Interobserver Variability in Sodium Fluoride Bone Scan Interpretation
Recommendations to Reduce Potential Errors in Classification and Treatment in Patients with Prostate Cancer
Bone scanning is an established imaging modality in the investigation of patients with prostate cancer (PCa). Traditionally, scanning has been performed on a gamma camera, utilizing technetium-labeled diphosphonates, such as 99mTc-methylene diphosphonate (MDP). In recent years, 18Fluorine-labeled sodium fluoride (18F-NaF), utilizing Positron Emission Tomography/Computed Tomography (PET/CT), has assumed an ever-growing role in the evaluation of osseous metastasis. PET/CT is performed on hybrid scanners which combine the advantages of metabolic/physiologic imaging (PET) and anatomic imaging (CT). This study was prompted by our observation of discordance between the clinical and biochemical findings and reported results of NaF bone scintigraphy in a group of our patients. These patients were under consideration for salvage therapy of localized PCa and/or assignment to either the M0 (no evidence of metastasis) or M1 (metastasis present) PCa. In this study we assess interobserver variability in NaF PET/CT scan interpretation, its effect on patient staging, and offer recommendations to improve reader performance.
Methods: A retrospective interpretation was performed of all patients who underwent NaF PET/CT bone scanning from April, 2012-January, 2013. Studies were re-read blindly by a board certified nuclear medicine physician, experienced in NaF PET/CT bone imaging (MK, reader A), and compared to the initial report, issued by one of three board certified radiologists (primary readers). Patients with discordant interpretations were then read blindly by a second board certified nuclear medicine physician, also experienced in NaF PET/CT (SO, reader B). All studies were performed on a 16-row PET/CT system (Discovery ST; GE Healthcare), following the intravenous administration of 370 MBq (10 mCi) of NaF. Manufacturer recommended processing and display were utilized.
Results: Seventy studies were included in the data base. Of these, 16 studies in 15 patients showed discordant results between the primary readers and reader A. Of these 16 studies, readers A and B were in agreement on 13, and had minor differences in recommendations for follow-up in the other 3. Based upon reader A and B agreement, 11/70 (16%) patients who were initially reported as M1 were felt to be M0. Two additional studies were identified in which there was marked discrepancy in the extent of osseous disease between the interpretation of the primary reader and readers A and B. Patient charts were then evaluated to determine the subsequent course of these patients. Available data were consistent with the findings of readers A and B.
Conclusions: Significant interobserver variability may exist in the interpretation of NaF PET/CT bone scans. In our study, we observed nearly a 20% discordance between the initial read and rereads. This may result in misclassification of patients with PCa with significant consequences. Experienced readers appear less likely to erroneously place patients in the M1 category. Suggestions to minimize interobserver variability are provided to help avoid this potential pitfall.