Breast Density Measurement May Help Predict Breast Cancer Risk

By Crystal Phend
SAN FRANCISCO, March 4 -- Incorporating breast density into assessment tools may add to the ability to predict breast cancer risk, but it's no slam-dunk, researchers here said.
Caution patients that the breast density risk prediction model is likely not ready for clinical use.
A simple risk algorithm incorporating breast density discriminated which women would develop breast cancer significantly better -- and "possibly clinically" better -- than the standard Gail model, reported Jeffrey A. Tice, M.D., of the University of California San Francisco, and colleagues in the March 4 issue of the Annals of Internal Medicine.
Like previous models, though, the breast density model had only modest ability to discriminate which women would develop breast cancer (concordance index 0.66 on a scale of 0.5 to 1.0).
Furthermore, the breast density model reclassified risk incorrectly more often than correctly compared with the Gail model although it still had a higher positive predictive value.
Since no single model can address all needs in breast cancer risk assessment, the researchers said, the best, most cost-effective approach might be to start with a simple model and family history then move to more detailed assessment for women at higher risk.
The breast density model "is convenient enough that it could be incorporated into routine breast cancer screening, and primary care physicians could use it to calculate an individual woman's breast cancer risk," they wrote.
"However, its accuracy must be further evaluated in independent populations before it can be recommended for clinical use," they added.
Radiographically dense breasts have consistently been implicated as a major risk factor for breast cancer partly because mammography is less sensitive in dense breasts.
To refine their breast density risk prediction model, the researchers analyzed data from mammography registries of the Breast Cancer Surveillance Consortium, which is a community-based sample broadly representative of the United States.
It included 1,095,484 women 35 or older who had had at least one mammogram with breast density measured using the Breast Imaging Reporting and Data System (BI-RADS) classification system.
The sample was ethnically diverse with 29% of the cohort of black, Asian, Hispanic, or other minority race or ethnicity.
Invasive breast cancer developed among 14,766 women over the median follow-up of 5.3 years.
Factors in the model for five-year risk of invasive breast cancer included age, race or ethnicity, and breast density. Family history and biopsy history were added to adjust incidence estimates when available.
The predicted incidence in the randomly-selected validation sample (60% of the cohort) was well matched to the observed incidence (1.41% versus 1.38%, expected-to-observed ratio 1.03, 95% confidence interval 0.99% to 1.06%).
The model's ability to accurately discriminate which women would develop breast cancer was only modest, though.
Concordance was 0.660 on a scale where 0.5 is no discrimination and 1.0 is perfect discrimination (95% CI 0.651 to 0.669).
However, this was statistically higher than that of the standard Gail model (0.613, 95% CI 0.604 to 0.622).
The breast density model slightly underestimated breast cancer rates in younger women, Asian women, and Hispanic women (expected-to-observed ratios 0.94, 0.95, and 0.94, respectively).
Adding breast density to risk predicted by age, race or ethnicity, family history, and history of breast biopsy reclassified 22% of women correctly either to a higher-risk category for those with cancer or to a lower-risk category for those without cancer.
However, the addition of breast density also incorrectly reclassified 16% of women.
Compared with the Gail model, the breast density model correctly reclassified 14% of women and incorrectly reclassified 35% of women.
And while the breast density model increased the true-positive rate from 28% to 53% and the positive predictive value from 2.3% to 2.4%, it also almost doubled the false-positive rate from 17% to 30%.
"Further comparisons of the two models in additional populations will help to clarify their relative value," the researchers said.
The study was supported by cooperative agreements with the National Cancer Institute-funded Breast Cancer Surveillance Consortium and by a Building Interdisciplinary Research Careers in Women's Health faculty development grant.
Dr. Tice reported receiving grant support from Building Interdisciplinary Careers in Women's Health. A co-author reported consultancies, honoraria, and grant support from Eli Lilly as well as grants from the Lilly Foundation.
Primary source: Annals of Internal MedicineSource reference:Tice JA, et al "Using clinical factors and mammographic breast density to estimate breast cancer risk: Development and validation of a new predictive model" Ann Intern Med 2008; 148: 337-347.