RESUMEN
AIM: To evaluate the use and reliability of the PET-based response criteria for interim PET (iPET) in terms of interobserver variability in pediatric and adolescent patients suffering from non-Hodgkin´s lymphoma (NHL). Particular attention was given to the identification of visual cutoff to separate patients with a favourable outcome. PATIENTS, METHODS: Retrospective analysis of PET-datasets of 18 children and adolescents suffering from NHL who underwent iPET after two cycles of chemotherapy for response assessment. Datasets were evaluated and rated in three independent review centers (RC) (blinded-read, intra-center consensus) using a visual 5-point response scale. Ratings were compared to clinical outcome. Pairwise interobserver agreement was analysed with Cohen's kappa-test (κ). Overall agreement (between attended RCs) was assessed with Fleiss' κ-test. RESULTS: Four patients suffered relapse (early, n = 2; late, n = 2). Per region analyses on interobserver variability revealed a "substantial" agreement (Fleiss' κ = 0.618). Per patient analyses revealed concordant iPET-ratings in eight patients: iPET-negative (iPET-), n = 5; iPET-positive (iPET+), n = 2; iPET-inconclusive (iPET±), n = 1. Discordant ratings were found in the remaining patients. Patients with early relapse were concordantly identified using mediastinal blood pool structures (MBPS, score ≥ 3) as visual cutoff between iPET+ or iPET-, respectively. However, patients with late relapse were not concordantly identified taking the MBPS as visual cutoff. CONCLUSION: The iPET interpretation using a dedicated PET-based response scale assured a low interobserver variability in per-region but not in per-patient analyses in a multicenter read. Using a sensitive read out (iPET+, score ≥ 3) a reliable identification of patients suffering relapse was limited to those with early relapse.