Reproducibility of International Harmonization Project (IHP)-Based Interpretations of Posttherapy PET/CT in Patients with Hodgkin's and Aggressive Non-Hodgkin's Lymphoma.

IHP-based interpretation of PET/CT post-tx of lymphoma is now widely used but there is little data regarding its reproducibility and accuracy. We determined interobserver variability using IHP criteria and compared its accuracy with semiquantitative approaches using the residual mass (RM) absolute standardized uptake value (SUV_max) alone and the ratio of the RM SUV_max to mediastinal blood pool (MBP) SUV_mean.

Three expert nuclear medicine physicians blinded to patient (pt) outcome independently visually interpreted post-tx PET/CT scans based on the IHP-criteria. RMs were scored on a scale of 0-4, where 3-4 was positive for RMs ≥ 2 cm whereas 1-4 was positive for RMs < 2 cm; scans without RMs were similarly scored, depending on whether there were PET positive or negative. RM SUV_max, MBP SUV_mean and the RM SUV_max to MBP SUV_mean ratio were determined. Since outcome data were available on all 50 pts (median follow-up of pts without progression = 49 months), the positive predictive value (PPV), negative predictive value (NPV) and accuracy for prediction of residual disease/progression were compared between the IHP-criteria and the semiquantitative approaches.

A total of 50 post-tx scans in 50 pts with aggressive NHL (n=24) and HL (n=26) were interpreted by the 3 readers. On a patient-by-patient basis, the 3 readers were in complete agreement with respect to final scan interpretation (i.e., pos/neg) 90% of the time using the IHP-criteria. Percentage of PET+ scans ranged from 20%-30% and PET+ RMs from 45%-55% for the 3 readers. On a per-patient basis, visual criteria resulted in PPVs of 47%-70%; NPVs of 86%-88% and accuracies of 74%-84% for the 3 readers. In contrast, cut-off RM SUV_max of 2.5 or cut-off RM SUV_max to MBP SUV_mean ratio of 1.75 resulted in PPV, NPV and accuracy of 75%, 86% and 84% with areas under the receiver operating characteristic (ROC) curve of 0.765 and 0.787, respectively.

Visual IHP-criteria provide fairly reproducible interpretations among nuclear medicine experts with fair to moderate diagnostic/prognostic accuracy. Compared with the IHP criteria, the semiquantitative approaches provide, at least for some readers, substantial improvement in PPV and overall accuracy. These approaches should, therefore be considered for a more consistent scan interpretations after standardization of the timing of imaging after FDG injection.

No relevant conflicts of interest to declare.