American Journal of Epidemiology Advance Access originally published online on April 2, 2008
American Journal of Epidemiology 2008 167(9):1037-1040; doi:10.1093/aje/kwn062
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Invited Commentary: Assessing Breast Density Change—Lessons for Future Studies
From the Cancer Genetics and Epidemiology Program, Lombardi Comprehensive Cancer Center, Georgetown University, Washington, DC
Correspondence to Dr. Celia Byrne, Cancer Genetics and Epidemiology Program, Lombardi Comprehensive Cancer Center, Georgetown University, S-148 Lombardi Building (LL), 3800 Reservoir Road NW, Washington, DC 20057-1465 (e-mail: cb252{at}georgetown.edu).
Received for publication December 17, 2007. Accepted for publication February 26, 2008.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMEASUREMENT OF BREAST DENSITY...SECULAR CHANGES IN EXPOSURESGENERALIZABILITY,...References |
|---|
Breast density is one of the strongest predictors of breast cancer risk, and quantitative measurement is fairly reproducible. However, to study change in breast density, other issues should also be considered. Most studies of breast density have relied on one assessment, yet the mammographic features of the breast that constitute breast density change with age and/or menopause. When measuring breast density change, issues related to assessment are of greater concern. In addition, because age-period and cohort effects are codefined, evaluation of age trends must also consider the possible explanations of period and cohort effects. The prevalence of different factors affecting breast density changed dramatically over the last 50 years. In this issue of the Journal (Am J Epidemiol 2008;167:1027–1036), Kelemen et al. evaluate how factors known to be related to breast density influence breast density change with age. These authors are to be complimented on their detailed analysis and consideration of many of these issues. They not only describe the averaged effects of age on breast density changes but also consider whether patterns of density change differ for women with different exposure histories.
breast; mammography; radiographic image interpretation, computer-assisted; risk factors
Abbreviations: BMI, body mass index; PMH, postmenopausal hormone
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMEASUREMENT OF BREAST DENSITY...SECULAR CHANGES IN EXPOSURESGENERALIZABILITY,...References |
|---|
In this issue of the Journal, Kelemen et al. (1) evaluated how factors known to be related to breast density influence the change in breast density with age. The researchers evaluated mammograms obtained between 1990 and 2003 from the Minnesota Breast Cancer Family cohort. The data covered an age range of more than 50 years (ages 40–90). Breast density information was derived from an average of 3.4 mammograms per woman to evaluate age trends. Thus, as described, these authors combined a mixture of short-term longitudinal data from each woman in a cross-sectional manner across women to model the trends for a wide age range. They used cubic spline models to allow for nonlinear trends between the ages of 43 and 77 years.
Studies have evaluated the effects of breast density change on breast cancer risk, with inconsistent findings (1–4). Some reported no additional influence on risk beyond the effects of baseline percent breast density (5), while others found an additional risk with a measure of change (4). Breast density is one of the strongest predictors of breast cancer risk, and quantitative measurement is fairly reproducible in most studies. Given the fourfold to sixfold magnitude of the associations between baseline percent density and breast cancer risk, the effects of change would need to be quite strong and persistent to be detected beyond those of baseline breast density. Furthermore, issues related to assessment of breast density become of greater concern when considering breast density change. Because of the high correlation between age and menopause transition and the difficulty in clearly categorizing menopause status, findings that menopause has an independent effect on change in breast density beyond the age effect have not been consistent across studies, with some reporting both age and menopause effects and others finding only age effects (2, 6, 7).
In many previous cross-sectional studies of breast density that used only one assessment of breast density for each woman, measured breast density was lower in women who were postmenopausal, older, heavier (higher body mass index (BMI)), and parous, particularly those with an early age at first birth (8, 9). Previous studies demonstrated that, despite potential changes in breast density with age and menopause, a premenopausal mammogram was about as strong a predictor of developing postmenopausal breast cancer as a postmenopausal mammogram (8, 10). Given those findings, clearly baseline breast density would likely be a strong predictor of subsequent breast density. Thus, given that baseline breast density was included when modeling breast density change, only factors that add to change beyond their influence on baseline breast density need to be considered.
Kelemen et al. (1) found age at menarche, parity, and age at first birth, as well as alcohol intake, not to be associated with breast density change, although these factors were associated with baseline breast density. Given that baseline percent density was shown to influence density change, such that those who start out with higher breast density have a greater absolute change, the influence of parity and age at first birth on change is already captured by baseline density. It is understandable that age at menarche and age at first birth and parity would likely influence baseline breast density, but since those factors generally are not changing (although some women have a later age at first birth or increased parity after starting to get mammograms), it is unlikely that they will substantially influence breast density change later in life.
However, women may change their exposure level to other related factors such as BMI, postmenopausal hormone (PMH) use, and pattern of alcohol intake during this time of life. Because BMI and either past or current combined estrogen and progesterone PMH use are strongly related to baseline breast density, detailed and accurate information regarding changing exposures at the time of each mammogram may be necessary to truly evaluate the impact of these factors on breast density change beyond baseline. However, because most women tend to gain weight as they age, variation in BMI change in the study population may be insufficient to evaluate an independent effect on breast density change. This study was not able to evaluate changes in alcohol intake patterns at different times in life given that women were asked only what their usual alcohol consumption was during their life (1). Given previously reported combined effects of alcohol consumption and PMH use on circulating levels of steroid hormones, this issue should be investigated further. Thus, to evaluate the impact on breast density change, change in factors such as BMI, PMH use, alcohol consumption, and others that vary over time must be measured serially and accurately and the distribution of change must vary sufficiently within the study population for the role of these factors to be evaluated.
Given the consideration of many to modify breast density as a mechanism to possibly alter an individual's risk profile, it would be important to know whether changes in physical activity or diet would modify changes in breast density. A recent Australian study, which did not find an independent effect of menopausal change on breast density after controlling for age, did find that free testosterone levels were associated with breast density change (6). Thus, despite the various factors considered in the study by Kelemen et al. (1) in evaluating those that influence breast density change, perhaps other factors should be considered in studies of other population groups.
|
MEASUREMENT OF BREAST DENSITY AND BREAST DENSITY CHANGE |
|---|
|
TOPABSTRACTINTRODUCTIONMEASUREMENT OF BREAST DENSITY...SECULAR CHANGES IN EXPOSURESGENERALIZABILITY,...References |
|---|
As in most studies using a quantitative assessment of breast density, the measurement is highly reproducible within this study; Kelemen et al. (1) took numerous steps to both assess variation in measurement throughout their study and consider the impact on the findings. Only one technician was used to assess all breast density measures, and the correlation for repeated assessment was >0.90. For an epidemiologic measure, one would often consider this assessment as very good. The median percent density for premenopausal women in this study was lower than that measured in other US, Canadian, and European study populations. The Minnesota Breast Cancer Family cohort may differ from other populations, or the assessment of breast density, while reproducible, may tend to compress the range of density downward. If the scale of percent density is compressed, then the ability to detect change separate from measurement variation will be more and more difficult. Reproducibility, although very important, may not be the only measure to consider when evaluating the assessment of breast density. The authors indicated that the error was uniform across levels of breast density, with the absolute difference between two measures up to 9 percent across all density levels. However, this degree of error appears to be greater than the median differences in density between the first and last mammograms evaluated for each woman in the study. Therefore, the concern is that meaningful changed may be missed.
Another factor of concern in studies of breast density is the limitation of the scanner used in digitizing the mammograms to capture the breast edge on the mammograms. The authors (1) noted this issue and recognized that it would tend to make the measured breast area smaller in this study and thus the percent density greater, which raises larger doubt regarding the lower mean breast density readings in this study.
One other aspect is that the techniques and practices used to acquire mammograms are constantly changing as well. So, to the degree that newer mammographic techniques are "clearer," then breast density would likely be lower in the more contemporary mammography. Since this is the same direction as seen with age, there is a concern that some small portion of the density change with age may be attributable to improved techniques. When comparing change in two groups, it is thus important that comparable techniques be used to acquire concurrent mammograms. However, many facilities in the United States are changing to digital mammography from film-screen mammography. Some will use both modalities concurrently, while others will change all at once. The mixing of modalities and trying to assess density change add many other layers of complication. One study suggested that breast density assessed from a digital format is comparable to that from film screen (11), while another study indicated that they are not comparable (12).
|
SECULAR CHANGES IN EXPOSURES |
|---|
|
TOPABSTRACTINTRODUCTIONMEASUREMENT OF BREAST DENSITY...SECULAR CHANGES IN EXPOSURESGENERALIZABILITY,...References |
|---|
In this study (1), the levels of breast density contributed by premenopausal women, women in menopausal transition, and postmenopausal women were smoothed together to depict the patterns of breast density change between ages 40 and 90 years. Because none of the women provided mammograms over that entire age span, one assumes that premenopausal women had the same level of breast density that postmenopausal women had when they were premenopausal (and vice versa). However, these younger women were more likely taller (a factor related to growth factor levels), to have ever used oral contraceptives (85 percent compared with 49 percent), to be nulliparous, and to have a family history of breast or ovarian cancer. All of these factors, as well as a possible younger age at menarche, are likely to make the breasts of younger women denser than those of the currently postmenopausal women when they were premenopausal. In addition, the younger women are both less likely to undergo a surgical menopause and more likely to have a higher BMI, which would potentially lower the breast density of the younger women. These possible birth cohort effects were acknowledged by Kelemen et al. (1) but should be considered when interpreting the findings. Given the major influence of baseline breast density on the pattern of change with age, these issues need to be studied.
|
GENERALIZABILITY, INTERPRETATIONS, AND FUTURE CONCERNS |
|---|
|
TOPABSTRACTINTRODUCTIONMEASUREMENT OF BREAST DENSITY...SECULAR CHANGES IN EXPOSURESGENERALIZABILITY,...References |
|---|
Despite these extra considerations of potential misclassification and other measurement issues, understanding breast density change and determining perhaps different patterns or rates of change, as presented by Kelemen et al. (1), are of great interest not only in potentially improving the prediction of breast cancer risk (13) but also in helping to understand the disease etiology. Two studies have now shown that use of PMH and higher BMI predict a slower decline in density with age, providing insight into how BMI influences breast cancer risk (1, 2).
Figure 2 of the Kelemen et al. (1) paper represents the unadjusted distribution of breast density measured for this specific population in this study. However, the concern about this manner of presentation regards the tendency of others to read this graph in a fashion similar to that of age-specific growth charts of children. Others have proposed creating age-standardized breast density charts so that physicians can show patients how their breast density level compares with that of other women their same age. The problem is that we do not have a standardized agreement on measuring breast density. Growth charts work for children in that there is a standardized agreement on units (for height and weight). We do not have standardized assessments of breast density. Although readings of different individuals might be highly correlated and rank individuals in a similar fashion from low to high density, there can still be substantial absolute differences in their readings (14).
In addition, this study (1) was derived from an unusual cohort identified via the breast cancer probands. In all, 63 percent of those for whom mammograms were obtained were blood relatives of one of the original breast cancer probands. Although not in this study, other studies showed an association between having a family history of breast cancer and increased breast density (8).
While the greatest breast density change occurs during the transition from premenopause to postmenopause, only 10 percent of this study population (n = 175) were in this category (1), although these women did contribute longer follow-up time and thus more mammograms to the study. Furthermore, since 60 percent of these women initiated or continued PMH use, the patterns of "nature" change with age may not be as clear. Therefore, to better identify the patterns of change in this group, future studies may want to include a higher proportion of women likely to be transitioning to menopause. In all studies that rely on measures of mammographic breast density, it must be recognized that if changes in the breast that occur earlier in life are the most relevant to breast cancer etiology, then studies limited to the ages at which women typically undergo mammography will miss this exposure assessment. Mammographic breast density in women over age 40 years is a strong predictor of breast cancer risk (8), but factors that determine breast density at age 40 years may have occurred earlier in a woman's life and their influence on later breast density change may be missed.
Understanding and being able to accurately measure breast density change will be important for many studies of etiology and potential interventions. This study by Kelemen et al. (1) illustrates the need for consistent and reliable measures of breast density and time-dependent covariates, the need for studies with more images over a longer time period for each woman to avoid the problems of cross-sectional analyses, and the concern for caution when interpreting findings. As more studies consider the assessment of breast density change as an exposure, intermediate endpoint, or outcome, these issues need to be considered.
|
ACKNOWLEDGMENTS |
|---|
Conflict of interest: none declared.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMEASUREMENT OF BREAST DENSITY...SECULAR CHANGES IN EXPOSURESGENERALIZABILITY,...References |
|---|
- Kelemen LE, Pankratz VS, Sellers TA, et al. Age-specific trends in mammographic density: the Minnesota Breast Cancer Family study. Am J Epidemiol (2008) 167:1027–36.
[Abstract/Free Full Text] - Maskarinec Pagano I, Lurie G, Kolonel LNG, et al. A longitudinal investigation of mammographic density: the multiethnic cohort. Cancer Epidemiol Biomarkers Prev (2006) 15:732–9.
[Abstract/Free Full Text] - Salminen TM, Saarenmaa IE, Heikkilä MM, et al. Risk of breast cancer and changes in mammographic parenchymal patterns over time. Acta Oncol (1998) 37:547–51.[CrossRef][Web of Science][Medline]
- van Gils CH, Hendriks JH, Holland R, et al. Changes in mammographic breast density and concomitant changes in breast cancer risk. Eur J Cancer Prev (1999) 8:509–15.[Web of Science][Medline]
- Vachon CM, Pankratz VS, Scott CG, et al. Longitudinal trends in mammographic percent density and breast cancer risk. Cancer Epidemiol Biomarkers Prev (2007) 16:921–8.
[Abstract/Free Full Text] - Guthrie JR, Milne RL, Hopper JL, et al. Mammographic densities during the menopausal transition: a longitudinal study of Australian-born women. Menopause (2007) 14:208–15.[CrossRef][Web of Science][Medline]
- Boyd N, Martin L, Stone J, et al. A longitudinal study of the effects of menopause on mammographic features. Cancer Epidemiol Biomarkers Prev (2002) 11:1048–53.
[Abstract/Free Full Text] - Byrne C, Schairer C, Wolfe J, et al. Mammographic features and breast cancer risk: effects with time, age, and menopause status. J Natl Cancer Inst (1995) 87:1622–9.
[Abstract/Free Full Text] - Vachon CM, Kuni CC, Anderson K, et al. Association of mammographically defined percent breast density with epidemiologic risk factors for breast cancer (United States). Cancer Causes Control (2000) 11:653–62.[CrossRef][Web of Science][Medline]
- Boyd NF, Guo H, Martin LJ, et al. Mammographic density and the risk and detection of breast cancer. N Engl J Med (2007) 356:227–36.
[Abstract/Free Full Text] - Jeffreys M, Warren R, Smith GD, et al. Breast density: agreement of measures from film and digital image. Br J Radiol (2003) 76:561–3.
[Abstract/Free Full Text] - Harvey JA. Quantitative assessment of percent breast density: analog versus digital acquisition. Technol Cancer Res Treat (2004) 3:611–16.[Web of Science][Medline]
- Kerlikowske K, Ichikawa L, Miglioretti DL, et al. Longitudinal measurement of clinical mammographic breast density to improve estimation of breast cancer risk. J Natl Cancer Inst (2007) 99:386–95.
[Abstract/Free Full Text] - McCormack VA, Highnam R, Perry N, et al. Comparison of a new and existing method of mammographic density measurement: intramethod reliability and associations with known risk factors. Cancer Epidemiol Biomarkers Prev (2007) 16:1148–54.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
Related articles in Am. J. Epidemiol.:
- Age-specific Trends in Mammographic Density: The Minnesota Breast Cancer Family Study
- Linda E. Kelemen, V. Shane Pankratz, Thomas A. Sellers, Kathy R. Brandt, Alice Wang, Carol Janney, Zachary S. Fredericksen, James R. Cerhan, and Celine M. Vachon
Am. J. Epidemiol. 2008 167: 1027-1036.[Abstract] [FREE Full Text]
This article has been cited by other articles:
|
|
 |
|
  D. S.M. Buist, M. L. Anderson, S. D. Reed, E. J. Aiello Bowles, E. D. Fitzgibbons, J. C. Gandara, D. Seger, and K. M. Newton Short-Term Hormone Therapy Suspension and Mammography Recall: A Randomized Trial Ann Intern Med, June 2, 2009; 150(11): 752 - 765. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||