Page 90 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

Chapter 4
Similar studies presented different agreement measures [14–16]. Itti et al. [14] (n = 114, 3 readers) reported pairwise Cohen’s κ-values (0.53–0.80) for I-PET after 2 cycles of R-CHOP or R-ACVBP in a retrospective cohort but did not report specific agreement measures. From their presented data, we calculated OA of 77%–90% between observer pairs, subdivided into a NA of 81%–91% and PA of 72%–89%. Horning et al. [16] (n = 38 patients, 3 readers) reported a Fleiss’ κ of 0.50 and percentage OA of 71% for I-PET after 3 cycles of R-CHOP, but specific agreement measures could not be extracted. Han et al. [15] reported κ-values of 0.41–0.52 and OAs of 82%–88% for I-PET (n = 55, after 3 cycles of R-CHOP) and EoT-PET (n = 57), respectively, as assessed by 2 readers. NA and PA as extracted from their presented data were 89% and 50% for I-PET and 92% and 59% for EoT-PET, respectively.
Taken together, it appears that NA was generally above 80% in all studies (probably at least partly related to the high prevalence of negative scans). However, PA seemed to have a wider range between studies. In our study, we found a Cohen’s κ -value of 0.65 and 0.71 for I-PET and EoT-PET, respectively.
Our data suggest that I-PET is more difficult to assess than EoT-PET. We found that the percentage of OA was lower for I-PET than for EoT-PET. The trend toward a lower NA for I-PET than for EoT-PET, could (in part) be caused by the higher number of negative scans at the end of treatment. In the study from Han et al., agreement measures also seemed generally higher for EoT-PET than for I-PET [15]. Treatment-related inflammation shortly after chemotherapy might hamper the identification of lymphoma-related 18F-FDG uptake.
In addition, we explored observer agreement as a function of disease location. Related to initially involved sites, we found the lowest percentages of OA for mesenteric, gastrointestinal, and skeletal sites in I-PET and EoT-PET. In these tissues, the local background of 18F-FDG varies between and within patients over time; uptake due to intercurrent inflammation and, for example (healing), pathologic fractures needs to be accounted for, and this is not always covered by the Lugano criteria. In I-PET, discrepancies in spleen and Waldeyer’s ring were more common. The short interval between the I-PET exams after the previous R-CHOP14 course [20] and the recent administration of granulocyte colony stimulating factor [21] in our study could cause false-positive uptake in these
88