Purpose: To determine how many cancers screen-detected at subsequent rounds were already visible on previous screening mammograms, and to study their radiological features.
Materials and methods: The previous screening mammograms of 100 women who had cancers screen-detected at subsequent rounds (group A), and the negative screening mammograms of 200 women (group B) who had a further negative screening test two years later, were mixed for a new reading. The two groups were similar for age and year of examination. These films were blindly reviewed by 5 radiologists. Then, mammograms were reviewed retrospectively, with knowledge of subsequent diagnostic results. Finally the A group findings were classified as: 1) true negative: no radiological signs; 2) minimal sign: a nonspecific abnormality is retrospectively visible at the site of subsequent cancer; 3) false negative: "she should have been recalled"; 4) misdiagnosis at assessment: the woman had been recalled, but the cancer was missed after the assessment procedures.
Results: 60% of cases were true negatives, 29% were minimal signs, 9% were false negative and 2% were misdiagnosed at assessment. The most common radiological sign found among false negative cases was an architectural distortion: opacities and calcifications were more frequent among minimal signs. Only 10 of 40 cancers retrospectively visible on previous mammograms had reached stage II at diagnosis. At blinded review, the radiologists found false abnormalities in a considerable number of healthy women (average: 29%).
Discussion and conclusions: Our study shows that mammography sensitivity can be improved. Cancer radiological signs may go undetected due to difficult interpretation (opacities, calcifications) or perception (architectural distortions). The use of a low threshold of suspicion (as in a reading test) in real screening might permit to detect more cancers (most of them, however, would not reach advanced stages at subsequent rounds), but might also lead to many unnecessary assessments and, probably, to some benign biopsies in healthy women. In conclusion, an attempt at improving mammography sensitivity by lowering the threshold of suspicion can not be directly recommended due to the considerable negative effects related to a loss in specificity. A reading test similar to the one presented in our study would be a useful training procedure for radiologists who are involved in a screening program.