Page 199 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

R-CHOP with lenalidomide in MYC+ LBCL
rearrangement status based on complete pathology/molecular reports. In case of equivocal documentation, FISH assays were repeated at the HOVON Pathology Facility. BCL2 and BCL6 FISH results were completed when sufficient material was available. In cases with sufficient material COO classification was determined by IHC (Hans algorithm) and by using gene expression profiling (Nanostring Lymph2CX assay: raw counts obtained by Nanostring gene expression analysis were uploaded at the Lymphoma/Leukemia Molecular Profiling Project website [2].
Imaging assessments and central PET-CT review
Contrast-enhanced CT scans and 18F-FDG PET scans combined with low-dose CT scans (PET-CT) were performed at baseline,after 3 cycles of treatment (interim PET (iPET-CT)), and at EOT. The EOT PET-CT scan was scheduled 6-8 weeks after the last lenalidomide administration.Treatment response at iPET-CT and EOT PET-CT was assessed according to the Lugano criteria using the visual Deauville 5-point scoring system [3,4]. Deauville scores of 1- 3 were interpreted as CMR, while scores 4 and 5 indicated stable or progressive disease. PET-CT scans were anonymized and uploaded to a Keosys (Imagys) web-based viewing and reporting system and centrally reviewed by two independent experienced nuclear medicine physicians of the HOVON Imaging Working Group who were blinded for survival outcome. In case of discordance, a third reviewer performed adjudication. PET-CT scans were performed and reviewed in compliance with EANM guidelines [5]. Patients with CMR at iPET-CT but with a positive EOT PET-CT scan were classified as progressive metabolic disease (PMD) at EOT, even when the EOT scan was in partial metabolic response (PMR) compared to the pre-treatment PET-CT scan.
Statistical analyses
In order to take the two-stage sampling nature intrinsic to the study design into account, the primary study endpoint was estimated using the method proposed by Jung [6] , which uses the design parameters and the interim analysis results. The design poses a one-sided hypothesis that the response rate is larger or equal to 60%, which we evaluated at a 5% significance level. For the construction of the corresponding two-sided 90% CI the method of Koyama was followed [7]. Both methods are implemented in the R software package “OneArmPhaseTwoStudy” [8]. The secondary survival endpoints were evaluated using the Kaplan-Meier method. Univariate logistic and Cox proportional hazards regression models were
197
7