Observer agreement of treatment responses on planar bone scintigraphy in prostate cancer patients: importance of the lesion assessment method

R. F. FONAGER; H. D. ZACHO; S. ALBERTSEN; J. FLEDELIUS; J. A. EJLERSEN; M. H. CHRISTENSEN; R. ALEKSYNIENE; J. A. BIURRUN MANRESA; L. J. PETERSEN

Barcelona

Congreso; 29th Annual Congress of the European Association of Nuclear Medicine; 2016

European Association of Nuclear Medicine

Aim: The aim of this study was to assess observer agreement in the evaluation of treatment responses of bone metastases using three different scoring methods in prostate cancer patients. Materials and methods: A total of sixty-three paired bone scans (BSs) were selected from prostate cancer patients with two or more BSs within twelve month at Aalborg University Hospital from January 2009 to November 2014. BS was performed before initiation of a therapy and minimum twelve weeks within treatment. Experienced nuclear medicine physicians, blinded to clinical and laboratory data, evaluated the whole-body BS by three different methods, A) standard clinical assessment (improved, stable, or worse), B) MD Anderson criteria (complete response, partial response, stable disease, or progression), and C) Prostate Cancer Working Group 2 criteria (non-progressionversus progression, defined as two or more new lesions). Individual BS were evaluated for presence (M1) or absence (M0) of bone metastases, superscan (absent/present), and number of lesions per region. Observer agreement was assessed by Cohen?s kappa and reported by Landis & Koch strength of agreement terminology. Results: There was substantial agreement on bone response when using Prostate Cancer Working Group 2 criteria (Cohen?s kappa at 0.84, 95% CI 0.69-0.99). Cohen?s kappa for standard clinical assessment and MD Anderson criteria were 0.52 (95% CI 0.36-0.69) and 0.56 (95% CI 0.40-0.71), respectively, corresponding to moderate agreement. However, the latter methods have more response categories and this is known to affect agreement negatively. Excellent agreement was demonstrated with regards to M0/M1 (Cohen?skappa 0.94, 95% CI 0.82-1.00), and substantial agreement was found in assessment of superscan (Cohen?s kappa 0.78, 95% CI 0.49-1.00). BlandAltman analysis showed a large variation in the assessment of number of lesions per region, e.g. in columna, which showed the largest variation, limits of agreement (Bland Altman analysis) ranged from -14 to 15, with median number of lesions of 7 (range 0 to >20). The average difference in number of lesions between readers though was low, maximum 1.0 in all regions, meaning that no reader assessed the number of lesions systematically higher or lower than another. Conclusion: Variations in BS response assessment strongly depended on the method of analysis. Separate counting of lesions on repeated BSs without access to prior scans cannot be recommended. Furthermore, variation in classification of progression versus nonprogression might have significant impact on clinical decision-making, emphasizing the need for a uniform approach in bone response monitoring.