Adolescent and early adulthood inflammation-associated dietary patterns in relation to premenopausal mammographic density

Mid- and later life adult diet and breast cancer risk have been extensively studied, but few associations have emerged [1]. Prior research supports that exposures in the years prior to the first birth, including diet, may be more influential in terms of breast cancer risk than exposures occurring later in life [2, 3]. During adolescence, the mammary glands undergo rapid proliferation, while the terminal structures of the mammary glands differentiate only after the first birth. Thus, this period has been identified as a critical window for establishing future breast cancer risk [3]. In prior analyses in the Nurses’ Health Study II (NHSII), adolescent and early adulthood dietary patterns have been associated with breast cancer risk. For example, a pro-inflammatory dietary pattern, associated with markers of inflammation (e.g., C-reactive protein), was associated with a 35% increased risk of premenopausal breast cancer among women in the highest quintile of a pro-inflammatory adolescent dietary pattern [4], while adolescent dietary patterns that captured an overall healthy diet (e.g., the Alternative Healthy Eating Index) were inversely associated with breast cancer risk [2]. Systemic inflammation, particularly when low-grade and chronic, is likely to create an imbalance in pro- to anti-inflammatory markers, causing the former to be overexpressed and leading to increased cellular proliferation, as well as genomic instability and cellular damage [5]. This overexpression of certain pro-inflammatory markers (e.g., interleukin-6, C-reactive protein, and tumor necrosis factor α) may be associated with an increased risk in tumor growth and proliferation in the breasts [6, 7], with one possibility being an increase in estrogen production leading to increased breast density [8]. Women may be particularly susceptible to these potential impacts of inflammation during adolescence and early adulthood when the mammary glands are undergoing rapid change and proliferation. Mammographic density has consistently been identified as a strong, independent risk factor for breast cancer [9, 10]. More recently, it has been examined as a potential intermediary of associations between established breast cancer risk factors (e.g., childhood and adolescent somatotype, biopsy-confirmed benign breast disease, age at menarche, family history of breast cancer) and breast cancer, with results from a case-control study nested within the NHS/NHSII cohorts showing a significant mediation of mammographic density on the relationship between early life body size and breast cancer risk in premenopausal women [10]. Most prior epidemiologic studies have examined the impact of current adult diet on mammographic density by focusing on specific food items or groups (e.g., milk, alcohol) [11‐13]; however, few studies have examined the association between dietary patterns and mammographic density [14, 15]. In the Minnesota Breast Cancer Family Study, a prospective cohort of family members (sisters, daughters, nieces, and grand-daughters) that began in 1990 [16], Tseng et al. reported a significant inverse association between a Mediterranean dietary pattern and mammographic density among current smokers; however, no association was observed among non-smokers [15]. In a cross-sectional study comparing two dietary patterns (Western and Mediterranean), Castello et al. reported a positive association between Western dietary pattern adherence and higher mammographic density among overweight-obese women (BMI ≥ 25), but no associations among women with a BMI < 25 [14]. No associations were observed between the Mediterranean diet and mammographic density. In both studies, the sample population consisted primarily of postmenopausal women with diet being assessed during adulthood. To our knowledge, no studies have examined the association between dietary patterns consumed during adolescence and later life mammographic density.

The aim of this study was to examine the association between adolescent and early adulthood dietary patterns and premenopausal mammographic density. Two different inflammatory-associated dietary patterns were examined: pro-inflammatory and the Alternative Healthy Eating Index (an anti-inflammatory dietary pattern). We also examined whether the associations between these dietary patterns differed by body mass index (BMI), an established risk factor for breast cancer that is also associated with mammographic density.

The median age at the time of mammogram was 44 years old (range 30–55 years) with a mean percent mammographic density of 41%. The greatest overall mean caloric intakes (kcal/day) were observed among women whose dietary patterns were strongly pro-inflammatory in adolescence (Table 1) and early adulthood [see Additional file 2]. Women in the highest quintile of the pro-inflammatory dietary pattern in adolescence had the highest overall mean BMI at the time of mammogram (Table 1). The mean physical activity levels (METS/week) were highest among women who were most adherent to the AHEI in adolescence and early adulthood and among those in the lowest quintile of the pro-inflammatory dietary pattern in early adulthood.

In age-adjusted analyses, we observed as adolescent pro-inflammatory dietary pattern score increased, percent mammographic density decreased (p_trend = 0.005) and non-dense area increased (p_trend < 0.0001) (Table 2). However, these associations were no longer significant in the multivariable-adjusted models (Table 2). No associations were observed between early adulthood dietary pattern and mammographic density, or for adherence to the AHEI dietary pattern and mammographic density in adolescent or early adulthood analyses (Tables 2 and 3).

Upon examination of individual covariates, we observed the greatest attenuation of the effect estimates occurred with adjustment for BMI at the time of mammogram. To more fully examine the impact of BMI on the dietary pattern-mammographic density association, we stratified the percent mammographic density analyses by BMI at the time of mammogram (< 25 vs. ≥ 25) (Table 4). The mean percent mammographic density estimates were lower across all tertiles of dietary pattern scores for women with a BMI ≥ 25 compared to women with a BMI < 25, as expected, but no significant association between dietary pattern and percent mammographic density was observed within either BMI strata. With the exception of the adolescent pro-inflammatory dietary pattern, statistically significant interactions between BMI and adolescent and early adulthood dietary patterns were observed (Table 4). However, this finding is likely a reflection of the strong association between BMI and mammographic density, rather than dietary pattern. Although this suggests the effect of dietary pattern on percent mammographic density could be different across BMI categories, our analyses were not powered to detect the associations between dietary patterns and percent mammographic density with strata of BMI. In our examination of smoking status (current vs. non-smokers) as having a potential interaction effect on the association between dietary pattern and percent mammographic density, we found no evidence of interaction for either dietary pattern in adolescent or early adulthood analyses. In secondary analyses, we examined the average of the adolescent and early adulthood dietary patterns in relation to mammographic density phenotypes among the women who completed both FFQs (n = 677). The results were similar to the main analyses [see Additional file 3]. Secondary analyses were also performed including the breast cancer cases in addition to the controls, and the results were not substantially different.

In this study, we did not observe significant associations between inflammation-associated dietary patterns in adolescence or early adulthood and mammographic density in models fully adjusted for breast cancer risk factors. Although significant associations were observed between an adolescent pro-inflammatory dietary pattern and mammographic density in some age-adjusted models, with a similar but non-significant pattern for the early adulthood pro-inflammatory dietary pattern, these associations did not remain after adjustment for BMI and other breast cancer risk factors. While, as expected, the distribution of percent mammographic density varied by BMI, no associations were observed between dietary patterns and mammographic density within the strata of BMI (< 25, ≥ 25).

To our knowledge, this is the first study to evaluate the relation between adolescent dietary patterns and mammographic density as prior studies have focused on individual foods/food groups and sources of macronutrients (e.g., animal fat). In previous work in the NHSII cohort, Bertrand et al. observed that adolescent animal fat intake was positively associated with premenopausal mammographic density, an association that remained after adjustment for red meat intake and adult intake of animal fat. In addition, adolescent red meat intake had a non-significant positive association with mammographic density [32]. Consistent with these results, the Dietary Intervention Study in Children (DISC) reported that higher adolescent intake of saturated fat was associated with higher breast density measured with non-contrast MRI at ages 25–29 (n = 177 women) [37]. In a study of US Chinese immigrant women (n = 201), Tseng et al. reported adolescent red meat consumption was significantly, positively associated with mammographic density in adulthood. Interestingly, the authors reported that red meat intake in this population was as much as 2.5 times lower, even for women in the highest consumption tertile, compared to consumption levels measured in Western populations, yet an association was still observed [38]. Other studies that have investigated diet in early life and breast density have reported no significant associations [39‐41].

A limited number of studies have examined the dietary patterns in relation to mammographic density, with a focus on mid-life dietary patterns. In the Minnesota Breast Cancer Family Study (n = 1286), the association between a Mediterranean dietary pattern (6 items considered beneficial: vegetables, legumes, fruits and nuts, cereals, fish, and monosaturated to saturated fat ratio; 2 items considered to not be beneficial: meat and dairy; and alcohol intake) and mammographic density was examined among a sample of predominantly postmenopausal women [15]. No overall association was observed between the Mediterranean diet and mammographic density; however, this result varied by smoking status; specifically, consuming a Mediterranean diet was significantly, inversely associated with mammographic density among current smokers, while no association was observed among never smokers. The authors reported vegetables, legumes, and cereals to be the individual food items driving the inverse association among current smokers [15]. More recently, a cross-sectional study evaluated the association between two dietary patterns, Western and Mediterranean, and mammographic density. In this study, Castello et al. observed no association between adherence to a Mediterranean dietary pattern and mammographic density, while consumption of a Western (unhealthy) dietary pattern in adulthood was associated with a higher mammographic density among women with a BMI > 25 [14]. Although these studies evaluated different dietary patterns than the current study, the Western dietary pattern is akin to the pro-inflammatory dietary pattern in that both have been associated with system inflammation, while the Mediterranean dietary pattern and AHEI are anti-inflammatory and overlap in some of the included food groups.

There are some important limitations of this study to be considered. First, measurement error could have impacted the adolescent diet assessment, since participants were asked to recall diet as much as 2 to 3 decades in the past [27, 32]. However, the validity and reproducibility of the NHSII adolescent diet recall have been previously demonstrated [22, 42], and prior studies have observed statistically significant associations between adolescent dietary intake, both individual foods groups (animal fat intake) [32] and dietary patterns (inflammatory, AHEI, and prudent) [2, 4] with later life outcomes including percent mammographic density and breast cancer risk, respectively. It may be that more, not less, robust associations are detectable given the level of measurement error inherent in the retrospective adolescent diet assessment. Another limitation is that the biomarkers used to derive the pro-inflammatory pattern were not obtained during adolescence, but in postmenopausal women. Thus, the foods that were identified as associated with these biomarkers may not be as relevant during the adolescent period resulting in additional dietary exposure misclassification.

A further limitation is related to the timing of the breast density assessment. Breast density measured through a screening mammogram may not capture the most relevant breast density measurement in regard to the impact of adolescent dietary exposures. Studies have shown that breast density declines with age [43], but these changes most commonly occur around menopause, as studies have reported declines among women who are all or mostly 40 years and older [44, 45]. Krishnan et al. evaluated the correlations between mammographic measures (dense area, percent dense area, and non-dense area) over time among women (n = 970, aged 24–83 years) identified from two studies (Melbourne Collaborative Cohort Study, and the Australian Breast Cancer Family Registry population-based case-control study) and reported that within-woman correlations taken at multiple year intervals (2, 4, 6, 7, and 10 years apart) were all highly correlated with normalized estimates ≥ 0.90 after adjustment for age and BMI [46]. Given that the current study included only premenopausal women in the analysis, that density tends to decline around menopause, and that evidence supports high within-woman correlations in measurements across time, we would expect our single premenopausal mammograms to be a reliable source of mid-life density measures for evaluating their association with adolescent/early adulthood dietary patterns. Further studies are needed to better understand the factors that influence breast density measures in the adolescent and early adulthood time periods, when mammograms are not routinely taken. Breast density measured during this time period may be essential to helping us more fully understand the association between adolescent diet and breast cancer risk.

An important strength of this study was the access to digitized mammograms, and the ability to measure dietary patterns at two different time points among the study population, adolescence and early adulthood. Information on the risk factors was also available for both time points, which allowed for necessary adjustments in multivariable analyses.

In conclusion, we observed that, after adjusting for covariates including BMI at mammogram, consuming a pro-inflammatory or AHEI dietary pattern during adolescence or early adulthood was not associated with premenopausal mammographic density. More research is needed to investigate early life diet in relation to breast density using prospectively collected dietary data and breast density measured prior to the age of most mammograms.