Variability in film reader estimates of breast density in the PERFORMS scheme

Abstract

Introduction All UK screening personnel are invited to take part annually in the PERFORMS self-assessment scheme where they make several judgements about series of challenging recent screening cases. As part of this process they assess the density of each case. Density is a factor known to be associated with a greater risk of developing breast cancer and thus accurate density judgements may well presage the facility to proffer improved follow-up for individual women. The present study examines the degree of variability amongst film reader estimates of breast density on a large number of cases. Methods Data were examined from the most recent 2 years of the PERFORMS scheme where breast density estimates were made for each case examined using a three-point rating scale of fatty, mixed density, and dense. These data comprised information from 444 individuals (mainly consultant radiologists, advanced practitioners and breast physicians) who had all examined the same 240 difficult cases. Results The inter-rater reliability, corrected for chance agreements, was assessed using kappa. Overall, the degree of agreement across cases on breast density category was significantly greater than no agreement (P <0.0001). However, only a moderate degree of inter-rater reliability was exhibited, kappa = 0.470. There were significant differences between the levels of agreement amongst the ratings of the radiologists, advanced practitioners and others (all P <0.05). Conclusions The low agreement rates between participants for density ratings were surprising. That there were differences between the occupational groupings may reflect breast screening experience.