Page 91 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

PET/CT interobserver agreement in DLBCL
organs. In an intra- and interobserver agreement study of baseline PET/CT from a mix of lymphoma subtypes, the lowest weighted κ-values were observed in hilar nodes, infraclavicular nodes, and bowel [22]. However, these sites were also those sites that were least frequently involved with lymphoma in their cohort, thus κ-values could have been low because of the low prevalence of these sites. For some specific nodal and extranodal localizations, only a few positive cases were identified; therefore, we decided to report on numbers of discrepancies and percentage OA only.
Baseline PET/CT provides more accurate staging than CT only [2,3] and serves as a reference to quantify tumor response (SUV, metabolically active tumor volume). In our study, in which baseline PET/CT was not mandatory according to prevailing guidelines at the start of the study [23,24], we found that observer variability in treatment evaluation was independent of the baseline imaging modality.
A strength of this study is that the I-PET and EoT-PET scans were assessed by 2 reviewers from a pool of 10, in contrast to previous studies with small, fixed numbers of reviewers. Scoring by a pool of reviewers represents the normal situation of 18F-FDG PET/CT lymphoma response assessment in clinical practice. A limitation is that scans rated as unclear were excluded from our main analyses; in 13 I-PET and 10 EoT-PET scans one of the reviewers rated the final conclusion as unclear, specifying in the free-text section that they were not certain that the residual 18F-FDG uptake was lymphoma-related. A similar conclusion was drawn by both observers in 2 EoT-PET scans. In analysis of a best-case (both reviewers agreed on negative or positive scores) and worst-case (discrepancy) scenario, we found that these results only slightly affected observer agreement: for I-PET and EoT-PET, percentages of OA were 88.1% and 91.9% in the best-case scenario, versus 85.4% and 89.7% in the worst-case scenario. In most of these unclear scans—13 of 13 I-PET scans and 6 of 12 EoT-PET scans residual 18F-FDG uptake in extranodal lymphoma sites caused the uncertainty, especially skeletal lesions in I-PET scans (perhaps because of enhanced bone marrow background uptake due to granulocyte colony-stimulating factor, or healing fractures or bone remodeling in previous lymphoma locations). Other reasons mentioned for unclear reads were missing baseline PET status, no contrast-enhanced CT scan available, inferior quality of a CT scan, a possible sarcoid like response, or uncertainty about a nonresponding lesion while all other lesions responded (Figs. 2 and 3). Another
89
4