The role of the reference radiologist. Estimates of inter-observer agreement and potential delay in cancer detection in the national breast screening study

Abstract

The Canadian National Breast Screening Study (NBSS) is a randomized controlled trial to assess the effect of screening on breast cancer mortality. The NBSS designated a single reference radiologist who blindly reviewed over the course of the study 5200 randomly selected two-view mammographic examinations of women not known to have breast cancer. He also reviewed 575 screening-detected breast cancer (SBC) cases and 102 interval breast cancer (IBC) cases. All cancers were histologically proven. As a result of the reviews, comments on inter-observer agreement, interpretation, and technical quality were conveyed on an ongoing basis to radiologists appointed to 15 NBSS screening centers. Agreement of the reference radiologist with center radiologists was better for breast cancer cases (kappa = 0.511, P less than .002) than for those not known to have breast cancer (kappa = 0.307, P less than .002). Observer error and technical problems led to delayed detection in 22% of SBCs and 35% of IBCs. Another 11% of SBCs and 58% of IBCs were probably mammographically occult. No similarly comprehensive review of mammography during a screening program has been published. Suggestions arising from the NBSS review were sometimes resisted by center radiologists. Measures are suggested which might facilitate acceptance of recommendations arising from audit mechanisms in mammography screening programs, thereby enhancing opportunities for mammographic excellence.