Interobserver variability in gross tumor volume contouring in non-spine bone metastases

BACKGROUND AND AIM: The optimal imaging test for gross tumor volume (GTV) delineation in non-spine bone metastases has not been defined. The use of stereotactic body radiotherapy (SBRT) requires accurate target delineation. Magnetic resonance imaging (MRI) and/or (18)fludesoxyglucose positron emission tomography (18FDG-PET) allow for better visualization of the extent of bone metastases and optimizes the accuracy of tumor delineation for stereotactic radiotherapy compared to computed tomography (CT) alone. We evaluated the interobserver agreement in GTV of non-spine bone metastases in a single center and compared MRI and/or 18FDG-PET and CT in GTV delineation. METHODS: Anonymous CT and MRI and/or 18FDG-PET obtained from 10 non-spine bone metastases were analyzed by six radiation oncologists at our center. Images acquired by CT and MRI and/or 18FDG-PET were used to delineate 10 GTVs of non-spine bone metastases in the pelvis, extremities, and skull. The cases showed different characteristics: blastic and lytic metastases, and different primary cancers (lung, breast, prostate, rectum, urothelial, and biliary). In both CT and MRI and/or 18FDG-PET, the GTV volumes were compared. The index of agreement was evaluated according to Landis and Koch protocol. RESULTS: The GTV volume as defined on MRI was in all cases larger or at least as large as the GTV volume on CT (P=0.25). The median GTV volume on MRI was 3.15 cc (0.027-70.64 cc) compared to 2.8 cc on CT (0.075-77.95 cc). Interobserver variance and standard deviation were lower in CT than MRI (576.3 vs. 722.2 and 24.0 vs. 26.9, respectively). The level of agreement was fair (kappa=0.36) between CT and MRI. The median GTV volume on 18FDG-PET in five patients was 5.8 cc (0.46-64.17 cc), compared to 4.1 cc on CT (0.99-54.2 cc) (P=0.236). Interobserver variance and standard deviation in CT, MRI, and 18FDG-PET were 576.3 versus 722.2 versus 730.5 and 24 versus 26.9 versus 27.0, respectively. The level of agreement was slight (kappa=0.08) between CT and 18FDG-PET. CONCLUSIONS: Interobserver variance in non-spine bone metastases was equal when MRI and PET were compared to CT. CT was associated with the lowest variance and standard deviation. Compared to CT GTVs, the GTVs rendered from MRI images had fair agreement, while the GTVs rendered from 18FDG-PET had only slight agreement. RELEVANCE FOR PATIENTS: The delimitation of the treatment volume in non-spine bone metastases with SBRT is important for the results determining its efficacy. It is therefore essential to know the variability and to manage it to achieve the highest quality of treatment.