Determination of breast density fluctuates considerably from one exam to the next in any given woman. Such variability can both perplex the patient and drive unintended consequences into her care pathway—effects that may be exacerbated by the direct-to-consumer density notifications required by law in (so far) 24 U.S. states.
That’s from a systematic review of the literature published online Jan. 12 in the Annals of Internal Medicine.
The review further showed that supplemental screening of women with dense breasts does indeed find additional breast cancer, but it also increases false positives.
Joy Melnikow, MD, MPH, of UC-Davis lead-authored the report on the research, which began with quality assessment and abstraction of 24 studies from seven countries.
The team deemed six studies to be of good quality, primarily using U.S. Preventive Services Task Force criteria supplemented by NIH methodology checklists and other guides.
According to USPSTF criteria, a good-quality study generally meets all pre-specified criteria, while fair-quality studies do not meet all criteria but have no important limitations. Poor-quality studies have important limitations that may invalidate results (e.g., inadequate or biased application of reference standard; population limited to very high-risk patients).
In three good-quality studies reporting reproducibility of BI-RADS density, Melnikow and colleagues found that 13 percent to 19 percent of women were recategorized between “dense” and “nondense” at subsequent screening.
Breaking down performance by imaging modality, the team found:
- Two good-quality studies reported that sensitivity of ultrasound for women with negative mammography results ranged from 80 percent to 83 percent. For specificity the range was 86 percent to 94 percent and, for positive predictive value (PPV), 3 percent to 8 percent.
- The sensitivity of MRI ranged from 75 percent to 100 percent; specificity, from 78 percent to 94 percent; and PPV, from 3 percent to 33 percent (three studies).
- Rates of cancer detection with digital breast tomosynthesis increased by 1.4 to 2.5 per 1,000 examinations compared with mammography alone (three studies).
Additionally, rates of additional cancer detection with ultrasound were 4.4 per 1,000 exams (89 percent to 93 percent invasive), while recall rates were 14 percent.
Use of MRI detected 3.5 to 28.6 additional cancer cases per 1,000 exams (34 percent to 86 percent invasive), and recall rates were 12 percent to 24 percent.
Overall, recall rates ranged from 7 percent to 11 percent, compared with 7 percent to 17 percent with mammography alone.
None of the studies under review looked at breast cancer outcomes.
Laying out the limitations of their work, Melnikow et al. note that good-quality evidence was sparse, studies were small and confidence intervals wide, and definitions of recall were absent or inconsistent.
“Moving from a ‘dense’ to a ‘nondense’ breast categorization may result in different mandated communications in states with breast density notification, as well as fluctuation in clinical recommendations for supplemental screening,” the authors conclude in their study discussion.
They also stress that tomosynthesis may correlate with lower recall rates, but studies are few and retrospective.
The team calls for “well-designed, long-term, prospective, comparative studies of supplemental screening” focused on clinical outcomes.
To this Melnikow added, in prepared remarks:
“It is important to be clear who actually has dense breasts. Also, when patients are told their breasts are either dense or not dense, they need to have confidence in that assessment."
The journal has posted the full study online for free.