Page 81 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

PET/CT interobserver agreement in DLBCL
Introduction
Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of malignant lymphoma, accounting for 30%–40% of non-Hodgkin lymphomas [1]. Current international guidelines [2,3] recommend 18F-FDG PET before therapy in typically 18F-FDG– avid lymphoma types—for example, Hodgkin lymphoma and DLBCL [4]—and to apply the Lugano response classification based on the Deauville score (DS) on a 5-point scale at the end of treatment. Application of 18F-FDG PET during therapy (interim-PET, or I-PET) allows PET-guided patient management, with success in Hodgkin lymphoma [5–8]. In DLBCL, the value of I-PET is less clear [9]: most I-PET–adapted trials in DLBCL did not demonstrate a strategy that overcomes treatment resistance [10], except for a phase II study with intensification after rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) at a 14-d cycle (R-CHOP14) in I-PET– positive patients to R-ICE and Z-BEAM autologous stem cell transplantation [11]. Therefore, I-PET is currently not used in clinical practice. An important prerequisite for these PET-guided studies is a consistent classification of the I-PET scans into a positive or negative category.
Similar to other disciplines, observer variation is the Achilles’ heel of radiology [12]. In the DS scoring system, the 18F-FDG uptake in potentially malignant tissue is rated versus normal 18F-FDG distribution in mediastinal blood pool and liver. Such a semiquantitative approach is less prone to observer variation than visual readings purely based on perception, knowledge, experience, and pattern recognition [12], possibly influenced by optical illusion effects [13]. There are few studies on interobserver agreement of DS in PET scans in DLBCL patients treated with rituximab-containing chemotherapy, reporting a 0.4–0.8 range of κ-values for I-PET [14–16] and 0.5 for DS in end-of-treatment PET (EoT- PET) [15].
In clinical practice and trials, it is essential to know the specific agreement, that is, the absolute probability of obtaining the same test result by different reviewers rating the same scan. In cases of I-PET–driven treatment escalation on a positive I-PET scan or I-PET– driven treatment deescalation on a negative I-PET scan, observer variation–driven misclassification might compromise the results of clinical trials or induce overtreatment or undertreatment, respectively.
79
4