Compositional Breast Density as a Risk Factor

Institution: University of California, San Francisco
Investigator(s): Steven Cummings, B.A., M.D. - Karla Kerlikowske, M.S., M.D. - John Shepherd, Ph.D. -
Award Cycle: 2002 (Cycle VIII) Grant #: 8XB-0146 Award: $100,000
Award Type: SPRC Exploratory
Research Priorities
Detection, Prognosis and Treatment>Imaging, Biomarkers, and Molecular Pathology: improving detection and diagnosis
Etiology and Prevention>Etiology: the role of environment and lifestyle
Etiology and Prevention>Prevention and Risk Reduction: ending the danger of breast cancer



Initial Award Abstract (2002)
The appearance of the female breast on a mammogram varies because of differences in the relative amounts of fat, connective and epithelial tissue. Fat appears radiolucent or dark on a mammogram, whereas connective, and epithelial tissues are radiologically dense and appear lighter or white. The radiologically dense area, or mammographic breast density, is perhaps the strongest and least recognized risk factor for breast cancer. Mammographic breast density provides a better estimate of breast cancer risk than do standard "risk factors," especially in postmenopausal women. Current methods that measure mammographic breast density are neither precise nor accurate enough to be used in individual women to classify their risk status, or to follow changes in mammographic density based on preventive interventions. There is a need for new technologies that are inexpensive, convenient, and precise that allow patients and their clinicians to better assess risk of breast cancer.

Our central hypothesis is that breast compositional density, measured using novel techniques developed for bone densitometry, can be used as an accurate and precise method for estimating risk of breast cancer. This study is a preliminary investigation into this hypothesis. Twenty-five women with cancer and 25 without cancer will have mammographic breast density measured using two novel techniques, Dual X-ray Absorptiometry (DXA) and Single X-ray Absorptiometry (SXA). Our SXA method incorporates a phantom that is automatically adjustable in height to match the compression thickness of the breast during a mammogram. This innovation of the investigators is patent pending. We will measure the precision of each technique and calibrate them to available fat and glandular tissue reference materials. The DXA and SXA techniques will be compared to conventional mammographic density measurements and risk of breast cancer will be calculated for each method.

To the investigators' knowledge, these novel densitometric techniques have never been applied to measuring mammographic breast density and offer the possibility of making measurement of breast density universally available to women who receive routine mammography.


Final Report (2006)
The purpose of this study was to develop x-ray absorptiometry techniques to measure breast compositional density as a risk factor for breast cancer. Other than age, mammographic breast density has been shown to be one of the strongest available indicators of breast cancer risk. The current approach to measure breast density (BD) is referred to as mammographic percent density. We proposed a more precise and accurate method based on tissue compositional density. Our approach was to use x-ray techniques developed for bone densitometry called dual and single x-ray absorptiometry (DXA & SXA). These methods used a two-compartment composition model of breast tissue as fat and non-fatty lean (gland, stroma) tissue to quantify breast compositional density in terms of percent fibroglandular density.

Aim 1 was to developed these procedures and methods. These methods are fully developed and outlined in this report. Aim 2 was to compare the different breast density techniques to each other by looking at the difference between cases and controls.

We recruited 18 women with breast cancer in one breast and 18 controls. Each woman underwent clinical screening mammography and DXA imaging of the unaffected breast. The controls had one additional craniocaudal mammogram (x-ray view from the head toward the toes) and two DXA scans of the other breast. The mammograms contained a novel SXA calibration “phantom’, consisting of atomically-equivalent adipose fat and glandular tissue, to calibrate each pixel to % fibroglandular density. In addition, we quantified mammographic density (ratio of dense area to total area) for each mammogram. The DXA scans were acquired on a standard bone densitometry but were analyzed using our own analysis algorithms. The cases and controls were not significantly different in terms of age, weight, height, fraction gone through menopause, or race (p values > 0.05). We didn't expect to have the power to show significant differences between these groups for mammographic density. However, the trend in all the measures, DXA, SXA, and mammographic density, was for the women with cancer to have an elevated breast density.

In addition, in work supported by this study using the SXA methods, we tested the association between breast density via SXA and age, family history of breast cancer, age at first live birth, menopause, postmenopausal hormone therapy and body mass index adjusting for age. We found that among 1764 women, 7% had a BDSXA measure of <10%, 33% of 10-25%, 41% of 26-50%, 17% of 51-75%, and 2% greater than 75%. We found strong relationships between BDsxA and age (P< 0.001), current postmenopausal hormone therapy (P= 0.05), menopause (P< 0.001), and body mass index (P< 0.001). BDSXA was not associated with age at first live birth or family history of breast cancer. We conclude that BDsxa is associated with characteristics that are known from previous work to be related to mammographic breast density, a strong predictor of breast cancer risk. This suggests BDsxA will be associated with breast cancer risk. Aim 3 was to quantify the least significant change (LCS) needed between measures to have 95% confidence that the change is not due to random errors. Using the repeated mammograms and DXA scans on the same breasts, we found the LSC to be smallest for DXA (3.3%), 5.4% for SXA, and 7.5% for mammographic density. Aim 4 was to contrast the density methods to each other. We found that each method was highly correlated to a similar Pearson's correlation coefficient of 0.8. In conclusion, we developed and demonstrated the SXA breast density tool and showed that it is accurate and linear over a broad range of breast densities. Recruitment of women in the weeks between diagnosis and surgery proved to be very difficult. Adding more subjects would have strengthened our findings in specific aims 2, 3, and 4. However, we are highly encouraged by our results.

Question: Why were both single and dual x-ray absorptiometry (DXA & SXA) used in this study? Answer: The investigators did not know which of the two methods would be better for monitoring change in breast density, or which had a higher absolute accuracy. DXA does not require compression of the breast and is easily done (after commercial development) with devices like bone densitometers. In the US, there are over 30,000 devices that potential could do breast densitometry. However, the investigators were very unsure if women would have a separate breast exam using x-rays no matter how low the dose. Thus, the team is looking at the same measure using the SXA method as part of clinical mammography.

SXA is more challenging since more assumptions are necessary for accurate interpretation of results and sometimes those assumptions are not quite correct (the team has incorporated what they have learned using the CBCRP grant into a cohort study of 30,000 women. While there is still MUCH more development to do, but just the fact that they were able to have SXA on all clinical mammograms at a busy clinical site like this speaks volumes for what is preferable. It appears that the paradigm that will be used is SXA or a derivative of it for standard of care risk assessment in the future, and DXA for more detailed investigations of breast density. For example, the investigators are using DXA in a Program Project Grant with the NCI on the biological basis of breast density


Symposium Abstract (2003)
Other than age, mammographic breast density has been shown to be one of the strongest available indicators of breast cancer risk. The current approach to measure breast density is referred to as mammographic percent density. It is the ratio of projected dense breast area to the total area. Although mammographic percent density is a widely used technique for large trials, it has serious limitations for clinical use. We have proposed a more precise and accurate method based on tissue compositional density. Our approach is to use x-ray techniques developed for bone densitometry called dual (DXA) and single (SXA) x-ray absorptiometry. These methods use a two-compartment composition model of breast tissue as fat and non-fatty lean (gland, stroma) tissue. The ratio of fat to lean mass is determined for each image pixel. Summing all pixels in the mammogram results in a total percent glandular density (%GD).

This study involves 25 women with uni-lateral breast cancer and 25 matched controls, each undergoing standard mammography and the SXA and DXA techniques. The DXA measures have previously been shown to be accurate with cadaver breast and will serve as a reference for comparisons to the SXA results. The DXA method takes a prone pendulous view of the breast by placing a positioning box on a densitometer, allowing only the breast to be scanned. Images are then analyzed using the Rat/ Small Animal analysis option recalibrated to percent glandular density. SXA images are acquired during a routine screening mammography visit. A reference phantom consisting of atomically-equivalent adipose fat and glandular tissue is placed in the mammography image field and conforms to the same thickness as the compressed breast. Breast glandular density can then be derived from the x-ray attenuation by interpolating between the two reference materials.

We have acquired in vivo results on three subjects. The SXA methodology was easily incorporated into a routine mammography screening visit, without interfering with the breast image or the positioning of the breast by the technologist. The DXA method was successful at capturing the entire image of the breast. We believe our methods can be well integrated in routine screening mammography or bone density assessment to identify a women’s individual breast cancer risk. Our goal is to describe the relationships of breast compositional density methods to mammographic percent density and BI-RADS categorical scoring systems in terms of cancer risk sensitivity, correlations, and the ability to monitor drug therapies.

Compositional breast imaging using a dual-energy mammography protocol.
Periodical:Medical Physics
Index Medicus: Med Phys
Authors: Laidevant AD, Malkov S, Flowers CI, Kerlikowske K, Shepherd JA.
Yr: 2010 Vol: 37 Nbr: 1 Abs: Pg:164-74

Measurement of breast density with dual X-ray absorptiometry: feasibility
Periodical:Radiology
Index Medicus: Radiology
Authors: Shepherd JA, Kerlikowske KM, Smith-Bindman R, Genant HK, Cummings SR
Yr: 2003 Vol: 223 Nbr: 2 Abs: Pg:554-7