|Jump to:||Authorized Access|||||Attribution|||||Authorized Requests|
- Study Description
Description of Cohort: The California Pacific Medical Center (CPMC) Breast Health Cohort is a cohort study based at CPMC and is linked to the San Francisco Mammography Registry, one of the sites of the NCI-funded Breast Cancer Screening Consortium (U01CA063740). CPMC is a community hospital in San Francisco, which has one of the highest volumes for mammography in San Francisco. Between September 2004 and June 2007, >90,000 mammograms were performed at CPMC. The CPMC breast health cohort collects demographic and risk factor data on women receiving mammography through participation in the San Francisco Mammography Registry, as part of the Breast Cancer Screening Consortium (U01CA063740). The SFMR database collects information from all sources, including a questionnaire on demographic and risk factor information, the clinical results of the breast examination, the measures of breast density by Dr. John Shepherd and the women who agreed to donate a blood sample. By merging these various sources of information we have very efficiently developed a large sample of women who have donated blood and have had a measure of mamographic density.
Blood Collection: Dr. Steve Cummings is leading an effort to collect and archive blood samples from women who are receiving mammography screening. All women who are sent for a screening mammogram at CPMC are considered eligible. Since the cohort began collecting blood samples in July 2004 until June 2007, samples have been collected from over 11,000 women.
Measurement of Breast Density: Dr. John Shepherd is currently measuring breast density in a large fraction of the cohort using an automated approach with single X-ray absorptiometry. Dr. Shepherd has established a link with the CPMC mammography center that allows him to collect routine digital mammography information. Using the data from the mammogram, Dr. Shepherd and his group have developed the single X-ray absorptiometry (SXA) technique for measuring density which is described in more detail below.
The table demonstrates the distribution of demographic variables and some breast cancer risk factors of women who donated blood and had a breast density measurement in the CPMC breast health cohort. Nearly 80% of the participants are Caucasian and most of the women are post-menopausal with a median age of ~52. Since it will be difficult to accrue a large enough sample from each ethnic group, our study will focus only on Caucasian women.
Table: Demographic variables, reproductive history and family history of breast cancer among 2962 women participating in the CPMC cohort study who contributed blood samples between 1994-1997.
Variable Median/Percentage Age (Median/IQR) 52 (46-59) Ethnicity Caucasian/White 0.76 Asian/Pacific Islander 0.141 Hispanic 0.029 Mixed Race/Ethnicity 0.039 African American/Black 0.022 American Indian 0.001 Other 0.009 First degree relative with breast cancer 0.17 Age at first birth Nulliparous 0.39 Age<20 0.043 Age>40 0.032 Age<30, ≥20 0.251 Age>30, ≤40 0.282
Measurement of Breast Density in Cohort: Measurement of breast density is accomplished using an automated technique for all mammograms obtained by Dr. Shepherd using Single X-Ray absorptiometry (SXA). SXA measurement of breast density is done on approximately 30% of all screening mammograms. Below we describe the method for measurement of breast density by SXA by Dr. Shepherd's group and its validation and association with breast cancer. As we demonstrate below, breast density, as measured by SXA, is an automated, highly reproducible measure of the density of breast tissue and is associated with breast cancer risk.
SXA for Quantifying Breast Density: Single x-ray absorptiometry (SXA) was initially developed for measuring bone density. SXA can determine the fraction of each of two densities simultaneously using the fact the sample is a constant thickness, the thickness in known, and the total attenuation is known. In applying this technique to breast density, we assume a two compartment model: fat and non-fatty (fibroglandular tissue). We use a reference material composed of various concentrations of two materials: one which is the same density as fat and another which is the same density as fibroglandular tissue. The reference material (phantom) is placed in the X-ray field with each mammogram. We have been able to implement this in a way that is unintrusive to the patient and technologist at CPMC.
Assuming this two-compartment model and a constant known breast thickness, we can then calculate the percent density at any region of the breast based on the assumption that % pixel grey-scale is proportion to the mass fractions of breast fat and lean tissue. If reference materials (a phantom) of fat and fibroglandular tissue are imaged with the patient's breast and the reference materials have the same thickness as the patient's breast, then the breast's grey-scale values can be converted to fat/fibroglandular mass fractions by interpolating between those two references. The total percent density is found by averaging the volume fraction over all breast pixels.
The phantom being used for breast density assessment at CPMC began to be used in September 2004. The phantom does not have to be manipulated by the technologist and stays attached on the mammography device during standard craniocaudal (CC) views. Thus it creates minimal to no interference with the clinical mammogram.
Reproducibility of breast density measures: Traditional measures of mammographic density require some human interpretation. A human reader outlines the area perceived to be dense and a computer then calculates the percent area outlined as a percent of the entire image. Thus, while traditional mammographic density is associated with breast cancer risk, it has some limitations. In a study by Drs. Shepherd, Kerlikowske, et al., the correlation coefficient (Pearson's R) between different readers was 0.8-0.9.
In contrast to the traditional mammographic density measurement, the SXA measurement is fully automated and, therefore, the reproducibility of the measurement is higher. Dr. Shepherd and colleagues have performed a replication study of SXA as a measurement of breast density. They have estimated the correlation coefficient of the SXA measurement of breast density to be >0.98. Thus, as expected for an automated measure, SXA is a highly reproducible measure of mammographic breast density.
Drs. Shepherd and Kerlikowske have recently analyzed the association between breast cancer risk and breast density as measured by SXA (Shepherd et al., Cancer Epi Biomarkers and Prev, 2011, PMID: 21610220). They found that women in the highest quintile of % volumetric density had an odds ratio of 4.1 (95% CI: 2.3 - 7.2) for breast cancer risk compared to women in the lowest quintile of volumetric density. Thus volumetric density appears to be a highly reproducible, automated measure of breast cancer.
- Study Type: Nested Case-Control
Number of study subjects that have individual level data available through Authorized Access: 982
- Authorized Access
- Publicly Available Data (Public ftp)
Connect to the public download site. The site contains release notes, manifests, documents, data dictionaries, variable summaries, and truncated analyses.
- Study Inclusion/Exclusion Criteria
Inclusion criteria: Women who have consented for genetic analyses and donated blood samples, have completed the SFMR risk factor questionnaire, have described their race/ethnicity as "Caucasian" or "white", and have at least one mammogram with SXA measurement of density and whose age and BMI-adjusted breast density measurement is either in the top 20%tile or bottom 20%tile of the distribution of breast density will be considered eligible.
Exclusion criteria: Women who do not complete key aspects of the SFMR including race/ethnicity information BMI and/or age/date of birth information will be considered ineligible.
- Molecular Data
Type Vendor Platform Number of Oligos/SNPs SNP Batch Id Comment Whole Genome Genotyping ILLUMINA HumanOmni1_Quad_v1-0_B 1051295 1049033
- Study History
The current study is a modified case-control design, where cases are defined as women in the highest quintile of age and BMI-adjusted breast density and controls are defined as women in the lowest quintile of age and BMI-adjusted breast density. To derive we begin with the % breast density as measured by SXA (described in the citations). For the sake of this study we used % density as opposed to total volumetric density. We then took the square-root transformed SXA % density (to make the distribution approximately normal). We then created a linear regression model between age, BMI and transformed breast density. We then took the residual of the linear model, as the age and BMI-adjusted breast density. The top and bottom quintiles of this measurement were used.
Age and BMI-adjusted measures are used since these are both felt to be "nuisance variables" that affect density greatly but negatively confound the association with breast cancer. Other non-genetic factors that affect breast density (parity history) only modestly affect breast density so their effect was not eliminated when creating the sampling approach.
- Selected publications
- Diseases Related to Study (MESH terms)
- Links to Other NCBI Resources
- Authorized Data Access Requests
- Study Attribution
- Elad Ziv, MD. University of California, San Francisco, CA, USA
- Steven Cummings, MD. California Pacific Medical Center Research Institute and University of California, San Francisco, CA, USA
- Karla Kerlikowske, MD. University of California, San Francisco, CA, USA
- John Shepherd, PhD. University of California, San Francisco, CA, USA
- National Cancer Institute, National Institutes of Health, Bethesda, MD, USA
- P01 CA107584. National Institutes of Health, Bethesda, MD, USA
- R01 CA120120. National Institutes of Health, Bethesda, MD, USA
- Johns Hopkins University Center for Inherited Disease Research (CIDR), Baltimore, MD, USA
Funding Source for Genotyping
- HHSN268200782096C. "NIH contract High throughput genotyping for studying the genetic contributions to human disease". National Institutes of Health, Bethesda, MD, USA
- HHSN268201100011I. "NIH contract High throughput genotyping for studying the genetic contributions to human disease". National Institutes of Health, Bethesda, MD, USA
Genotyping Quality Control
- Genetics Coordinating Center. Dept. of Biostatistics, University of Washington, WA, USA
- Principal Investigator