 194 (27 5%) to ABC; 128 (17 1%) were unclassified. As in other series (Scott et al, 2015), patients in the GCB group were significantly (P < 0 05) younger (median age 66 0 years), had better survival (5-year OS 72 9%), and were more likely to have a MYC gene rearrangement (MYC-R, 12 1%) than those in the ABC group (median age 70 5 years, 5-year OS 53 7%, MYC-R 5 5%); the remaining prognostic characteristics in the two groups are comparable. Burkitt lymphoma displays germinal centre B cell gene expression characteristics (Swerdlow et al, 2017); accordingly it is perhaps not surprising that members of the MHG subgroup were, almost exclusively, identified as GCB by the 3group classifier (46/50). Separation of these cases widened the survival disparity between the ABC and GCB groups (Fig 1); the 5-year OS being 76 9%, 54 4%, 41 8% and 68 3% in the GCB, ABC, MHG and unclassified groups respectively. Indeed, the survival of patients in the MHG group is substantially worse than that of those remaining in the GCB group (P < 0 001), and significantly worse than Correspondence