CHICAGO - Dense breasts and small lesion size might be regarded as major reasons why signs of breast cancer are missed on a mammogram. Now, a comprehensive review of all breast cancers detected at a dozen institutions suggests that reality is far more complex.
The study, presented by Dr. Robyn L. Birdwell and Dr. Debra M. Ikeda of Stanford University at a Tuesday morning RSNA scientific session, retrospectively reviewed 1,083 breast cancers detected by screening mammography from 1994 to 1996. Of these, 493 patients had prior mammograms available for review.
An expert examined the pairs of prior and current mammograms and determined that 286 (58%) cancers were visible in retrospect.
Five blinded radiologists then judged only the prior exams, and a majority of these readers recommended recall imaging in 111 cases (39%). For these 111 cases, two expert radiologists compared the pairs of prior and screen-detected mammograms to determine why these cancers were missed.
Thirty percent of lesions were characterized by calcifications alone, while 70% were masses. For calcified lesions, the most common reasons for missed cancers were dense breasts (34%) and overlooked calcifications (31%). For masses, the most common problems were distracting lesions (43%) and lesions on the edge of the film or tissue (42%). Film technique played a potential role in 45% of cases.
"The mammographic characteristics of cancers that were missed at screening mammography were surprising in one sense because they were not atypical," Dr. Birdwell said. "They were located predominantly in the upper-outer quadrant, almost half of them were in less-dense breasts, the mean of the mass and calcification size was greater than 1 cm, the masses typically were irregular and spiculated, and the calcifications were clustered and pleomorphic."
Several strategies could be used to reduce detection errors, Dr. Birdwell said, including continuing medical education, reviewing errors, double reading, improving the screening environment by minimizing distractions, and using computer-aided detection.
A parallel analysis was conducted on the mammograms for which recall imaging was recommended by two or fewer of the five blinded radiologists. Of the 169 cases included in that analysis, experts would have recalled only 20%.
"The other 80% of cancers displayed perceptible but nonspecific findings that don't warrant recall as judged by expert radiologists," Dr. Ikeda said. "In this series, failure to identify these features on prior studies is not considered a diagnostic error."
By Deborah BarnesAuntMinnie.com contributing writer
December 1, 1999
