Background and Aims: Mammographic density (MD) is a risk factor for breast cancer which can also visually mask breast cancers, making them harder to detect on mammography. For mammographic breast screening programs these effects jointly contribute to increased interval cancer rates and decreased screening program sensitivity and specificity for women with denser breasts. A more risk-based approach to population breast screening could potentially use MD to help direct supplemental or alternative screening technologies to these women, thereby reducing the risk of interval cancers. The benefits of such interventions would need to be balanced against potential increases in false positive outcomes.To help understand the evidence for consideration of MD in risk-based breast screening, we conducted a systematic review of studies to assess how accurately a given MD measurement tool can stratify women according to their risk of a subsequent interval breast cancer or other pre-determined screening program indicator.
Methods: To be eligible, studies needed to report interval cancer rates or program sensitivity (separately for invasive breast cancer or DCIS), or false positives, program specificity, or missed cancers, for each MD category determined by an MD measurement tool, in populations screened with digital mammography alone (excluding studies confined to high-risk groups). Medline, Embase, Cochrane Database of Systematic Reviews databases were searched1/1/2008-19/10/2021). Two reviewers assessed articles for inclusion, extracted data and assessed the risk of bias using QUADAS-2. Outcomes with 95% confidence intervals were plotted by reported MD categories, and to standardise comparisons between studies, by MD category midpoint percentile.
Results: Of 1675 unique records, 20 studies were eligible for inclusion, reporting on three MD assessment tools (15 on BI-RADS, 5 on Volpara, and one on mammographic texture resemblance). All three tools were more likely to discriminate between the lower MD groups than between the higher MD groups in terms of observed rates of invasive interval cancer, false positives and program specificity, however, tool performance varied between settings. No study reported simultaneously clear discrimination of both high and low risk groups for interval cancers. All studies were determined to be at high risk of bias due to outcome assessment.
Conclusions: There was sufficient evidence supporting the potential use of BI-RADS and Volpara MD assessment for directing targeted screening technologies aiming to reduce interval cancers. However, their performance varies between studies including how well either higher risk or lower risk groups are identified. This suggests that validation in the target setting would be required prior to implementation.