Discrepant screening mammography assessments at blinded and non-blinded double reading: impact of arbitration by a third reader on screening outcome

To determine the value of adding a third reader for arbitration of discrepant screening mammography assessments. We included a consecutive series of 84,927 digital screening mammograms, double read in a blinded or non-blinded fashion. Arbitration was retrospectively performed by a third screening radiologist. Two years' follow-up was performed. Discrepant readings comprised 57.2% (830/1452) and 29.1% (346/1188) of recalls at blinded and non-blinded double readings, respectively. At blinded double reading, arbitration would have decreased recall rate (3.4 to 2.2%, p  < 0.001) and programme sensitivity (83.2 to 76.0%, p = 0.013), would not have influenced the cancer detection rate (CDR; 7.5 to 6.8 per 1,000 screens, p = 0.258) and would have increased the positive predictive value of recall (PPV; 22.3 to 31.2%, p  < 0.001). At non-blinded double reading, arbitration would have decreased recall rate (2.8 to 2.3%, p  < 0.001) and increased PPV (23.2 to 27.5%, p = 0.021), but would not have affected CDR (6.6 to 6.3 per 1,000 screens, p = 0.604) and programme sensitivity (76.0 to 72.7%, p = 0.308). Arbitration of discrepant screening mammography assessments is a good tool to improve recall rate and PPV, but is not desirable as it reduces the programme sensitivity at blinded double reading. • Blinded double reading results in higher programme sensitivity than non-blinded reading. • Discrepant readings occur more often at blinded compared to non-blinded reading. • Arbitration of discrepant readings reduces the recall rate and PPV. • Arbitration would reduce the programme sensitivity at blinded double reading