Editorial: Time for Change in Digital Breast Tomosynthesis Research
Key findings
- When digital breast tomosynthesis (DBT) was first being evaluated, it was reasonable to compare sequential cohorts of women, but that approach can introduce bias
- In a recent editorial, Daniel B. Kopans, MD, explains why it's too soon to claim synthetic two-dimensional mammography plus DBT can replace full-field digital mammography plus DBT
- Dr. Kopans describes a more accurate way to compare full-field digital mammography with synthetic images
Digital breast tomosynthesis (DBT) was developed for screening, not diagnosis. Its major benefits were identified only once it was used as a breast cancer detection tool in very large studies.
Subscribe to the latest updates from Radiology Advances in Motion
In an editorial published in Radiology, Daniel B. Kopans, MD, professor of Radiology at Harvard Medical School, founder of the Breast Imaging Division in the Department of Radiology at Massachusetts General Hospital, a pioneer in the field of breast imaging and the inventor of the DBT technique, traces the history of this research and shows why some of the more recent conclusions in the literature lack sufficient evidence.
In particular, he suggests a simple, direct method to research whether replacing full-field digital mammography (FFDM) with synthetic two-dimensional images is warranted.
Early Research
In the early evaluation of a new test, it's reasonable to compare its performance in sequential cohorts. Once several large practices began to use DBT for screening, they triaged some women with DBT and compared their results with those of women who had another test.
Eventually, data sets were large enough to objectively compare women screened using FFDM in an earlier period to women screened with DBT in more recent times. These larger data sets confirmed what initial work at Mass General and elsewhere had suggested: DBT was superior to FFDM for cancer detection and was associated with lower recall rates.
However, different cohorts may have different prior probabilities of breast cancer. Additionally, the triage approach introduces the possibility of selection bias.
Recent Research
Based on similar retrospective studies, some researchers now claim synthetic two-dimensional mammography plus DBT can replace FFDM plus DBT. Dr. Kopans responds that the lack of statistically significant differences between the two techniques could be attributable to an insufficiently large number of cases.
Furthermore, in screening studies, missing "only a handful" of cases could have major implications for mortality.
Every DBT study should contain full FFDM mediolateral oblique and craniocaudal projections, Dr. Kopans recommends. For women 40 and older, even the extrapolated radiation risk is extremely low. He urges his colleagues not to replace FFDM images with synthetic images unless it can be shown directly that the two technologies are comparable or one is superior.
Future Research
Dr. Kopans advises that rather than compare breast cancer detection rates in different cohorts, researchers of DBT should compare endpoints in the same women:
- For each patient, collect the FFDM image, then have the system process the synthetic image.
- Interpret the images in a random fashion. For one patient, "lock in" the results from the FFDM image before interpreting the synthetic image. For the next patient, reverse the order of interpretation. Alternatively, randomize the process of determining which image will be reviewed first.
Every screening examination is a ready-made opportunity for comparison, Dr. Kopans notes. Prospective evaluation of the same woman at the same point in time should replace sequential studies whenever possible.
view original journal article Subscription may be required
Learn more about the Division of Breast Imaging
Learn more about research in the Department of Radiology