Page 89 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

PET/CT interobserver agreement in DLBCL
Figure 3. Example of discrepancy between reviewers’ assessment of stomach on, from left to right, axial attenuation-corrected PET, low-dose CT, and fused PET/CT images.
(A) Baseline 18F-FDG PET/CT with clear localization of lymphoma in stomach. (B) I-PET/CT after 4 cycles of R-CHOP14. Reviewer 1 did not give final DS score and commented that stomach was “DS 4 but could be physiologic uptake”. Reviewer 2 scored this scan negatively (DS 2). (C) EoT-PET/CT after 6 cycles of R-CHOP14. Reviewer 1 still commented on stomach but now scored negatively. Reviewer 2 again scored scan negatively (DS 2).
Discussion
Our study presents the interobserver agreement of DS results for I-PET and EoT-PET from a central review of a large multicenter randomized clinical trial in DLBCL. We found high percentages of OA (88%–92%) and NA (92%–95%) for both I-PET and EoT-PET using a DS of at least 4 for test positivity, at a lower (74%–76%) PA.
Most studies on interobserver agreement primarily report Cohen’s κ and some present percentage of OA in addition. Cohen’s κ is a relative measure, and the values are low in relatively low-prevalence situations (e.g., of residual lymphoma sites). Therefore, we report percentage OA, which is independent of differences in prevalence. In addition, we report specific agreement measures, which reflect the absolute probability that another reviewer gives the same conclusion as a colleague, specified for positive and negative test results [18]. In other words: 74% PA implies that if one reviewer rates an I-PET scan as positive, the probability that another reviewer will provide the same result is 74%.
87
4