Introduction
Established risk factors for the development of breast cancer include components incorporated into the Gail model, breast mammographic density, and cellular atypia. Mammographic density is an important biomarker of risk for the development of breast cancer and, because it is modifiable, it is a potential response biomarker as well. Cytologic atypia is an established risk factor for the development of breast cancer. A prospective study conducted in high-risk women employed random periareolar fine needle aspiration (RPFNA) to sample breast tissue [
1]. It revealed that women with RPFNA atypia had a fivefold increased risk for subsequent clinical development of ductal carcinoma
in situ (DCIS) or invasive cancer as compared with those without atypia, and RPFNA atypia stratified risk based on the Gail model [
1]. Both mammographic density and cellular atypia are risk biomarkers that can stratify estimates based on the Gail model but they have limitations, particularly when they are used as surrogate markers of response, which include interpretive variance (both biomarkers), lack of categorical change (cellular atypia), and lack of change with some effective interventions compared with placebo in postmenopausal women (mammographic density) [
2,
3].
Increased proliferation is a fundamental process in carcinogenesis. Shabaan and coworkers [
4], in a cross-sectional study, observed that women with increased Ki-67 in foci of hyperplasia were at increased risk for breast cancer. Reduction in proliferation has been shown to correlate with response to antihormonal agents in cancer treatment trials [
5]. Ki-67 expression is currently being used in phase II breast cancer chemoprevention trials. The rationale behind the use of Ki-67 as a response biomarker in phase II proof-of-principle trials would be stronger if correlation could be established with development of cancer or with other biomarkers associated with a substantial increase in risk (atypical morphology and high mammographic density, for instance). On the other hand, if no strong correlation could be established, then these two biomarkers may be regarded as independent and potentially complementary risk or response biomarkers. We have previously shown that Ki-67 expression in benign epithelial cells is positively correlated with epithelial cell number and cytomorphologic abnormality in women at increased risk for breast cancer [
6]. In this analysis, we examine the correlation between Ki-67 and mammographic density.
Discussion
Breast mammographic density is reflective of the amount of epithelium, stroma and breast fluid relative to fat (which is radiolucent). The volume of stroma and collagen in most women influences density to a greater extent than the amount of breast epithelium [
11,
12]. Mammographic density is positively associated with several other risk factors and biomarkers, including breast intraepithelial neoplasia [
13], serum insulin-like growth factor-I and growth hormones in premenopausal women [
14], serum prolactin and combined estrogen plus progestin HRT in postmenopausal women [
15,
16], and family history of breast cancer [
17].
Boyd and coworkers [
10], in a case control study using computer-assisted measurements, found that statistically significant increases in breast cancer risk were associated with increasing mammographic density. The increment in relative risk for breast cancer for each percentage increase in density was 2% (
P < 0.0001) and the relative risk for greater than 75% density relative to no density was 4.04 (95% confidence interval 2.12 to 7.69). Breast density is favorably modulated by some but not all drugs/interventions that are effective in the prevention and adjuvant treatment of breast cancer. Tamoxifen was associated with a 14% reduction in absolute breast density over 54 months, as compared with an 8% reduction for placebo-treated women in the IBIS-1 (International Breast Cancer Intervention Study-1) trial [
18], in a cohort of women with greater than 10% density. Changes in mammographic density favoring tamoxifen were significant only in premenopausal women and those under the age of 55 years. Similarly, two years of a low-fat diet was demonstrated to reduce the area of breast density for premenopausal but not postmenopausal women [
19], despite the observation in the WINS (Women's Interventional Nutritional Study) study [
20] that such an intervention significantly reduced the risk for recurrence, including contralateral breast cancer, in postmenopausal women [
20]. Although breast density is clearly a risk factor for both premenopausal and postmenopausal women, its accuracy in predicting response to a preventive intervention is less clear, particularly for postmenopausal women. It seems plausible, based on the evidence, that interventions associated with reduced density are likely to be effective preventive agents.
Because prospective prevention intervention studies with cancer as an end-point are expensive and lengthy, surrogate response biomarkers are often used in phase II chemoprevention trials, in which favorable modulation of biomarker by an agent is taken as support for that agent's ability to reduce the incidence of cancer. A response biomarker ideally should also be a risk biomarker in addition to being modifiable. Mammographic density is an established risk factor for breast cancer, as noted above, and is modifiable. Ki-67 expression in benign breast cells obtained from RPFNA is also a reversible, a potential risk and a possible surrogate response biomarker. In a cohort of 147 high risk women, we previously showed that cytomorphologic atypia in benign breast cells obtained by RPFNA is associated with increased Ki-67 expression [
6]. Median Ki-67 expression was 2.8% in women with RPFNA atypia as compared with 1.1% in women without atypia. In the present study, which now includes 344 women, this correlation between Ki-67 and cytomorphologic atypia persists and corroborates our previously reported data. Whereas proliferation appears to be linked to cytologic atypia, it is not clear whether there is a link between mammographic density and proliferation or cytomorphology. Study of such a correlation is important because both mammographic density and Ki-67 are currently being used in breast cancer prevention trials as surrogate response biomarkers [
21,
22].
We found no correlation between mammographic density and Ki-67 or mammographic density and cytomorphologic atypia in benign breast cells in a cohort of high-risk women for whom sufficient cells from RPFNA were available for both cytomorphology and Ki-67 testing. No study has previously been undertaken to identify such a correlation. Many epidemiologic studies have evaluated a relationship between mammographic density and benign breast histology, with some studies showing an association between histology and mammographic density, whereas others have shown no such association. In a cohort of women taking part in the Canadian NBSS (National Breast Screening Study) study [
23], proliferative breast disease was found to be more frequent in women with greater breast density. A similar association was noted in a study reported by Bland and coworkers [
24]. Two other studies [
25,
26], on the other hand, identified no correlation between histology and mammographic density. In a nested case-control study within the prospective Breast Cancer Detection Demonstration Project, percentage mammographic density and benign breast disease histology were found to be distinct breast cancer risk factors. The risk associated with benign breast disease was not explained by the effects of percentage breast density, and the risk associated with percentage breast density was not explained by benign breast histology, suggesting a lack of correlation between the two risk factors [
27]. Several cross-sectional studies have described an association between histology and mammographic density, with different results [
25,
28‐
31]. Fisher and coworkers [
28] compared histology and mammographic appearance of breast in women with cancer and women with fibrocystic disease, and found no association between epithelial change and mammographic density. They found that mammographic densities were associated with fibrosis in breast parenchyma. A similar lack of association was described by another study [
25]. In contrast to these studies finding no association, Bright and coworkers [
30] reported associations between mammographic density and epithelial hyperplasia when xerographic and histologic findings in women with benign breast disease were compared. Similarly, Urbanski and colleagues [
31] described an association between atypia and extensive mammographic density.
Association between proliferation in benign breast and mammographic density is less well studied. In a recently reported study, Ki-67 (MIB-1) expression was assessed in areas of low, medium, and high mammographic density in benign breast tissue obtained from reduction mammoplasties. Contrary to what might be expected, Ki-67 expression in epithelial cells was less in the areas of medium and high density as compared with the areas of low density [
32]. In another prospective study of association between mammographic density and benign histology [
33], mammographically dense and nondense (fatty) tissues contained similar frequencies of hyperplasia with atypia and proliferative activity, as determined by S-phase percentage. These latter observations suggest a lack of strong correlation between mammographic density and proliferative activity within the breast. Our findings with random tissue sampling are consistent with these findings.
Our cohort of 344 women includes 114 women who were on HRT, which could be a potential confounding factor. However, only 16 women were taking a combination of an estrogen and a progestin. Combined estrogen plus progestin HRT, and not estrogen alone, is associated with increases mammographic density in postmenopausal women [
15,
16]. We therefore do not believe that HRT status had any significant impact on our results, namely a lack of correlation between mammographic density and Ki-67 expression. Our cohort also included 39 women with a history of prior breast cancer. Cancer treatment such as endocrine therapy or premature menopause from chemotherapy could potentially have a confounding effect on our results. However, only 11 women in our cohort had invasive cancer, eight women received chemotherapy, and four women took tamoxifen. Given the small number of women receiving interventions that could have confounding effect on mammographic density, we do not believe that inclusion of these women in the cohort influenced our findings. Furthermore, we ran an analysis excluding these women with prior cancer and there was no difference in the results.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
QJK contributed to the study design, read all the mammograms and drafted the manuscript. BFK performed the statistical analysis and contributed to the study design. AOD organized and collected the data for analysis. CMZ made all the cytology assessments. PS performed RPFNA to obtain cytology specimens. CJF performed RPFNA, contributed to the study design, and served as a mentor for the entire project.