Mammographic Breast Density Response to Aromatase Inhibition

Breast density, a mammographic reflection of the fat, epithelial, and stromal composition of the breast, has been shown to be a promising marker for breast cancer risk in healthy women, with a higher percentage of breast density (greater than 50%) associated with a 3- to 5-fold increase in risk of breast cancer compared with lower levels (1). High mammographic breast density (MBD) at time of breast cancer diagnosis is also associated with increased risk of a local cancer recurrence (2, 3), and, among women with ductal carcinoma in situ, with an increased risk of a second breast cancer (4, 5).

Importantly, breast density may be modifiable. MBD has consistently been shown to increase with use of estrogen plus progestin menopausal therapy (6–8) and to decrease with exposure to tamoxifen (9–12). However, these changes in MBD do not occur in all women, suggesting that the variability could reflect differential response to these therapies. In fact, women at high risk for breast cancer within the IBIS-1 study, who experienced at least 10% reduction in MBD while on tamoxifen, had a reduced risk of breast cancer (relative risk = 0.51) compared with women who had no change in their densities (13); women on placebo who experienced similar reductions in MBD, however, did not have a decreased risk of breast cancer. A recent study in Korean women confirmed these findings among women in the adjuvant setting; a greater reduction in MBD with an average 13 months of tamoxifen or aromatase inhibitor therapy was associated with recurrence-free survival. Compared with those experiencing the largest reduction in MBD (>10%), there were 1.33, 1.92, and 2.25 increased risks of recurrence with 5% to 10%, 0% to 5%, and less than 0% reductions in MBD, respectively (14). These data suggest that change in MBD may be useful as a surrogate marker to identify women who may or may not benefit from certain endocrine therapies.

Aromatase inhibitors are a class of pharmaceutical agents established as adjuvant therapy for postmenopausal women with early-stage estrogen receptor-positive (ER) breast cancer (15). These agents are also beneficial in the prevention setting in terms of decreasing the incidence of primary breast cancers (16). Aromatase inhibitors block the conversion of androstenedione to estrone (E1) and testosterone to estradiol (E2) by cytochrome P450 (CYP) 19, aromatase, and have been shown to profoundly decrease levels of E1 and E2 in both serum and breast tissue (17–20). Our group recently showed that overall aromatase expression was increased in tissue cores taken from mammographically dense regions compared with nondense regions of the breasts of healthy women (21). This would suggest that MBD is influenced by aromatase expression and local estrogen synthesis and potentially could be used as a biomarker to assess the impact of aromatase inhibitor therapy.

If aromatase expression is increased in mammographically dense tissue, we hypothesize aromatase inhibitors will decrease overall MBD. The studies to date examining the influence of aromatase inhibitors on MBD are mixed (14, 22–28), with the majority finding no association (Table 1); however, most of these studies had small cohorts of patients. The largest study to date of 175 women on aromatase inhibitor therapy by Kim and colleagues showed small reductions in MBD with aromatase inhibitor use (average 3% at 1 year of therapy); however, their study was not able to evaluate the influence of body mass index (BMI), had a high proportion of prior chemotherapy use, a younger population (median age 49 years), and did not consider type of adjuvant aromatase inhibitor therapy (14).

In this report, we present data from the largest study to date to examine the influence of aromatase inhibitors on MBD in postmenopausal women. We examine changes in MBD among women on 3 clinical trials of patients with early-stage breast cancer treated with an aromatase inhibitor, evaluate factors associated with MBD changes, and compare them with MBD changes in an age-matched cohort of women undergoing routine mammography screening. We also examined whether our findings differ by the 2 classes of aromatase inhibitor therapy, steroidal versus nonsteroidal aromatase inhibitors.

Women with early-stage postmenopausal breast cancer enrolled onto National Cancer Institute of Canada Clinical Trials Group (NCIC CTG) study MA.27, North Central Cancer Treatment Group (NCCTG) N063I, or Mayo Clinic (MC) MC0532 who provided consent for mammogram retrieval were eligible for this study. All women were receiving anastrozole (1 mg orally once daily) or exemestane (25 mg orally once daily) in the adjuvant setting. Eligibility was determined following mammogram retrieval. A woman was considered to be ineligible for this study if she had aromatase inhibitor therapy for less than 6 months; bilateral breast cancer; breast implantation; surgery in noncancerous breast; lacked a mammogram taken within 12 months of the start of aromatase inhibitor; or lacked a mammogram taken after at least 6 months of aromatase inhibitor use that was the same image format (film vs. digital) as the pre-aromatase inhibitor mammogram. This last criterion was put in place as percentage density has been shown to differ with image formats (29). If multiple mammograms were available after the start of aromatase inhibitor, the mammogram closest to the day of 1 year of aromatase inhibitor use and with the same image format as the pre-aromatase inhibitor mammogram (baseline mammogram) was used. This image is referred to here as the year 1 mammogram.

Information on prior chemotherapy, prior postmenopausal hormone therapy use (estrogen alone or combination estrogen and progesterone therapy), baseline weight, and weight gain was available from self-reported questionnaires or in-person interviews. Knowledge of the blinded treatment arm (either anastrozole or exemestane) was also made available for patients enrolled to NCIC CTG MA27.

Because MBD naturally declines in postmenopausal women as they age, it was essential to have a comparison population closely matched with the cases initiating aromatase inhibitor therapy. Controls were drawn from participants in the Mayo Mammography Health Study (MMHS), a cohort of women having routine screening mammography at the Mayo Clinic (Rochester, MN), who had at least 1 screening mammogram before their enrollment mammogram (30). The MMHS enrolled more than 50% of the female residents of Minnesota, Wisconsin, and Iowa having screening mammography between October 2003 and September 2006 at the Mayo Clinic, who were over the age of 35 years, and had no prior history of breast cancer. To be considered as a potential control, a women had to have had a screening mammogram at least 12 months before enrollment in the MMHS (baseline mammogram) that was the same format as the mammograms taken at the time of enrollment into MMHS (the 1 year mammogram). Matching was done using a greedy algorithm providing the closest match available to the case. For each case, a control was chosen whose age at her baseline mammogram was within 5 years of the age of the case at her baseline mammogram; BMI category was the same as the case (underweight/normal vs. overweight/obese); interval between baseline and 1-year mammograms was within 120 days of that of the case; baseline mammographic density was within 10% quantitative MBD of the case (31).

All baseline (pre-aromatase inhibitor) and 1-year mammograms were digitized using an Array Digitizer with 12-bit grayscale depth. The pixel size was 0.130 mm × 0.130 mm for both the 18 cm × 24 cm and 24 cm × 30 cm films. The craniocaudal views of the noncancerous breast (or corresponding breast side in controls) at both time points were digitized on each woman and all mammograms were scrubbed of patient identifiers. Batch files were created so that baseline and year 1 images could be displayed simultaneously on side-by-side screens for assessment. The baseline and 1-year images were randomly assigned to left- or right-sided screen. A 5% repeat set of images was included within each batch file for assessment of reliability.

Percentage breast density (dense area divided by total area × 100) was estimated using a computer-assisted thresholding program that has routinely been used in several studies of breast density, including our own (32–35). Percentage density was assessed on batch files consisting of paired mammograms from each case and their matched control over a 4-month period by 1 evaluator (F.F. Wu) trained by Norman F. Boyd (University of Toronto, Toronto, Canada), an acknowledged leader in the field of MBD. Our evaluator has over 10 years' experience in density estimation, consistently shows associations with breast cancer using this density measure (34, 36, 37) and has shown high correlation with Dr. Boyd on over 600 images (data not shown).

Briefly, 2 thresholds are set by a trained programmer; the first separates the breast from the background and the second separates dense from nondense tissue. Both images were viewed simultaneously and assessed for percentage density, allowing the evaluator to flag pairs for which adequate comparisons could not be made (i.e., breast capture at 2 time points were substantially different). To assess the intrareader reliability in terms of determining change in percentage density, the evaluator read pre- and posttreatment images from 46 women at 2 different sessions. Bland–Altman plots were constructed to compare estimates of change at these 2 sessions. The difference in the change in breast density between the first and second sessions was within 5% for all, but 3 of 46 women [mean, 6.5%; 95% confidence interval (CI), 0.0%–13.7%].

We first examined all cases on aromatase inhibitor therapy to assess what factors are associated with the increased likelihood of a 5% or more reduction in percentage MBD (event) using multivariate logistic regression.

We next conducted analyses to examine whether there were differences in the 1-year change in density between the women on aromatase inhibitor therapy and their matched controls. A paired t test was used to assess whether the difference in the change in MBD over 1 year between the case and her matched control was significantly different than 0 for the following patient cohorts: (i) all the matched pairs; (ii) the matched pairs where both the case and her control had a baseline mammographic density of at least 10%; (iii) the matched pairs on exemestane; (iv) the matched pairs on anastrozole; (v) matched pairs where the case was on aromatase inhibitor therapy for at least 12 months; and (vi) matched pairs who had the same status of prior postmenopausal hormone therapy use (either both never or both ever). SAS 9.2 was used for all statistical analyses.

A total of 574 postmenopausal patients with breast cancer starting adjuvant aromatase inhibitor therapy were consented to participate in this study, where 505 patients (88.0%) were enrolled on the NCIC CTG MA27; 57 (9.9%) on MC0532 and 12 (2.1%) on NCCTG N063I. After mammogram retrieval process, 187 women were found to be ineligible due to: lacking a mammogram taken at least 6 months following start of aromatase inhibitor (89 pts.) or at most 12 months before starting an aromatase inhibitor (2 pts.); discontinuing aromatase inhibitor treatment before a mammogram being taken (17 pts.); and having mammograms in different formats (digital vs. film) at the 2 time periods (79 pts.). A total of 387 cases were eligible and used in case-only analyses. However, controls could not be matched for 18 of these women, thus a total of 369 matched pairs from MA27 (n = 342), NCCTG N063I (n = 7), and MC0532 (n = 20) were used for case–control comparisons (Table 2).

Of the 387 postmenopausal breast cancer women receiving an average of 10 months of adjuvant aromatase inhibitor therapy, 14% (n = 56) experienced at least a 5% decrease in their MBD. This increased to 20% for the 280 cases who had a baseline percentage density of at least 10%. Multivariate analyses showed the likelihood of experiencing a reduction of at least 5% in MBD with aromatase inhibitor therapy was increased for cases with a baseline density of 15% or more (OR, 10.4; 95% CI, 4.0–26.9; P < 0.0001), who had 12 months or more of aromatase inhibitor treatment (OR, 3.18; 95% CI, 1.68–6.03; P = 0.0004) and had any prior hormone therapy exposure (OR, 1.95; 95% CI, 1.03–3.67; P = 0.04). Baseline BMI, change in BMI between baseline and year 1 mammograms; age, time between mammograms and treatment arm were not associated with a reduction in MBD with aromatase inhibitor use (data not shown).

Table 2 describes characteristics of the 369 cases and matched controls used for the paired analyses, by parent trial and overall. Three fourths of the women were overweight or obese. Pre-aromatase inhibitor MBD was less than 10% for 29% of cases (37% controls) and 30% or more for 17% of cases (14% controls). Prior use of postmenopausal hormones was balanced between the cases and controls (53%).

The distribution of change in MBD between baseline and 1-year for the 369 cases and their matched controls is shown in Table 2 and Fig. 1. The median change in density at 1 year among cases was −1.0% [interquartile range (IR), −3.2% to 0.8%] compared with −0.8% (IR, −2.8% to 0.4%) among their matched controls. Moreover, 82% of the cases and 88% of their controls remained within 5% of their baseline MBD, as shown by the box in Fig. 1. Figure 2 shows the change in MBD by baseline mammographic density.

The primary endpoint was the difference in the change in MBD over 1 year between the case and her matched control. A difference value of 0 indicates that the case and her control had an identical change in breast density, whereas a negative value indicates that the case had a larger decrease in density than her control. The median difference for all 369 pairs was −0.1% (with 10th percentile of −5.9% and 90th percentile of 5.2%). Thus, there was no evidence to conclude that the change in MBD over 1 year differs between a case and her matched control whether considering all matched pairs (n = 369; P = 0.51); the subset of matched pairs where both the case and her matched control had a baseline MBD greater than 10% (n = 222; P = 0.55); those cases on MA27 randomized to either aromatase inhibitor therapy (n = 178; P = 0.41; n = 164; P = 0.56) or among those concordant on prior hormone therapy status (n = 240; P = 0.85). Among the 90 matched pairs where the case had taken aromatase inhibitors for at least 12 months, though, there was some suggestion of a greater difference between case–control pairs [median difference for pairs = −0.90% (95% CI, −1.62 to −0.10) P = 0.08], but these results only approached statistical significance.

Within this large study of aromatase inhibitors and breast density, we found that 14% of cases initiating aromatase inhibitor therapy had a MBD reduction of at least 5% at an average of 10 months of therapy. Cases that experienced a reduction of 5% or more in MBD with aromatase inhibitor therapy were more likely to have a higher baseline MBD, to have used aromatase inhibitor therapy for more than 12 months, and used postmenopausal hormones before aromatase inhibitor therapy. However, when comparing the cases on aromatase inhibitor therapy with matched controls of the same age, BMI, and baseline MBD, there was no evidence of an association of change in breast density with aromatase inhibitor therapy use. Results were similar for exemestane and anastrozole, when restricting to pairs who were concordant on postmenopausal hormone status, and those with at least 10% MBD at baseline.

Our investigations of factors associated with MBD reductions among the aromatase inhibitor cases identified higher baseline MBD, longer duration of endocrine treatment, and prior hormone therapy use. Kim and colleagues (14) also found higher baseline MBD and longer duration associated with MBD reductions, but they did not evaluate prior hormone therapy use. In previous work, we examined factors that predict changes in MBD among healthy women, and found prior hormone therapy use and higher baseline density were associated with greater reductions in MBD in women of mammogram age (38).

Our finding of no overall association between aromatase inhibitor therapy and MBD is consistent with 5 of the previous studies that have examined the influence of aromatase inhibitors on breast density among postmenopausal women (Table 1). These previous studies were composed of breast cancer cases on 5 years tamoxifen, randomized to letrozole or placebo (22); cases randomized to letrozole, anastrozole, or placebo as adjuvant therapy (24, 26); healthy women with some visible density randomized to exemestane or placebo (25); and healthy women on postmenopausal hormone therapy who took letrozole (23, 27). Surprisingly, results did not vary for the studies of aromatase inhibitor therapy that required visible density (25) or at least 25% density (24) on the baseline mammogram, which is consistent with our findings of no difference between cases and matched controls who had at least 10% baseline density.

Four studies did find significant reductions in women on aromatase inhibitor therapy. These included 40 postmenopausal women on combination hormone therapy who experienced a statistically significant reduction in percentage density when taking letrozole for a median of 24 months (n = 18) compared with women who did not take letrozole over this same time period (n = 22; 6.8% vs. 1.4% reduction, respectively; ref. 27) and studies comprised of small numbers of breast cancer cases (26) and high-risk women (28), which examined changes in MBD at 6 and 12 months of therapy and found evidence of reduction in MBD at 12 months (Table 1). However, for the latter studies, there was no comparison group evaluated, so these changes are difficult to interpret as MBD is expected to decrease with age (38), as seen in the controls in the current study. The recent study by Kim and colleagues (14) also examined change in MBD at an average 13 months for 175 women on aromatase inhibitor therapy and 890 women on tamoxifen therapy, finding an overall average reduction of 5.9% (range, −17.2% to 36.9%) with either therapy and a smaller mean reduction (−3.1% ± 6.3%) at 1 year with aromatase inhibitor therapy compared with tamoxifen (−6.5% ± 7.1%). Also, women on aromatase inhibitor therapy who experienced a reduction in MBD of less than 5% over the first 13 months of therapy had a 7 times greater risk of recurrence at 68.8 months of follow-up than women who had a 5% or greater reduction (14), although this was not statistically significant (95% CI, 0.90–56.37) and the CIs were wide.

About the studies noted earlier, there were some inconsistencies in eligibility, populations, and analyses (Table 1). The study by Kim and colleagues (14) was conducted in a Korean population, and some Asian populations have been shown to have smaller or similar amount of dense area but higher percentage density than Caucasian women because of their smaller breast size (39, 40). Also in the study by Kim and colleagues, 16% of women had a baseline MBD more than 50% and 76% of women more than 25% MBD; on the other hand, only 2% had a baseline MBD of 10% or less. Thus, these women who were younger (median of 49 years compared with 63 years in our study) and started with much higher baseline MBD compared with our postmenopausal population that had 29% to 37% of women under 10% baseline MBD. This could have resulted in differing ability to detect changes, as 34% of the Korean cases on aromatase inhibitor experienced a MBD decrease of 5% or greater compared with 14% in our population. Another major difference in the Kim and colleagues study was the high proportion of patients who received adjuvant chemotherapy (77%), which might influence MBD reduction through ablative effects on ovarian function in the premenopausal women as the menopausal transition is consistently associated with decreases in density (38, 41). Our study population was postmenopausal at the time of baseline mammogram and the small proportion of chemotherapy use (31%) in cases after the baseline mammogram would not have affected their ovarian function. Other differences in the Kim and colleagues' (vs. current) study included the use of digital instead of film mammography to assess changes in density and the inability to account for BMI in their analyses.

An association of greater reductions in MBD with longer time on aromatase inhibitor therapy was not seen consistently across studies. We and others (14) showed duration of therapy was a factor associated with MBD reduction among the cases. Also, results from our case–control comparisons among the 90 pairs who had used aromatase inhibitors at least 12 months showed evidence of a greater decrease among cases than controls, although this was not statistically significant (P = 0.08). The studies by Prowell and Smith, which examined changes at 6 and 12 months of therapy in the same women (n = 50 and 16, respectively), also found greater reduction with the longer duration of therapy. However, 2 of 3 studies that examined the longest duration of aromatase inhibitor therapy (24 months) found no association between aromatase inhibitor use and density change (refs. 24, 25; Table 1). Time on therapy has been shown to be relevant in studies of tamoxifen and MBD, where reductions in density were seen with 12 to 18 months of tamoxifen therapy versus placebo but were even stronger with 54 months of therapy (10).

Even if there is a subgroup of women who experiences changes in density with aromatase inhibitor use that potentially translates to reduced risk or recurrence, it will be difficult to identify these women in the clinical setting. Postmenopausal women for whom aromatase inhibitors are currently indicated generally have lower MBD, and reductions with aromatase inhibitor therapy are often small (<5%) and within the noise of the mammographic density measure. In fact, in our study, there were only 2% of women on aromatase inhibitors who experienced reductions in density of at least 10% (Table 2). Even in the study by Smith and colleagues (28), which saw decreases among 11 of 16 patients on letrozole at 12 months (Table 1), only 2 of 16 patients experienced a decrease of at least 10% at 12 months (28). However, other imaging modalities, such as MRI, may be able to better characterize small changes in breast density with aromatase inhibitor therapy use, as has been seen with studies of tamoxifen (56). Even with the ability to detect changes, it is not clear how a clinically meaningful reduction in density will be defined, as the majority of women experience a decline in density as they age, with the greatest changes during perimenopause (38, 42).

The lack of association in our study between aromatase inhibitor therapy and density change was surprising in light of the numerous studies that have shown mammographic density reductions resulting from tamoxifen therapy in Caucasians (9, 10) and the recent study of reductions in MBD with aromatase inhibitor use among younger Korean women (14). Cuzick and colleagues (10) in the largest study of tamoxifen and density to date (818 women total) showed a decrease in breast density with tamoxifen use compared with placebo among both premenopausal and postmenopausal women, but the reduction was much greater among those under 45 years compared with those over 55 years [mean change of −13.4% (95% CI, −8.6% to −18.1%) vs. −1.1% (95% CI, −3.0% to 5.1%), respectively]. This is similar to the magnitude of change in density at 1 year among cases on aromatase inhibitor therapy over age of 55 years in the current study [mean change of −1.04% (95% CI, −1.4%, −0.63%)]. Importantly, the Cuzick and colleagues' study required a baseline MBD of at least 10% to allow the possibility of a density reduction, and 98% of the population in Kim and colleagues (14) also had MBD more than 10%, which may have been responsible for detection of the relatively small change among the postmenopausal aromatase inhibitor users. However, even in our examination of the 90 case and control pairs with increased baseline densities (more than 10%), we did not find evidence of a statistically significant association between aromatase inhibitors and density reduction. The study by Cuzick and colleagues (noted earlier) required 12 to 18 months of therapy and Kim and colleagues had an average of 13 months of therapy. Our findings that women who had taken at least 12 months of aromatase inhibitor therapy had a difference that approached statistical significance raises the possibility that longer follow-up of patients would have revealed a difference. However, the difference even in these patients was small and, for MBD to be a useful predictive biomarker of aromatase inhibitor effect, changes would need to be seen in a reasonably short period of time.

To date, then, studies would indicate that tamoxifen is associated with reductions in MBD, even among postmenopausal women, whereas aromatase inhibitors show either a lesser or no effect on MBD. However, this differential influence of endocrine therapies on MBD does not directly translate to therapeutic efficacy. In fact, aromatase inhibitors have been found to be better than tamoxifen in reducing recurrence (15, 43) and contralateral breast cancer (44) in postmenopausal women, despite the lack of their effect on MBD. Importantly, we do not understand the molecular mechanism underlying the positive association of MBD and breast cancer risk, nor how tamoxifen influences MBD. Tamoxifen competitively binds to ERs on breast tumors and other tissue targets, producing a nuclear complex that decreases DNA synthesis and inhibits estrogen effects. Tamoxifen function can be regulated by a number of different variables including growth factors. Aromatase inhibitors, however, have only one function, which is the blockade of the aromatase enzyme, and thus the conversion of androstenedione to estrone and testosterone to estradiol. Furthermore, tamoxifen and aromatase inhibitors have different effects on systemic levels of hormones and gene expression at the tissue level (20, 26, 45–48). Difference in mechanisms of action between aromatase inhibitors (decrease ligand, i.e., estrogen) and tamoxifen (blocks the ER), therefore, may be an explanation for the difference in impact on MBD.

Thus, decreases in MBD with tamoxifen, but to a lesser extent with aromatase inhibitors, would suggest that MBD is influenced by regulation of the ER, but not necessarily the concentration of estrogens in the breast tissue. In fact, studies of postmenopausal patients with breast cancer on aromatase inhibitors (anastrozole and letrozole) showed suppression of estrone, estradiol, and estrone-sulfate levels from pretreatment levels (20, 26, 48). However, tamoxifen did not result in similar suppression of estrogens, and even showed an increase in estrone-sulfate (48). We would have expected a greater effect of aromatase inhibitors than tamoxifen on MBD if mediated solely through estrogens. We hypothesize, then, that reduction in estrogens alone is sufficient to reduce breast cancer risk in postmenopausal women but is not sufficient to impact MBD. This is consistent with the mixed findings from studies that have examined blood estrogen levels with MBD (49–52) and data from Prowell and colleagues that showed suppression of estrone-sulfate with 1 year of aromatase inhibitor therapy was uncorrelated with changes in MBD over the same time period (26). On the other hand, studies of ER-α expression in the breast epithelial tissue and MBD (53–55), which would support a mechanism by which tamoxifen reduces MBD, have shown no evidence of an association. Thus, at present, the mechanistic basis for differential effects of tamoxifen and aromatase inhibitors on density is unknown and is the subject of future research. Our current efforts to identify genes involved in involution and breast density may shed light on the response of MBD to these hormonally related therapies.

There were several strengths to our study, including the large sample size, knowledge of actual start date of aromatase inhibitor therapy for the trials, ability to examine 2 types of aromatase inhibitor therapy, validated measure of breast density, and the closely matched control group. We also recognize limitations of the observational design, our inability to match on prior postmenopausal hormone therapy due to the prioritization of the other matching variables and the analysis of mammograms taken within 1 year before the study instead of at the time of randomization. However, our closely matched control group will help minimize bias in comparisons of aromatase inhibitor therapy versus controls; the distribution of prior hormone use was equal in the cases and controls and separate analyses within the pairs concordant on prior hormone therapy were similar to the overall results; and finally, analyses showed that interval between mammograms was not associated with having a change in breast density.

In summary, we present results from the largest study of aromatase inhibitors in postmenopausal White women examining the influence of aromatase inhibitors on breast density. We found several factors associated with reduction of density among cases on aromatase inhibitor, including higher baseline density, prior hormone therapy use, and longer time on therapy. However, we found no evidence of an association of change in breast density with aromatase inhibitor therapy use among the cases compared with controls closely matched on baseline breast density, age, baseline BMI, and interval between mammograms. Our findings may be a consequence of the lack of influence of the drug on the overall dense tissue and/or the limitations of mammography in detecting small changes that may occur among postmenopausal White women with already low breast densities. These findings indicate that MBD is not likely to be a clinically useful biomarker, at least in the short term necessary for clinical use, for predicting the value of aromatase inhibitor adjuvant therapy in White women.

P.E. Goss has an expert testimony from Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C.M. Vachon, V.J. Suman, A.U. Buzdar, J.E. Olson, R.M. Weinshilboum, P.E. Goss, J.N. Ingle

Development of methodology: C.M. Vachon, V.J. Suman, A.U. Buzdar, C.R. Elliott, L. Shepherd

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.M. Vachon, A.U. Buzdar, G. Ursin, C.R. Elliott, L. Shepherd, J.N. Ingle

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.M. Vachon, V.J. Suman, K.R. Brandt, M.L. Kosel, P.E. Goss, J.N. Ingle

Writing, review, and/or revision of the manuscript: C.M. Vachon, V.J. Suman, M.L. Kosel, A.U. Buzdar, L.M. Flickinger, G. Ursin, L. Shepherd, R.M. Weinshilboum, P.E. Goss, J.N. Ingle

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.M. Vachon, F.-F. Wu, L.M. Flickinger, C.R. Elliott

Study supervision: C.M. Vachon, A.U. Buzdar, L.M. Flickinger, C.R. Elliott

The authors thank the women on trials MA27, N063I, and MC0532, who consented to this research.

This study was supported, in part, by grants from the NIH, Specialized Program of Research Excellence (SPORE) in Breast Cancer, P50 CA116201, R01 CA97396, R01 CA140286, and Mayo Pharmacogenomics Research Network (PGRN), U01 GM61388. P.E. Goss is supported by the Avon Foundation, New York.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.