These findings provide evidence for substantial socioeconomic disparities in the recent decline in breast cancer incidence. The greatest reductions in breast cancer incidence between 1998–2001 and 2003–2006 were observed among women living in counties with higher levels of income and education. Counties which had the highest DCIS rates during the years preceding the national peak in invasive breast cancer incidence tended to have the largest decline in incidence, suggesting an important role for screening mammography utilization.
These disparities were driven mainly by reductions in the high incidence of invasive breast cancer among women in counties with high socioeconomic status. While this is the first study to report disparities in the recent decline in incidence according to income and education, the positive association between socioeconomic status and breast cancer incidence is well known [21
]. This association is largely due to increased utilization of screening mammography [23
] and variation in known risk factors (e.g. reproductive patterns) [24
Our observation of an association between socioeconomic status and DCIS incidence is consistent with the known association between socioeconomic status and screening mammography. The National Health Interview Survey [18
] found that in the year 2000 approximately 55% of women aged 40 or older with income below the poverty level had a mammogram within the past two years, compared to 74% of women with income more than twice the poverty level. A similar difference is observed by education; approximately 58% of women without a high school diploma were screened within the past two years, compared to 77% of women with at least some college education. Notably, screening utilization has varied according to race as well. Uptake of screening mammography utilization occurred earlier in whites compared to blacks, though current levels now appear to be quite similar [18
The associations between socioeconomic status, DCIS rates, and the decline in invasive breast cancer incidence are consistent with the hypothesis that a saturation of screening mammography utilization contributed to a decline in incidence. With screening utilization reaching a plateau, years of intense screening may have largely depleted the pool of prevalent invasive breast cancers. Additionally, invasive cases which would have otherwise been diagnosed after the year 2000 were instead diagnosed at an earlier date as in situ lesions. With a deficit of advance-stage cases due to early detection [10
] and a deficit of new early stage cases as fewer women are being screened for the first time, a decline in incidence could be expected.
Notably, socioeconomic status is also associated with use of postmenopausal hormones [9
]. In the nationally representative National Health and Nutrition Examination Survey conducted between 1988 and 1994 [25
], 46.8% of women older than age 60 with 16 years of education reported ever using hormone replacement therapy, compared with 41.3% of women with 12 years of education. Similarly, 39.6% of women with $50,000 or more of income per year reported having used hormones, compared to only 24.2% among women with less than $10,000 of income per year. A more recent report [9
] from the Cancer Research Network described the changes in hormone therapy use after publication of the results of the WHI randomized trial of estrogen plus progestin hormone therapy [4
]. Though hormone use was more common among higher socioeconomic groups before and after the WHI results, similar percent reductions in use (approximately 40 – 45%) were observed across socioeconomic groups following the publication of the WHI results in July 2002. While this study was perhaps the largest of its kind (over 200,000 women), it was restricted to women in five health maintenance organizations and thus may not be generalizable to the general population. Thus, the potential ability of trends in hormone use to explain disparities in the decline in breast cancer incidence remains unclear. Notably, we observed similar associations between screening (as reflected by DCIS incidence) and trends in both ER-positive and ER-negative breast cancer. This finding is not likely to be attributed to patterns in hormone use, which is a risk factor for ER-positive breast cancer only [27
The decreased incidence of invasive breast cancers is likely multi-factorial. While cessation of hormone therapy and a plateau in screening utilization may play major roles, other factors may also contribute [28
]. For example, increased use of chemoprevention therapies may contribute to decreased invasive breast cancer incidence, which may also disproportionately affect groups according to socioeconomic status.
Since individual-level data on screening utilization is not available within the SEER database, county-level incidence of DCIS was used as a proxy. It should be recognized that DCIS incidence is also influenced by other factors aside from screening – most notably, variation in population risk factors [29
] – though mammography utilization is a very strong predictor of DCIS incidence and is perhaps the strongest risk factor [6
]. In addition, it is important to distinguish between community-and individual-level indicators of socioeconomic status [22
]. With county-level data only, we were unable to evaluate trends in breast cancer incidence among women according to individual-level income and education. Finally, the SEER database does not include information on postmenopausal hormone use. Thus we were unable to directly examine the contribution of trends in hormone use to disparities in the decline of breast cancer incidence.
In conclusion, our results indicate that the recent decline in invasive breast cancer incidence has been greatest among women living in counties with high socioeconomic status. The observed association between DCIS incidence and the decline in invasive breast cancer suggests an important role for utilization of screening mammography. Future studies may be able to improve our understanding of this relation by examining trends in breast cancer incidence according to individual-level indicators of socioeconomic status, postmenopausal hormone use, and mammography utilization. Sophisticated simulation models, such as those developed by the CISNET consortium [30
], may be particularly useful in evaluating the complex contributions of screening and postmenopausal hormone use to trends in breast cancer incidence. An improved understanding of the factors driving the recent decline in invasive breast cancer incidence may help to enhance this desirable trend among all socioeconomic groups, while also permitting better prediction of future trends in breast cancer incidence and mortality.