Diet across the Lifespan and the Association with Breast Density in Adulthood

Studies have shown inconsistent results regarding the association between dietary factors across the lifespan and breast density and breast cancer in women. Breast density is a strong risk factor for breast cancer, and the mechanism through which it influences cancer risk remains unclear. Breast density has been shown to be modifiable, potentially through dietary modifications. The goal of this paper is to summarize the current studies on diet and diet-related factors across all ages, determine which dietary factors show the strongest association with breast density, the most critical age of exposure, and identify future directions. We identified 28 studies, many of which are cross-sectional, and found that the strongest associations are among vitamin D, calcium, dietary fat, and alcohol in premenopausal women. Longitudinal studies with repeated dietary measures as well as the examination of overall diet over time are needed to confirm these findings.


Introduction
Breast cancer (BC) is the most commonly diagnosed cancer and the second leading cause of cancer death among women [1]. Alcohol consumption, physical activity, elevated aer menopausal body mass index (BMI) [2], age at menarche and menopause [3], and family history and genetic mutations [4] are a few of the well-established BC risk factors. In addition, breast density (BD), or the amount of dense �broglandular tissue present in the breast, has been related to BC risk; women who have breast densities of 75% or more have up to a 4-5-fold increase in BC risk [5]. Consequently, BD is oen thought of as an intermediate on the BC development continuum that can be measured, assessed, and targeted for potential cancer prevention strategies [5][6][7][8]. Even so, little is known about the mechanism through which BD may affect breast cancer risk [9]. Breast tissue develops mostly during puberty and continues to undergo changes throughout several life stage events, such as pregnancy [3,10,11]. is paper will examine research on diet and diet-related factors captured across the lifespan and the association with adult BD.

Methods
A literature search of the PubMed database of the United States National Library of Medicine was conducted to �nd human studies that evaluated the associations between BD measures and diet isn the form of either single nutrients or whole dietary patterns. Both observational and diet intervention studies conducted at any stage of the lifespan were considered. Observational studies were included if they had recorded individual's dietary intake of foods or energy with dietary assessment tools such as a dietary recall (DR), food frequency questionnaire (FFQ), food record (FR), or other relevant assessment tools. Relevant studies were identi�ed using the following search terms in multiple combinations: "adolescent diet and breast density, " "diet and breast density, " "childhood and breast density, " "diet and parenchymal patterns" and "mammographic breast density and diet. " e search was limited to full-text publications written in English. As illustrated in Figure 1, a total of 77 studies were identi�ed. Aer all exclusions, 28 studies were included in this paper.

Measurement of Breast
Density. BD can be measured two (2D) and three dimensionally (3D), with the most common being through 2D mammography. Mammography measures the area of dense tissue (ADT) and the total area of the breast. Percent dense area (PDA) is oen reported and is estimated as the proportion of dense �broglandular tissue area to total breast area [9]. Area of nondense tissue (ANDT), which is primarily adipose tissue, also can be estimated. Magnetic resonance imaging (MRI) and ultrasound also are used to measure BD. ese 3D modalities measure volume of dense tissue (VDT) and percent dense volume (PDV). Percent densities measured by mammography and MRI are highly correlated in the general population and among women who have low breast densities ( ) [38], but this correlation is attenuated among women with higher mammographic density greater than 50 percent ( ) [38]. In addition to quantitative measures of PDA and ADT, semiquantitative and qualitative measures are frequently reported. Either the Wolfe classi�cation, which has been further classi�ed into Tabár, or the breast imaging-reporting and data system (BI-RADS) classi�cation is used [39][40][41]. ese measures oen classify the breast on a four-to �ve-level scale ranging from low-to-high levels of �broglandular tissue. While all methods are able to assess BD, quantitative methods provide more consistent results and a larger gradient of risk. Qualitative measures oen have intervals that are too large (fewer categories) and do not capture true risk gradients [42].

�on�ietary Factors �at �n�uence Breast Density.
In general, PDA is higher in premenopausal women compared to postmenopausal women as well as in postmenopausal women who use hormone replacement therapy (HRT), and in both pre-and postmenopausal women with a lower BMI; and is lower in women who are parous, experience their �rst birth at a younger age, or are smokers [9,43,44]. Correlates of ADT are less well-studied, but in one study the ADT was inversely associated with age and BMI [44]. e nondense compartment of the breast is adipose tissue, and higher adiposity, frequently measured by BMI, attenuates the ratio of dense tissue area to total breast area. Other characteristics associated with BD may in�uence estrogen, insulin-like growth factor-I (IGF-I), or insulin-like growth factor binding proteins (IGBPBs) that affect �broglandular tissue proliferation [15,45]. Alternatively, some characteristics, such as parity, could have direct effects on breast morphology which are re�ected in PDA.

Dietary Factors
. is review will focus primarily on diet and diet-related factors and their potential effects on both PDA and ADT with only limited attention to endogenous risk factors that are well-studied and not modi�able. A summary of these �ndings can be found in Tables 10 and  11. Observational studies and clinical trials that evaluated dietary intakes during childhood, adolescence and adulthood are described.

Childhood Diet and Adult Breast Density
Much of breast development occurs during puberty; thus, factors such as childhood diet that in�uence the timing of puberty could potentially affect BD [46,47]. ree studies have examined dietary habits during childhood and the effect on BD in adulthood. Mishra and colleagues [13,14] conducted two studies in a nationally representative longitudinal British sample to examine the association of childhood diet with BD. Childhood diet was assessed at age four years by a single dietary recall completed by mothers and later linked to mammographic BD measures collected at approximately �y years of age from pre-, peri-and postmenopausal women. Aer controlling for relevant confounders, the investigators observed no association between PDA and childhood calcium [14], or total energy intake or with three dietary patterns ((1) breads and fats, (2) fried potatoes and �sh, and (3) milk, International Journal of Breast Cancer 3 fruit and biscuits). A limitation of these studies is that a single dietary recall was used to assess diet, which could have contributed to the null results since multiple recalls are typically required to adequately assess usual diet [48]. Additional time points for dietary data collection, such as during adolescence, may have provided more insight into the effect of early diet on BD.
Haars and colleagues [12] examined the association between short-term transient caloric restriction (i.e., 6-8 mos.) during the Dutch Famine (when women were aged 2-33 years) and adult BD in e Netherlands DOM-project. �hile this study does not necessarily �t within our inclusion criteria, it is included in this paper because of the limited data available on children. Levels of caloric restriction were retrospectively assessed through three questions regarding hunger, cold, and weight loss and categorized as absent, moderate, or severe famine exposure (FE). Degree of famine exposure at 2-9 years of age was signi�cantly inversely associated with ANDT; mean ANDT were 77.8 cm 2 , 87.7 cm 2 and 53.1 cm 2 in unexposed, moderately, and severely exposed, respectively ( trend = 0.03). Although not signi�cant, the women who were severely energy restricted at this age also had a larger ADT and higher PDA. However, because only 15 subjects were severely restricted, results should be interpreted cautiously.
e three studies that examined childhood diet and its effect on adult BD measures did not �nd associations with PDA or ADT although, in the study of the Dutch famine, severe caloric restriction early in life was signi�cantly inversely associated with ANDT later in life [12]. In this cohort, women who were severely calorically restricted had higher levels of both IGF-1 and IGFBP-3 postmenopausally than those who were not restricted [49]. us, one mechanism through which caloric restriction at young ages could potentially in�uence adult BD may be via differential programming of the somatotrophic axis resulting in long-term effects on growth factors such as IGF-1 and IGFBP-3 that are associated with breast density [50]. However, the small sample size and indirect diet assessment limit the inferences that can be drawn from this study. Taken together, the limited data available do not provide strong support for a role of childhood diet in determining breast density, but additional large prospective studies are needed before �rm conclusions can be made.

Adolescent Diet and Adult Breast Density
Most of breast development occurs during puberty, and diet during this time could have long-term effects on BD in adulthood. One of �ve studies we found a signi�cant association between diet during adolescence and BD in adulthood [16]. In the study by Tseng et al. [16], higher red meat intakes between the ages of 12-17 years were signi�cantly associated with increased adult PDA in 201 Chinese-American female immigrants. Aer adjusting for degree of acculturation and other relevant covariates, women with the highest red meat consumption were at 3 times the odds of being in the highest PDA category compared to those with the lowest red meat consumption. �hen strati�ed by menopausal status, red meat intake remained signi�cantly positively associated in postmenopausal, but not premenopausal women.
e remaining four studies, including 3 observational studies and one clinical trial, found no associations between dietary components or alcohol consumption during adolescence and BD in adulthood [12,15,17,18]. Two studies used data from the large Minnesota Breast Cancer Family Study Cohort (MBCFSC) to examine the role of adolescent diet and alcohol consumption on BD in pre-and postmenopausal women. Diet for girls at ages 12-13 years was collected retrospectively 50 years later via a 29-item FFQ focusing on highfat foods (e.g., meats and other animal fat sources, snacks, and desserts). Intakes of fruits, vegetables, �sh, and chicken were also analyzed. In the �rst study, Sellers et al. [15] observed no signi�cant associations between any of these food groups and BD in multivariate analyses strati�ed by menopausal status. In the second analysis, Vachon et al. [17] evaluated alcohol consumption prior to age 18 via a self-reported questionnaire collected when the majority of the women were in their sixties. "Never drinkers" had lower mean PDA than "ever drinkers" (22.2 ± 14.3% versus 26.5% ± 15.9%); however, these results were attenuated and not signi�cant aer adjustment for age, BMI, H�T use, age at �rst birth, and parity [17]. In the study by Haars et al. [12] described above, short-term caloric restriction in girls age 10-18 years was not associated with adult BD measures. In a clinical trial, Dorgan et al. [18] examined the long-term effects of a dietary intervention to lower fat and increase �ber intake during childhood and adolescence (the Dietary Intervention Study in Children-DISC) and observed no differences in the VDT or PDV between those participants who received the behavioral intervention and the control group [18]. us, similar to childhood diet, the limited data available do not provide much support for a role of adolescent diet in determining adult BD, but additional research is needed.

Adult Diet and Adult Breast Density
e majority of studies that have evaluated associations of diet with BD assessed the effects of adult diet. A total of 26 epidemiological studies and randomized controlled trials that examined dietary intake and BD among adult women are included in this paper.

Total
Energy. ree studies examined the association of total energy intake in adulthood with BD measures (Table 9). In a nationally representative British cohort total energy intake around age 36 years was signi�cantly positively associated with PDA and ADT at age of 51 years in pre-and postmenopausal women [13]. Sala et al. [29] similarly found that total energy intake was signi�cantly positively associated with PDA. e odds ratio (O�) for being classi�ed in the highest PDA category for women in the highest versus lowest tertile of energy intake was 1.79 (95% CI: 1.09-2.91). In analysis strati�ed by menopausal status, energy intake was associated with signi�cantly higher PDA in postmenopausal women only [29]. Finally, in the Dutch famine study described above, caloric restriction in adulthood was not associated 4 International Journal of Breast Cancer with several BD measures suggesting that exposure to shortterm caloric restriction may be more important in children.

Dietary Fat.
Eight studies [24,25,27,[29][30][31][32]51] have examined the association between dietary fat and BD in adulthood. ree studies showed a signi�cant positive association with total fat and BD measures. Nagata and colleagues [31] showed signi�cant associations in a Japanese sample with mean PDA being 15.5% in the highest quartile of total fat intake compared to 9.9% in the lowest quartile ( trend = 0.04). In a sample of 31 BC patients, women in the highest quartile of total fat (mean % energy ( ) = 42.04) compared to the lowest quartile of intake (mean % = 34.72) were signi�cantly more likely to be classi�ed as a P2 + D� (high density) pattern compared to the N1 + D1 (low density) pattern ( 0.0 ) [25]. Qureshi et al. [32] showed a positive trend for the relationship between total fat with increased ADT in a large Norwegian population of postmenopausal women although it did not reach statistical signi�cance.
Individual fatty acids have also been examined with saturated fatty acids (SFAs) generally being positively associated with increased BD measures. In an analysis based on 645 pre-and postmenopausal women ages of 40-62 years enrolled in the Canadian National Breast Screening Study (CNBSS), SFA intake was signi�cantly positively associated with PDA. Mean PDA was 44.2% in the highest quartile of SFA intake compared to 38.6% in the lowest ( trend = 0.009) [30]; however, menopausal status was not controlled for or strati�ed by in this analysis. Similar �ndings also were reported in pre-and postmenopausal Japanese women; mean PDA was 16.5% in the highest quartile of SFA intake compared to 7.3% in the lowest ( trend = 0.02) [31]. Qureshi and colleagues [32] also showed a positive trend with SFA and PDA in a Norwegian population of postmenopausal women although statistical signi�cance was not reached. Nordevang and colleagues [25] observed that women who consumed a mean % of 19.27 from SFA in the highest quartile were more likely to be classi�ed as having a high-risk PDA compared to those who consumed a mean % of 15.42 from SFA in the lowest quartile ( 0.0 ). In contrast, a signi�cant inverse association was observed with SFA in a subset of 283 premenopausal women from the MBCFSC; mean PDA was 37% in those with the highest SFA intake compared to 44% in the lowest consumers aer controlling for relevant confounders ( trend = 0.03) [51]. No associations with dietary fat were observed in postmenopausal women alone in this study.
e essential PUFA, linolenic acid, was inversely associated with PDA in a mediterranean population of both preand postmenopausal women. Women in the highest tertile of intake had 31% lower odds of being classi�ed as high PDA [24]. Elevated PUFA consumption in a sample of BC patients (mean % = 5.65 versus 4.70) and n-6 fatty acids (Mean % = 4.69 versus 3.81) was also signi�cantly associated with being classi�ed as a P2 or D� (high density) versus an N1 or P1 (low density) Wolfe parenchymal pattern ( 0.0 ). Vachon et al. [51] examined a sample of both pre-and postmenopausal women in the MBCFSC and observed women in the highest quartile of PUFA intake had 4% higher PDA compared to those in the lowest quartile ( = 0.0 ). Similar results were observed with the PUFA : SFA ratio in this study.
Finally, Nordevang and colleagues [25] found that women within the highest quartile of MUFA (mean % = 14.22) were more likely to have a high PDA compared to those in the lowest quartile (mean % = 11.98, 0.0 ). Far fewer associations between dietary factors and BD measures were observed in postmenopausal women, with only increased consumption of MUFAs being signi�cantly associated with high PDA even though the difference in MUFAs as percent energy between the high and low density groups was small (mean % = 12.9 versus 12.3, 0.0 ). A small number of randomized controlled trials (RCTs) have also been conducted to examine dietary fat and BD and have yielded mixed results [35][36][37]. ese studies will be further discussed in the "RCT" section of this paper. [52] concluded that there was strong evidence for a positive association between alcohol and BD in both pre-and postmenopausal women [52]. Alcohol may in�uence BD through decreasing the concentration of sex-hormone binding globulin and disturbing estrogen metabolism, increasing serum estrogen metabolites, raising oxidative stress in tissue, and leading to an increase in breast tissue proliferation [53]. e relationship between alcohol and BD may also be related to its positive association with IGF-1 and a negative association with IGFBP-1 that has been shown in post-, but not premenopausal, women [54]. Total alcohol consumption in a multiethnic cohort was associated with a 1-2% higher PDA among pre-and postmenopausal alcohol consumers (median alcohol consumption in the highest consumers = 12 drinks/wk) when compared to abstainers; however, this association failed to reach statistical signi�cance [28]. In a mediterranean cohort of both pre-and postmenopausal women, both total wine consumption and total alcohol consumption were signi�cantly positively associated with a 31% and 42% higher odds of having an elevated PDA, respectively [24]. A similar observation was made with total alcohol consumption in premenopausal women with "Never Drinkers" having a mean PDA of 39% compared 45% for consumers of 3.9 g/d and 42% for consumers of >3.9 g/d ( trend = 0.08). When the type of alcohol was examined, comparable results were observed with white wine in postmenopausal women only; however, an inverse association was observed with red wine in postmenopausal women with "nondrinkers" having a mean PDA of 34% compared to 32% for those consuming 1 serving/wk and 28% for those consuming ≥2-4 svg/wk ( trend = 0.02) [51]. e authors suggest that the difference between white and red wine may be due to the polyphenols that are present in red wine, which have been shown to have chemoprotective effects [51]. Tseng et al. [27] and Sala et al. [29] also looked at alcohol intake in pre-and postmenopausal women and found no associations with BD measures. iso�avones and their association with PDA and concluded that soy products have a little-to-no in�uence on BD measures regardless of the amount of iso�avones they are consuming in the range 0.1-120 mg/d [55]. A meta-analysis of several RCTs that examined that the effect of soy and BD measures was also conducted and will be discussed in the "RCT" section of this paper.

Calcium and Vitamin D.
Vitamin D and calcium have been linked to cellular growth and differentiation in breast tissue [56,57] and may in�uence the amount of dense tissue in the breast. Four cross-sectional studies found a signi�cant inverse association between vitamin D and calcium intake, alone or in combination, with BD measures [21][22][23]25] in premenopausal women. Nordevang et al. [25] found that lower intakes of calcium (1165 versus 1433 mg/10MJ) were signi�cantly associated with an increased PDA. When examining dietary vitamin D and calcium, Bérubé et al. [22] observed that premenopausal women in the highest categories of both vitamin D (≥100 IU/d) and calcium (≥750 mg/d) intake had 72% lower odds of having high PDA. When intake from both diet and supplements was considered, simultaneous increases of 400 IU of vitamin D/d and 1000 mg of calcium/d were associated with an 8.5% (95% CI: 1.8-15.1%) decrease in PDA in premenopausal women [21]. e association in postmenopausal women was considerably weaker [22] or null [21]. Diorio et al. [23] found comparable results; as dietary vitamin D and calcium increased by 100 IU/d and 250 mg/d, respectively, PDA decreased by 1.8% ( ). Similar results were found when intake from food and supplements were analyzed together.
Out of the remaining seven studies, two included only postmenopausal women and neither found an association between vitamin D and calcium intake and BD [19,20]. An additional four studies reported signi�cant associations between vitamin D and calcium overall; however, the results in postmenopausal women were considerably weaker than observed in premenopausal women [14,26,27]. Masala et al. [24] observed that Mediterranean women with a higher calcium intake had 33% lower odds of having a high-risk mammographic pattern. No association was observed with vitamin D; however, vitamin D intake in this population was very low [24]. In a nationally representative British cohort, an inverse association between calcium intake and PDA, which were both measured among women in their 50's, was observed. Calcium intakes ≥1180 mg/d compared to 699 mg/d resulted in a 0.53 (95% CI: 0.03-1.02) standard deviation decrease in PDA [14]. No additional associations were observed with the ADT or ADNT in this study. Tseng and colleagues [27] conducted a cross-sectional analysis using a 126-item FFQ to examine several dietary factors including vitamin D and found that, aer controlling for menopausal status, high-risk women (women with at least one 1st or 2nd degree relative with breast or ovarian cancer) with higher vitamin D intake had 50% lower odds of having high PDA when comparing the highest to the lowest tertile. Finally, serum 25[OH]D and dietary calcium intake obtained from an FFQ in a sample of women from the MBCFSC (73% postmenopausal) were not associated with either PDA or ADT [26]. While the overall trend failed to reach signi�cance, the study did demonstrate that women with the highest mean intake of both calcium (>1,385 mg) and 25(OH)D (>86.2 nmol/L) had the lowest PDA and ADT aer adjusting for age, BMI, parity, age at �rst birth, and physical activity. Vachon et al. [51] also reported no associations for calcium and vitamin D from both dietary and supplemental sources with PDA in this cohort.
Overall, this research suggests that vitamin D and calcium are inversely associated with BD in premenopausal women. It is critical to note that as calcium and vitamin D increased from 500 mg/d and 100 IU/d to >1,750 mg/d and >700 IU/d, respectively, PDA decreased in a doseresponse fashion with clinically relevant decreases in PDA between 8 and 12% among premenopausal women [21,23]. is is comparable to the effect of selective estrogen receptor modulators such as tamoxifen [58]. Importantly, Brisson et al. [59] examined serum vitamin D [25(OH)D] levels and found that PDA was lowest in the fall (39%) and highest in the spring (45%) ( ), which was consistent with the rise and fall in serum vitamin D across the seasons. Few studies account for season in which BD was assessed. However, it may be important to consider endogenous vitamin D synthesis in response to sunlight in addition to that contributed by food sources. e biologically active form of vitamin D may decrease BD via its antiproliferative properties or tissue-speci�c effects due to breast tissue possessing 1--hydroxylase, which converts inactive 25(OH)D to active 1,25(OH) 2 D [60]. e localized production of 1,25(OH) 2 D helps to regulate cell growth and promote terminal differentiation which promotes cellular resistance from carcinogenic factors [60]. Premenopausal women have higher levels of estrogen, insulin-like growth factor (IGF), and insulin-like growth factor binding proteins (IGFBPs), which may be associated with increased BD [61,62]. Vitamin D, calcium, and IGFBP-3 have been proposed to increase each other's bene�cial antiproliferative and proapoptotic effects [23]; however, vitamin D alone may help to combat the proliferative effects of estrogen and IGF when these hormones and growth factors are available in abundance, such as in premenopausal women.

Carbohydrates, Protein, and
Other. Ten studies have evaluated intakes of carbohydrates, protein, and many other nutrients and their association with BD measures. Eight studies [16, 24, 27, 30-32, 34, 51] used validated FFQs to assess nutrient intake; Sala et al. [29] and Nordevang et al. [25] conducted extensive dietary history interviews. Tseng and colleagues [27] found that, in a sample of 90 women with a sporadic family history of BC, total and animal protein intakes above the median intake had from 3 to 4 times the odds of an increased PDA; these associations were not observed in women with a strong hereditary pattern (1st or 2nd degree relative) of BC [27]. As mentioned previously, red meat intake during adolescence was signi�cantly positively associated with PDA in adulthood; however, there was no association with red meat intake during adulthood in a sample of 201 Chinese-American immigrants [16]. Although few signi�cant associations are observed among postmenopausal women; both Nagata et al. [31] and Sala et al. [29] found signi�cant associations in both Japanese and European populations, respectively, when evaluating carbohydrates and protein. Sala and colleagues [29] found that protein and carbohydrate were positively associated with PDA in all women. When, stratifying by menopausal status, signi�cant positive associations emerged between protein, total meat, and carbohydrates and PDA in postmenopausal women only with those consuming the most having 2.2-2.5 times the odds of having a high-risk PDA. Nagata and colleagues [31] also found that protein was signi�cantly positively associated with PDA with women in the highest quartile of intake having approximately 7% higher PDA than those in the lowest quartile. However, in contrast to the study by Sala, carbohydrates were signi�cantly inversely associated with PDA in 253 postmenopausal Japanese women with those in the highest quartile having 6% lower PDA than the lowest consumers [31]. No associations were observed in premenopausal women [31]. Among pre-and postmenopausal women in the CNBSS, mean PDA was 37.9% in those in the highest quartiles of �ber intake compared to 43.0% in the lowest quartile, and the difference was signi�cant [30]. Comparable results were found in a sample of 31 Swedish premenopausal BC patients; lower consumption of carbohydrate and �ber was associated with higher PDA [25].
In a study evaluating dietary factors and mammographic patterns in a Mediterranean population, both pre-and postmenopausal women in the highest tertiles of the following foods and nutrients had 27-34% lower odds of having a high PDA: total vegetables, cheese, -carotene, vitamin C, and potassium, whereas women in the highest tertile of tomato sauce intake had 34% higher odds of having a high PDA [24]. Similar results with high cheese intake were observed in a sample of 491 premenopausal women in this study [24]. Consistent with these �ndings, total dairy intake was signi�cantly inversely associated with PDA in premenopausal women in the MBCFSC aer controlling for relevant confounders [51]. Among pre-and postmenopausal women in the CNBSS, women in the highest quartiles of carotenoid intake had a 5.4% lower mean PDA when compared to the lowest quartile [30]. Comparable results were found by in sample of 31 Swedish premenopausal BC patients and found that lower consumption of carotene was associated with increased PDA [25].
Only one study to date has examined multivitamin/ multimineral (MVMM) supplement intake and BD outcomes. Bérubé and colleagues [34] found that current premenopausal supplement users had a signi�cantly higher adjusted mean PDA of 45% compared to 42.9% of past or 40.2% of never users ( trend = 0.009). No association was observed in postmenopausal women. Vachon et al. [51] also found that dietary vitamin E and supplemental vitamin C were signi�cantly positively associated with PDA in premenopausal women with the highest consumers having a 4-5% higher PDA than the lowest consumers. Supplemental vitamin B12, on the other hand, was positively related to PDA in postmenopausal women [51].
In conclusion, the foods or nutrients that were shown to be inversely associated with BD may be, in part, tied to IGF/IGFBP levels and oxidative stress reduction. BD has been associated with increased levels of oxidative stress as evidenced by malonoyldialdhyde (MDA) excretion [63] and IGF/IGFBP, particularly in premenopausal women [50,61]. �ower intakes of �ber, carotene, and calcium have also been associated with increased breast densities. Carbohydrate intake has been associated with both lower and higher BD measures women. ese con�icting results may be attributed to the fact that the types of carbohydrate are oen not accounted for and �ber content may in�uence the way that different carbohydrates affect the IGF/IGFBP pathway and oxidative stress. Finally, higher intakes of total dairy and cheese consumption in premenopausal women are associated with lower BD measures, which may be due to the high amounts of calcium and vitamin D in these products.

Dietary Patterns.
Analysis of dietary patterns has recently gained popularity in dietary assessment research, as they capture total diet and are more stable over time than the consumption of single nutrients or foods [64]. Two studies were conducted that examined a posteriori dietary patterns and their association with BD and one study examined the in�uence of Mediterranean Diet (measured by Mediterranean diet scale (MDS)) on BD measures. Dietary patterns were analyzed cross-sectionally in a British cohort and the MBCFSC [13,33]. Aer combining data collected from food records collected at ages 36 and 43 years, four patterns emerged in the British cohort ((1) low-fat and high �ber; (2) alcohol and �sh; (3) high fat and sugar; (4) meat, potatoes, and vegetables). However, none of these patterns was associated with PDA [13]. In the MBCFSC, three dietary patterns emerged from data from a 153-item FFQ ((1) fruit, vegetable, and cereal; (2) salad, sauce, and pasta/grain; (3) meat and starch). Only the fruit, vegetable, and cereal pattern was inversely associated with PDA in premenopausal women; however, it did not reach statistical signi�cance [33].
Smoking has been associated with decreased PDA because of its antiestrogenic effects [65]. When all women included in the sample were strati�ed by smoking status, adherence to the fruit, vegetable, and cereal pattern was signi�cantly inversely associated with PDA in smokers ( = 0.0 ) [33]. e salad, sauce, and pasta/grain pattern was also nonsigni�cantly inversely associated with PDA in smokers [16]. ese patterns are the highest in antioxidant-containing foods, which may bene�t women who are under higher oxidative stress, such as smokers.
Tseng et al. [66] cross-sectionally evaluated the MBCFSC using the MDS. e women were scored based on their consumption of vegetables, legumes, fruits and nuts, cereals, �sh, and the ratio of monounsaturated fatty acids (M�FA) to saturated fatty acids (SFA) as reported on a 153-item FFQ. For each unit increase in the MDS, PDA was decreased by 1.68% ( = 0.000 ) among current smokers but not among nonsmokers aer controlling for relevant confounders including menopausal status [66]. Vegetables, legumes, and cereals were the components of the MDS that had the strongest association with PDA in this population [66]. Overall, it appears that dietary patterns high in antioxidant-containing foods are inversely associated with BD in smokers, who may be experiencing a higher level of oxidative stress than nonsmokers. Other research has shown a positive association between BD and MDA, which is a marker for lipid peroxidation and oxidative stress [63].

Randomized Controlled
Trials. e epidemiological evidence described above suggests that diet is associated with BD measures and that BD has the potential to be modi�ed. As a result, researchers have conducted clinical trials to examine the association between speci�c dietary factors with BD outcomes (Table 8). Boyd et al. [35] �rst examined a low-fat, high-carbohydrate 2-year dietary intervention in 817 women with PDAs ≥50%. ose who were randomized into the intervention group received intensive instruction to consume 15% of calories from fat, 20% from protein, and 65% from carbohydrate while the control group received general dietary advice and instruction to maintain their current intake of fat. Aer two years, the average reduction in PDA was 6.1% and 2.1% in the intervention and control groups, respectively, ( ) [35]. e effect of the intervention remained signi�cant aer controlling for age, weight change, and menopausal status [35]. Aer strati�cation by menopausal status, signi�cant changes in PDA were only observed in women who were either premenopausal throughout the study or who were premenopausal at baseline but transitioned into menopause by the end of the study, with the greatest change in density occurring in the latter group. Consumption of fat and cholesterol was signi�cantly positively associated with change in ADT in this subgroup, whereas protein and cholesterol were signi�cantly positively associated with change in PDA [37]. Martin et al. [36] completed a similar larger clinical trial with longer followup that included 461 women who were premenopausal at entry and postmenopausal aer two years. Several BD measures were assessed (change in breast area, ANDT, ADT, PDA) premenopausally at baseline and later in the postmenopausal phase. Like the previous trial, this trial focused on women with high PDA ≥50% and the intervention group received the same dietary manipulation [35]. is study did not replicate the previous �ndings from Boyd et al. [35]. Aer two years, no change was observed in the intervention group and a slightly lower PDA was observed in the control group; the treatment group difference was not signi�cant [36]. e authors suggest that these unexpected results were likely due to an increase in the ANDT that occurred with weight gain in the sample.
As previously described in the vitamin D and calcium section, a one-year calcium and vitamin D supplementation trial was conducted through the WHI to examine the effects on mammographic PDA in postmenopausal women [20]. Despite the associations observed in observational studies, no change in mammographic PDA was observed with supplementation. e authors suggest that very low PDAs at baseline could have led to a ��oor effect� where further supplementation of vitamin D and calcium had no additional bene�t. Finally, studies that have examined soy and iso�avone consumption and mammographic PDA have also yielded mixed results. Hooper et al. [67] conducted a meta-analysis of eight RCTs including 1287 total women that compared the administration of supplemental iso�avones versus a placebo for at least six months. Results from the meta-analysis showed a modest nonsigni�cant increase in PDA (mean difference: 1.83%; 95% CI 0.25-3.40) in premenopausal, but not postmenopausal, women as iso�avone intake increased; however, there was limited evidence of a clear dose-response relationship over the range of iso�avone intake of 40-120 mg/d.

Conclusions
Data from observational studies suggest that the strongest associations between diet and BD measures are among vitamin D, calcium, dietary fat, and alcohol and are found in adult premenopausal women. However, the few clinical trials that have evaluated these associations have failed to demonstrate a signi�cant change in breast density with various dietary interventions. is could be because the foods/nutrients evaluated truly do not in�uence breast density or could be due to aspects of the study design including duration of the intervention, dose, sample size, or inclusion of predominantly older women in whom breast tissue may be less susceptible to dietary in�uences.
6.1. Limitations. is paper has critically examined 28 studies and has identi�ed strengths and weaknesses as well as highlighting several potential directions for new research to advance the �eld. Many of these studies are cross-sectional in nature and oen focus just on PDA. In addition to this, the majority of women who receive mammograms overall and in these studies are >40 y; an association between dietary factors and BD measures could be undetected if the critical dietary exposure occurred much earlier in life (and was not measured) before breast tissue is fully differentiated and potentially more vulnerable to exogenous in�uences.
e majority of studies included in the paper assessed BD using 2D mammography. Even though estimates of BD obtained by mammography and 3D modalities such as MRI are highly correlated in the general population and in women with less dense breasts [38], correlations are substantially lower in women with more dense breasts in whom density can be more accurately measured using 3D modalities.
Many studies examined the association of diet with PDA but not the ADT. Fewer associations are observed with the ADT compared to PDA; however, results should be reported International Journal of Breast Cancer when available in order to be more comprehensive, improve comparisons across studies, and enhance interpretability in relation to potential physiological mechanisms. Very few studies controlled for the phase of the menstrual cycle at the time of mammography. Because data on variation of breast density over the menstrual cycle are con�icting [68][69][70][71], it seems prudent to consider menstrual cycle day in analyses of breast density when possible. Finally, several methods were used to evaluate BD. Even though many studies used a semiautomated method to reduce variability and error, standardization of assessment would facilitate comparisons across studies.

Future Directions.
To date, most studies of the association of diet with BD have been cross-sectional. Longitudinal studies that measure diet and BD over the life course are needed. Studies that evaluate the in�uence of diet during adolescence, when most breast development occurs, on adult BD could be particularly enlightening. Support for an association of diet with BD from observational studies is stronger for premenopausal women. However, a limited number of short-term clinical trials do not show conclusive evidence that dietary factors in�uence BD. Clinical trials in younger women could be informative and may provide more de�nitive results. Lastly, more research on dietary patterns as they relate to BD are needed.