Accuracy and inter-reader agreement of breast MRI for cancer staging using 0.08 mmol/kg of gadobutrol

Abstract

BackgroundEvidence on gadolinium brain accumulation after contrast-enhanced MRI prompted research in dose reduction. PurposeTo estimate accuracy and inter-reader reproducibility of tumor size measurement in breast MRI using 0.08 mmol/kg of gadobutrol. MethodsWe retrospectively analyzed all women who underwent 1.5-T breast MRI for cancer staging at our department with 0.08 mmol/kg of gadobutrol. Two readers (R1 and R2, 12 and 3 years-experience) measured the largest lesion diameter. Accuracy was estimated both as correlation with pathology and rate of absolute (>5 mm) overestimation and underestimation, inter-reader reproducibility using the Bland–Altman method. Data are given as median and interquartile range. ResultsThirty-six patients were analyzed (median age 56 years, 49–66) for a total of 38 lesions, 24 (63%) mass enhancement, 14 (37%) non-mass enhancement. Histopathological median size (mm) of all lesions was 15 (9–25): 13 (9–19) for mass lesions, 19 (11–39) for non-mass lesions. On MRI, R1 measured (mm) 14 (10−22) for all lesions, 13 (10–19) for mass lesions, 19 (11–49) for non-mass lesions. MRI-pathology correlation was very high for all lesion categories (ρ ≥ 0.766). On MRI, R1 overestimated lesion size in 6 cases (16%), and underestimated in 3 (8%); R2, overestimated 7 cases (18%) and underestimated 3 cases (8%). At inter-reader reproducibility analysis (mm): bias 0.9, coefficient of reproducibility 13 for all lesions; −0.1 and 6 for mass lesions; 2.5 and 20 for non-mass lesions. ConclusionsBreast MRI may be performed using 0.08 mmol/kg of gadobutrol with high accuracy and acceptable inter-reader agreement.