Jacky Luiten

Recall and outcome of screen‐detected microcalcifications of mammography screening | 69 5 Follow‐up of recalled women During a two year follow‐up period (until the next biennial screening examination), we collected the radiology reports, type of biopsy methods with their outcome, and breast surgery reports of all recalled women. The screening organization routinely received the follow‐up data from the hospitals at which the women were analyzed after recall. To complete two year follow‐up, one of the radiologists (LD, with >25 years of experience in breast imaging) and several radiology residents collected additional reports, which were not received by the screening organization, through visits at these departments. All data were then entered into a database, which was created for quality control of the screening program and scientific purposes by the radiologist. The quality of data entry was not reviewed. If a woman was recalled for more than one ipsilateral lesion or for one lesion in the right breast and one in the left breast (bilateral) during the same screening round, the mammographic lesion with the highest suspicion at mammography screening was considered as the index lesion for recall. For the purpose of this study, we scored one screening abnormality per recalled woman. If a woman was recalled again in a subsequent screening round, this counted as a new recall. A total of 60 women were recalled twice, and two were recalled three times. Only four women, one of whom had microcalcifications, experienced a repeated recall within the same two year period. Screen‐detected cancers were divided into DCIS and invasive cancers. Lobular carcinoma in situ is considered a benign lesion. Details on the methods for the detection of interval cancers in our screened cohort have been published previously. 17,18 Statistical analysis The main outcome measure of this study was the number of microcalcification recalls per 1000 screening examinations and positive predictive value (PPV) of microcalcification recalls during two decades of mammography screening. These trends are shown as graphs and reported as absolute numbers, proportions, PPVs with 95% confidence intervals (CI), and rates per 1000 screening examinations for women screened from 1997 until 2017. Evaluation of the CIs over time, treating nonoverlapping CIs as evidence of difference, then provides conservative assessment of trends 19 that fit the graphical presentation and does not go beyond the limitations in the data. To allow clear interpretation of the data, results are presented separately for first (initial) screening examinations of participating women and subsequent screening examinations. The limitations of the data did not