Observer variation in interpreting 18F-FDG PET/CT findings for lymphoma staging.

Abstract

Many studies demonstrate a high accuracy for PET in staging lymphoma, but few assess observer variation. This study quantified agreement for staging lymphoma with PET/CT. The PET/CT images of 100 patients with lymphoma who had been referred for staging were reviewed by 3 experienced observers, with 2 observers reviewing each series a second time. Ann Arbor stage and individual nodal and extranodal regions were assessed. Weighted kappa (kappa(w)) and intraclass correlation coefficient were used to compare ratings. Intra- and interobserver agreement was high for Ann Arbor stage (kappa(w) = 0.79-0.91), number of nodal regions involved (intraclass correlation coefficient, 0.83-0.93), and presence of extranodal disease (kappa = 0.74-0.86). High agreement was also observed for all nodal regions (kappa(w) > 0.60) except hilar (kappa(w) = 0.56-0.82) and infraclavicular (kappa(w) = 0.14-0.55). Lower agreement was observed for bowel involvement (kappa(w) = 0.37-0.71). Experienced observers had a high level of agreement using PET/CT for lymphoma staging, supporting its use as a robust noninvasive staging tool. Further research is needed to evaluate observer variability for restaging during and after chemotherapy.