Molecular profiling improves classification and prognostication of nodal peripheral t-cell lymphomas: Results of a phase III diagnostic accuracy study

Abstract

Purpose The differential diagnosis among the commonest peripheral T-cell lymphomas (PTCLs; ie, PTCL not otherwise specified [NOS], angioimmunoblastic T-cell lymphoma [AITL], and anaplastic large-cell lymphoma [ALCL]) is difficult, with the morphologic and phenotypic features largely overlapping. We performed a phase III diagnostic accuracy study to test the ability of gene expression profiles (GEPs; index test) to identify PTCL subtype. Methods We studied 244 PTCLs, including 158 PTCLs NOS, 63 AITLs, and 23 ALK-negative ALCLs. The GEP-based classification method was established on a support vector machine algorithm, and the reference standard was an expert pathologic diagnosis according to WHO classification. Results First, we identified molecular signatures (molecular classifier [MC]) discriminating either AITL and ALK-negative ALCL from PTCL NOS in a training set. Of note, the MC was developed in formalin-fixed paraffin-embedded (FFPE) samples and validated in both FFPE and frozen tissues. Second, we found that the overall accuracy of the MC was remarkable: 98% to 77% for AITL and 98% to 93% for ALK-negative ALCL in test and validation sets of patient cases, respectively. Furthermore, we found that the MC significantly improved the prognostic stratification of patients with PTCL. Particularly, it enhanced the distinction of ALK-negative ALCL from PTCL NOS, especially from some CD30< PTCL NOS with uncertain morphology. Finally, MC discriminated some T-follicular helper (Tfh) PTCL NOS from AITL, providing further evidence that a group of PTCLs NOS shares a Tfh derivation with but is distinct from AITL. Conclusion Our findings support the usage of an MC as additional tool in the diagnostic workup of nodal PTCL.