Optimising breast cancer screening reading: blinding the second reader to the first reader's decisions

Area of Science:

• Radiology
• Oncology
• Public Health

Background:

• Breast cancer screening guidelines in Europe recommend analysis by two readers.
• The impact of blinding readers to each other's decisions in mammography is not well-established.

Purpose of the Study:

• To investigate whether blinding the second reader in mammography affects reader behavior and screening outcomes.
• To assess the influence of blinding on recall rates, cancer detection, and positive predictive value (PPV).

Main Methods:

• A cohort study analyzing data from 1,119,191 women across 43 UK screening centers.
• Multilevel models and Pearson's chi-squared test were used to evaluate the effect of blinding on reader concordance, recall rates, cancer detection, and PPV.

Main Results:

• When the second reader was not blinded, they were more likely to agree with the first reader's recall decision, indicating potential alliterative bias.
• Blinding the second reader was associated with a lower overall recall rate (OR 0.923) and a higher positive predictive value (22.1% vs. 20.6%).
• No significant difference in cancer detection rates was observed between blinded and unblinded conditions.

Conclusions:

• Reader 2's decisions can be influenced by Reader 1's when not blinded, potentially leading to alliterative bias and bypassing arbitration.
• Blinding the second reader in mammography is associated with a slightly higher PPV, though potential confounding by center characteristics exists.
• Ensuring independent review by both readers is crucial for optimizing the accuracy and effectiveness of double-reading in breast cancer screening.