Blinded double reading yields a higher programme sensitivity than non-blinded double reading at digital screening mammography: a prospected population based study in the south of The Netherlands

Abstract

PURPOSE: To prospectively determine the screening mammography outcome at blinded and non-blinded double reading in a biennial population based screening programme in the south of the Netherlands. METHODS: We included a consecutive series of 87,487 digital screening mammograms, obtained between July 2009 and July 2011. Screening mammograms were double read in either a blinded (2nd reader was not informed about the 1st reader's decision) or non-blinded fashion (2nd reader was informed about the 1st reader's decision). This reading strategy was alternated on a monthly basis. Women with discrepant readings between the two radiologists were always referred for further analysis. During 2 years follow-up, we collected the radiology reports, surgical correspondence and pathology reports of all referred women and interval breast cancers. RESULTS: Respectively 44,491 and 42,996 screens had been read either in a blinded or non-blinded fashion. Referral rate (3.3% versus 2.8%, p<0.001) and false positive rate (2.6% versus 2.2%, p=0.002) were significantly higher at blinded double reading whereas the cancer detection rate per 1000 screens (7.4 versus 6.5, p=0.14) and positive predictive value of referral (22% versus 23%, p=0.51) were comparable. Blinded double reading resulted in a significantly higher programme sensitivity (83% versus 76%, p=0.01). Per 1000 screened women, blinded double reading would yield 0.9 more screen detected cancers and 0.6 less interval cancers than non-blinded double reading, at the expense of 4.4 more recalls. CONCLUSION: We advocate the use of blinded double reading in order to achieve a better programme sensitivity, at the expense of an increased referral rate and false positive referral rate.