Determination of metabolic tumor response by FDG PET is highly reproducible with the same or different readers pre- and post-therapy

Objectives To compare the reproducibility of percent (%) change in the standardized uptake value (SUV) when PET scans are read by the same vs. different readers pre- and post-therapy.

Methods Maximum SUV of 52 tumors was determined independently by 4 PET readers on FDG PET/CT pre- and post-therapy. %-change in SUV between scans was calculated using 1) the same reader pre- and post-therapy and 2) 12 combinations of different readers pre- and post-therapy. EORTC PET criteria were used to determine metabolic response of each tumor. Reproducibility of %-change in SUV was described by the coefficient of variation (COV). The kappa statistic was used to describe inter-rater agreement in response classifications.

Results The means (± sd) of the average %-change in SUV for all tumors with the same (-45±34%) vs. different readers (-43±38%) pre- and post-therapy were not different (p=0.87) and were strongly correlated (r=0.94, p<0.01). No significant difference in mean COV was found for %-change in SUV determined by the same (16.7±36.2%) vs. different readers (14.3±69.4%; p=0.90) pre- and post-therapy. Agreement of metabolic response classification was excellent for the same (kappa 0.937) and different readers (kappa 0.907) pre- and post-therapy.

Conclusions Determination of %-change in SUV between pre- and post-therapy scans is highly reproducible regardless if the same or multiple readers measure SUV at different time points of assessment. Classification of tumor metabolic response based on changes in SUV was also highly reproducible. Across multiple readers, SUV is a robust parameter for tumor response assessment.