Background
Breast density may vary based on ethnicity, resulting in different density category distributions globally. We questioned whether radiologists skew mammographic density categorization according to BI-RADS®-based expectations versus local population/cultural influences. We evaluated variability in mammographic density categorization among radiologists practicing in Indonesia, the Netherlands, South Africa, and the USA.
Methods
200 screening mammograms performed from September-December 2017 were reviewed from four global locations for a total of 800. Three radiologists in each location/“team” provided consensus density assessments using BI-RADS®. Inter-reader agreement was compared using Gwet’s AC2 across all four density categories and Gwet’s AC1 for not-dense versus dense combined categories. Variability of distribution was calculated using the Stuart-Maxwell test across all four density categories, McNemar’s test for not-dense vs dense, and logistic regression for readers from a particular country on their own 200 mammograms versus other teams.
Results
For all 800 mammograms, inter-reader agreement across four density categories is 0.86 and not-dense vs dense categories is 0.66. Density category assessment distribution was significantly different when comparing teams in pairs and source readers vs other teams combined (p<0.001).
Conclusion
When comparing readers from four geographical locations, despite almost-perfect/substantial agreement of density categorization, distribution across density categories is significantly different.