Purpose: To retrospectively investigate whether different review designs have an influence on the estimate of missed interval cancer in a population-based breast cancer screening program.
Materials and methods: The Norwegian Breast Cancer Screening Program invites women aged 50-69 years to undergo biennial screening mammography. The current study was part of the evaluation and scientific aspects of the screening program and thus was covered by the general ethical approval of the screening program as a part of the Cancer Registry of Norway. All participants signed an informed consent that specified that data related to their screening visit could be used for evaluation and scientific purposes. Six radiologists (9-34 years of experience in mammography) reviewed previously obtained bilateral two-view screening and diagnostic mammograms of 231 interval cancers, 117 screening-detected cancers, and 373 normal cases. Four review designs were used: individual and paired blinded review and individual and consensus informed review. A five-point interpretation scale was used to reclassify the cancers into missed cancers, minimal signs, and true cancers. The number and proportion of subgroups were estimated with 95% confidence intervals.
Results: Of 231 interval cancers, 46 (19.9%) were reclassified as missed cancers with the mixed blinded individual review and 54 (23.4%) were classified as missed cancers with the mixed blinded paired review. Eighty-three cancers (35.9%) were classified as missed cancers with individual informed review, and 78 (33.8%) were classified as missed cancers with consensus informed review. Thirty-nine cancers (16.8%) were reclassified as missed when four or more radiologists assigned a score of 2 or more (probably benign or more suspicious); three cancers (1.3%) were reclassified as missed when a score of 4 or more (probably malignant or more suspicious) was assigned.
Conclusion: The percentage of interval cancers classified as missed ranged from 1.3% to 35.9% according to review design. To encourage learning, a review protocol should include both blinded and informed designs.