Towards risk-based breast cancer screening

Evidence shows that risk-based screening improves the balance of benefits and harms of screening for breast cancer, compared with simple age-based screening.^1–3 How to best implement risk-based breast cancer screening is a topic of active research. This research has incorporated risk factors such as family history of breast cancer, patient history of breast biopsy, and breast density.^1–3 High breast density poses a particular challenge as it is common and both increases breast cancer risk and decreases the sensitivity of mammography for the detection of breast cancer. Further, breast imaging is necessary to measure breast density, so identifying the optimal time to assess density in order to guide a screening program is critical if breast density is to be used to determine optimal screening strategies.

In this issue, Shih and colleagues propose a single measure at age 40 of breast density using the Breast Imaging Reporting and Data System (BI-RADS) with subsequent screening intervals and duration based on this measure (4 ADD SHIH CITATION HERE). Radiologists assign breast density into one of four BI-RADS categories: almost entirely fatty, scattered areas of fibroglandular density, heterogeneously dense or extremely dense. In Shih and colleagues’ proposal, the first two categories are considered “non-dense” and the last two “dense.” Clinical BI-RADS density assessment is subjective with inter-rater and intra-rater reliability resulting in moderate to substantial agreement (5). Shih and colleagues’ model identified the most cost-effective approach was baseline mammography at age 40 years for every woman followed by annual mammography from ages 40-75 for women with dense breasts and biennial mammography from ages 50-75 years for women with non-dense breasts. This strategy resulted in the most deaths averted, but also more screening examinations, false-positive examinations, and overdiagnosis.

Several aspects of Shih’s results are surprising and inconsistent with prior studies.^1–3,5 For example, modelling studies of annual screening strategies have often found them to be dominated or not cost-effective. Generally, screening every year results in a higher ratio of harms to benefits because few added deaths are averted - compared with biennial screening - while the added cost of annual screening is large. The screening strategy proposed by Shih and colleagues would perform annual screening on 60% of women, which raises questions about how it could be cost-effective.

In addition, Shih and colleagues make a number of generous, and possibly overly optimistic, assumptions in their models. They report the number of deaths averted per 1,000 women aged 50-75 years screened biennially is 10.2, which is higher than the 7 per 1000 women screened that others have reported.³ Their model therefore assumes a higher benefit of screening even prior to implementing a risk-based screening strategy, and further magnifies the number of deaths averted by using the higher sensitivity of mammography for prevalent screens for women with dense breasts for all subsequent screening rounds. By modelling an increase in the risk of preclinical breast cancer incidence in women with dense breasts combined with a high relative detection rate in these women, Shih and colleagues’ analyses yield a high rate of screen-detected cancers in women with dense breasts and a high number of deaths averted (13.2 deaths per 1,000 women). By contrast, in a study by Trentham-Dietz and colleagues, only annual screening in women aged 50-75 years with dense breasts and two first-degree relatives with breast cancer achieved a comparably high number of deaths averted (14.3-14.7 per 1,000 women screened) (6). Assuming a high number of averted deaths in Shih and colleagues’ proposed strategy results in the strategy being cost-effective despite the high number of screening examinations, false-positive examinations and overdiagnosis.

The fundamental problem with measuring breast density and assigning screening frequency based on one reading is that not everyone with dense breasts is at increased risk of breast cancer. Only 24% of women with dense breasts are at high risk of a missed invasive cancer within one year of negative mammogram (interval cancer >1/1000 mammograms), and these women can be identified using predictions from the Breast Cancer Surveillance Consortium (BCSC) risk model combined with breast density categories from BI-RADS (7, 8). This observation means most women with dense breasts can undergo biennial screening and need not consider annual screening or supplemental imaging. Screening women with dense breasts who are not at increased breast cancer risk subjects these women to more harms from false-positive tests, without additional benefit. Thus, we caution against using breast density alone to determine if a woman is at elevated breast cancer risk.

An attractive alternative is to focus on overall risk to select screening strategies, rather than just on breast density, because this approach better balances the benefits and harms of screening. For example, given that most guideline groups recommend screening from age 50-74, identifying women in their forties who have the risk of a woman aged 50-59 years is one way to determine who might benefit from earlier initiation of screening. For a forty-year-old woman to reach the level of risk of the average woman in her fifties, she would need to have dense breasts and a family history of breast cancer or a history of breast biopsy, or she would need to have scattered fibroglandular breast density, a family history of breast cancer and a history of breast biopsy (1,8). Thus, women who have a first-degree relative with breast cancer or a history of breast biopsy could be offered screening in their forties and, if mammography shows dense breasts, they could continue biennial screening through their forties. Such women with non-dense breasts could resume biennial screening at age 50.

Breast density is an important risk factor to include in risk-based screening strategies because it is both a strong and prevalent risk factor accounting for a large proportion of breast cancers (9) However, we argue that breast density should be combined with age and other risk factors when developing risk-based screening strategies that optimize benefits and minimize harms. We believe that until a more robust risk-based strategy is identified, the frontier curves presented in Shih and colleagues’ analyses support screening biennially from ages 50-74 years.