Page 137 - 89Zr-Immuno-PET:Towards a Clinical Tool to Guide Antibody-based Therapy in Cancer

Interobserver reproducibility of 89Zr-immuno-PET
considering the lower signal to noise ratios for 89Zr-immuno-PET compared to 18F-FDG-PET. For example, interobserver variability of 14% for SUVmean has been reported for manual tumor delineation of pulmonary lesions on 18F-FDG-PET (12).
For 89Zr-immuno-PET, this is the first study to report interobserver reproducibility of tumor uptake measures. Several factors should be considered to determine to which extent these results are generalizable. Interobserver reproducibility was determined for three different 89Zr-labeled mAbs (rituximab, cetuximab and trastuzumab), at different time points (D3, D4, D6) and different injected doses (74MBq for 89Zr-rituximab vs 37MBq for 89Zr-trastuzumab and 89Zr-cetuximab). This study was not designed to assess how these factors individually impact interobserver variability. Instead, the results obtained reflect a broad range of uptake characteristics, which can be used as a general estimate of the measurement error due to interobserver variability in VOI delineation. Future, larger studies can focus on factors that influence tumor contrast (e.g. tumor localization, differences in uptake characteristics between mAbs).
Although ICC are reported, reliability is dependent on the range in tumor 6 uptake and therefore not directly generalizable to other studies. In addition, tumor
uptake and interobserver variability are influenced by the disproportionate high
number of lesions in patient 2. Therefore, ICC values for this lesion-based analysis
cannot be applied to determine whether we can reliably detect differences between patients. Improved tumor contrast, in combination with a broad range in tumor uptake, is expected to result in improved interobserver reproducibility for all tumor uptake measures.
Another aspect to consider is that all observers used the same quantification software and a standardized operating procedure (no use of thresholds or fixed size VOI). Use of different software platforms without a standardized procedure may result in lower interobserver reproducibility. In addition, generalizability could be hampered if the three observers would have read the images in a systematically different way. In this study, there was no indication for such a systematic difference between the three observers.
These results suggest that interobserver agreement for SUVmean is sufficient to consider this uptake measure to quantify tumor uptake in a larger tumor area (opposed to only the maximum voxel or very small sample of the tumor as defined by SUVpeak). However, manual tumor delineation is a laborious task. As the concept of total lesion mAb uptake is of interest, the feasibility of semi-automatic VOI
135