Nearly a quarter of interval breast cancers could have been caught at the initial mammogram, according to a new analysis published Monday in Academic Radiology.
Such cancer cases typically crop up between two scheduled rounds of screening, for various reasons that could include clinician oversight or aggressive growth. Wanting to better understand the factors leading to these instances, overseas scientists recently dove into the data searching for clues.
Altogether, researchers pinpointed 1,010 interval cancer cases recorded as part of the BreastScreen Norway program between 2004 and 2016. About 24% (or 246) were deemed missed at the initial screening, with an average time between exams of about 14 months. Possible remedies to address this issue could include shortening intervals or implementing supplementary screening techniques, experts advised.
“Education, including self-assessment and training schemes, and participation in reviews could be ways to improve the screen-reader’s sensitivity to more subtle findings,” first author Tone Hovda, MD, with the Department of Radiology at Vestre Viken Hospital Trust in Drammen, Norway, and colleagues wrote April 26. “These strategies may also improve the radiologists’ perception and interpretation and increase their awareness of possible pitfalls.”
Hovda et al. classified each interval instance as true (no findings on prior screenings), occult (no findings at screening or diagnosis), minimal signs (minor, nonspecific findings), or missed (obvious findings). They additionally analyzed mammographic findings, along with density, time since prior screening and other differing characteristics between these groups.
About 48% of cancers were deemed as true or occult, the authors reported, while 28% showed minimal signs. Patients averaged age 61 at their cancer diagnosis, and there were no observed differences in breast density between the groups. However, a higher percentage of density was found in women with occult cancers. Among the 246 missed cancers, one-third were masses and one-third were asymmetries at prior screening. No difference in histopathological characteristics were observed between true, minimal signs and missed cancer.
Hovda additionally suggested double reading, optimizing image quality, and improving positioning as other ways to reduce these misses.
“During the past years, artificial intelligence based on deep convolutional neural networks shows promising results in breast diagnostics and screening,” the team noted. “And if AI in the future demonstrates the ability to detect abnormalities in images not perceived by radiologists, or not even detectable by the human eye, it may be possible to lower the interval cancer rate.”
You can read the rest of the study in Academic Radiology here.