Effect of recall rate on earlier screen detection of breast cancers based on the dutch performance indicators

Background. The recall rate (i.e., the rate at which mammographically screened women are recalled for additional assessment) in the Dutch breast screening program (0.89% in 2000 for subsequent examinations) is the lowest worldwide, with possible consequences including higher rates of late-detected (i.e., "missed") interval and screen-detected cancers. To estimate the effect of changes in recall rate on earlier detection of cancers, we carried out a blinded review of interval and screen-detected cancers in the Dutch screening program. Methods: A total of 495 sets of screen-negative mammograms (prediagnostic mammogram and the immediate previous mammogram) were collected from women participating in the biennial Dutch screening program. Of these, 250 were from control subjects, and 245 were from women who were subsequently diagnosed with breast cancer (123 interval and 122 screen-detected cancers). These mammograms were read by 15 radiologists who specialize in screening mammography and were blinded to outcome. Mean detection sensitivities for different false-positive rates were calculated using a linear mixed model. These results were used to calculate the effect of recall rate adjustment on earlier detection of cancers and numbers of false-positives. Results: Increasing the recall rate to 2.0% would increase the detection rate from 4.20 parts per thousand to 4.52 parts per thousand due to the earlier detection of interval cancers. Moreover, 0.54 parts per thousand of the screen-detected cancers would be detected 2 years earlier (late screen-detected cancers). At recall rates of 3.0% and 4.0% the detection rate would increase to 4.58 parts per thousand and 4.63 parts per thousand, respectively, and 0.64 parts per thousand and 0.72 parts per thousand, respectively, of the screen-detected cancers would be detected 2 years earlier. For each 1.0% incremental increase in recall rate above 5.0%, the detection rate would increase by approximately 0.03 parts per thousand, with positive predictive values decreasing to below 10%. Conclusion: Breast cancer can be detected earlier by lowering the threshold for recall, especially for recall rates of 1%-4%. With further recall rate increases, cancer detection levels off with a disproportionate increase of false-positive rates.