Comparison of single reading with double reading of mammograms, and change in effectiveness with experience

In the prevalence round of screening for breast cancer at a single centre, the benefit of the second radiological report has been assessed using data from 33,734 women, their screening results and knowledge of the interval cancers. The service was set up under the UK National Breast Screening Programme, and the data show evidence of learning by both individuals and the team as a whole. Of the 269 cancers detected by screening, 33 would not have been diagnosed if the only report available had been the first. The recall recommendation rate associated with the first report was 6.9% and 236 cancers were detected. The recall recommendation rate associated with all queries of all the reports was 10% (3354 queries). Had these 3354 queries all been investigated (instead of the 1423 actually recalled) only a further three cancers would have been detected. Sensitivity of the programme as a whole was substantially better than that of individual radiologists, while the specificity was maintained. The decision pathway by which recalls were agreed between the two radiologists resulted in a low recall rate (4.2%) for the programme as a whole, and is a critical factor in gaining the benefit of the improved sensitivity without a concomitant deterioration in the specificity. With the passage through the prevalent round, recall rates steadily fell, the malignant to benign biopsy ratio improved and sensitivity increased. The second radiological report yielded 14% additional cancers diagnosed and contributed very significantly to good sensitivity and so to the effectiveness of screening. Economic analysis of the results will be reported in a further paper.