18F-FDG-PET/CT for polymyalgia rheumatica: agreement and diagnostic accuracy of routine PET scan report vs. standardized PMR PET scores

BACKGROUND: 18F-FDG-PET/CT may reveal widespread inflammation of musculoskeletal structures in polymyalgia rheumatica (PMR). Currently, scans are subjectively analysed based on the overall gestalt of the scan. Standardized PET scores may potentially aid the interpretation of the scans for suspected PMR. Here, we compared the agreement and diagnostic accuracy of routine PET scan reports vs. the most validated PET scores for PMR.

METHODS: 68 consecutive patients with suspected PMR (treatment-naïve, n  = 29; already treated, n  = 39) undergoing 18F-FDG-PET/CT were included. In glucocorticoid-treated patients, complete tapering was pursued prior to the scan. Conclusions of routine PET scan reports were interpretated by three independent readers as "PMR", "not PMR" or "unclear". The Leuven and Leuven/Groningen scores were determined. Agreement of scan report interpretation, and agreement of routine scan reports and PET scores were determined. Sensitivity and specificity were determined for the routine scan report and the two scores, with the clinical diagnosis established after 6 months follow-up as the reference standard.

RESULTS: A diagnosis of PMR was made in 45/68 patients. Routine scan reports were uniformly rated by all three readers in 54 (78%) cases. Following a consensus meeting, scans were rated as "PMR" in 43 cases, "unclear" in 10 cases and "not PMR" in 15 cases. The routine scan report showed a sensitivity of 82% and specificity of 74%, if "unclear" cases were considered negative for PMR. The Leuven and Leuven/Groningen Scores showed similar diagnostic accuracy. Agreement between the routine scan report and PET scores was good (Cohen's kappa 0.60-0.64), if "unclear" cases were excluded from the analysis. Among 8/10 "unclear" cases, the PMR PET Scores accurately distinguished between PMR/PMR-mimicking inflammatory conditions and non-inflammatory conditions. Agreement and diagnostic accuracy of routine scan reports and PET scores were better among treatment-naïve patients than those that had been treated previously.

CONCLUSION: Our study reveals that routine PET scan reports for suspected PMR can be interpreted differently between readers. Although the routine PET scan reports and PMR PET scores did not always agree, they demonstrated similar diagnostic accuracy, with the highest accuracy observed in treatment-naive patients. The Leuven and Leuven/Groningen score could especially be helpful for cases in which the nuclear medicine physician is uncertain.