Aims: To validate and improve the existing algorithm (proposed by Hans et al.) to classify diffuse large B-cell lymphoma (DLBCL).
Methods and results: Tissue microarrays constructed from 81 patients with DLBCL were studied by immunohistochemistry for expression of CD10, Bcl-6, MUM1, Bcl-2, cyclin-D2, FOXP1 and PKC-gamma proteins. Cases were classified as either germinal centre B-like (GCB) or non-GC according to Hans et al. An alternative classification was also employed, in which cases positive for either CD10 or Bcl-6 were considered as a GC subgroup and cases negative for both CD10 and Bcl-6 were considered as a non-GC subgroup. GC was further subdivided into favourable GC (negative for both Bcl-2 and cyclin-D2) and unfavourable GC (positive for either Bcl-2 or cyclin-D2). The 5-year event-free survival (EFS) amongst patients classified as favourable GC versus 'others' was 49.5% and 7.3%, respectively (log rank P < 0.0001). Similarly, the 5-year overall survival (OS) amongst patients classified as favourable GC versus 'others' was 58.6% and 13.7%, respectively (log rank P = 0.0001). The difference in survival was independent of the international prognostic index.
Conclusions: In this group of patients the risk stratification based on the new algorithm was better than that proposed by Hans et al.