Adolescent Dietary Intake and Breast Cancer in Adulthood: A Systematic Review and Meta-analysis

Abstract

Adolescence represents a key opportunity for breast cancer prevention, as the rapid proliferation of breast tissue during puberty creates a critical window of vulnerability for the development of cancerous cells. With increasing research on adolescent dietary factors and breast cancer risk, we conducted a systematic review and meta-analysis to summarize the associations between adolescent diet and risk of breast cancer in adulthood, as well as benign breast disease (BBD) and high mammographic breast density, which are markers for breast cancer. We searched Web of Science, Ovid MEDLINE, Cochrane CENTRAL, and Embase for epidemiological studies assessing dietary intakes in adolescent girls (aged 10–18 y), published through 16 October, 2024, with no language or time restrictions. Study quality was assessed using the Newcastle-Ottawa Scale, and results were pooled using random-effects models. The review included 51 studies, mostly from the United States, with the majority relying on adult recall of adolescent diet, and only 20 studies were assessed as high quality. Higher adolescent intakes of fruits and vegetables [relative risk (RR): 0.90; 95% confidence interval (CI): 0.82, 0.99; n = 3 studies], soy (RR: 0.67; 95% CI: 0.55, 0.82; n = 3), dietary fiber (RR: 0.78; 95% CI: 0.67, 0.92; n = 3), and vegetable fat (RR: 0.76; 95% CI: 0.66, 0.88; n = 2) were associated with lower risks of breast cancer in adulthood. No significant associations were observed for meat and poultry, fish, processed meat/fish, eggs, dairy, milk, grains, alcohol, total fat, animal fat, and isoflavone. Additionally, greater consumption of dietary fiber (RR: 0.64; 95% CI: 0.50, 0.82; n = 2) and vitamin D (RR: 0.77; 95% CI: 0.62, 0.95; n = 2) during adolescence was associated with lower risks of BBD, whereas no dietary associations were observed for mammographic breast density. Our findings underscore the importance of both diet and timing in breast cancer prevention. Future well-designed prospective life course studies are needed to strengthen this evidence base.

This systematic review and meta-analysis was registered with PROSPERO (CRD42024532597).

Statement of significance.

To our knowledge, this is the first comprehensive review to review and quantify the association between adolescent intake of various dietary components and breast cancer risk in adulthood. Current dietary strategies for breast cancer prevention are largely informed by epidemiological evidence from studies involving adult populations. Our findings offer preliminary insights into potential pathways linking dietary exposures during adolescence to breast cancer development. We also identify critical gaps in adolescent research, emphasizing the need for continued high-quality research targeting the formative years to inform future prevention efforts.

Introduction

The global burden of breast cancer is continuing to rise rapidly, and there is an important need for national breast cancer programs to go beyond a focus on screening and consider primary prevention strategies [1]. Apart from genetic predisposition, established modifiable factors that could be targeted in public health programs include reducing alcohol consumption, promoting physical activity, reducing excess body weight, and encouraging breastfeeding, which are all associated with a reduced breast cancer risk [2].

In addition, there is a significant corpus of literature examining the role of diet and breast cancer risk. A recent umbrella review revealed that adherence to a prudent dietary pattern, characterized by high intakes of vegetables, fruits, legumes, cereals, and seafood, confers a decreased risk of breast cancer [3]. On the associations of individual foods and food groups, a dose–response meta-analysis found that increased intakes of fruits, vegetables, cheese, and soy foods were associated with a reduced risk of breast cancer, whereas red and processed meat increased risk of breast cancer [4]. Nonetheless, the existing literature has been primarily focused on the female adult population.

Intriguingly, previous epidemiological studies have suggested that dietary exposures during critical periods across the lifespan, such as from menarche to first childbirth, may have a greater impact on breast cancer risk accumulation. The leading hypothesis is that the rapid proliferation of breast tissue during puberty makes adolescence a critical period of vulnerability for the development of subsequent cancerous cells, suggesting that adolescence represents a key window for breast cancer prevention [5,6]. Subsequently, this has prompted a growing body of research to investigate the influence of dietary factors during adolescence on subsequent breast cancer risk [7,8].

In parallel, some studies have examined the role of adolescent diet in benign breast disease (BBD) and high mammographic breast density, both of which are well-defined markers of breast cancer [9,10]. Specifically, BBD refers to a group of non-cancerous breast disorders characterized by abnormal cell growth within the breast tissue. It can be classified into 3 categories based on histological findings: 1) nonproliferative or slow-growing lesions, such as simple breast cysts, 2) proliferative lesions without atypia, which involve increased growth of morphologically normal cells, including usual ductal hyperplasia, intraductal papilloma and sclerosing adenosis, and 3) proliferative lesions with atypia, which involve increased growth of morphologically abnormal cells, such as atypical ductal or lobular hyperplasia, lobular carcinoma in situ. Compared with nonproliferative lesions, the risk of developing breast cancer is approximately twice as high in proliferative lesions without atypia, and 4 times higher in those with atypia, due to the heightened cell turnover, which raises the likelihood of accumulating genetic mutations during replication [11,12]. However, dietary exposures during adolescence are thought to influence circulating estrogen levels and insulin-like growth factors, which in turn may affect breast density. The presence of higher mammographic density may be associated with increased breast cancer risk, primarily through a more protumorigenic immune microenvironment, which may aid escape from immune regulation for early tumor cell variants [7,10,13]. Given that both BBD and high mammographic breast density are highly prevalent [[14], [15], [16]], understanding the influence of dietary factors associated with both conditions may provide insights into the etiology of breast carcinogenesis and possible strategies for targeting those identified to be at risk [11,13].

To our knowledge, the associations between adolescent diet and risk of breast cancer have been summarized in 2 narrative reviews (published in 2013 and 2016) [17,18], 1 systematic review conducted over 20 y ago [19], a 2022 meta-analysis specifically on adolescent milk intake [20], and a 2013 narrative review on adolescent risk factors for BBD [21], but none on breast density. Evidence across current reviews was, however, generally inconsistent and inconclusive, suggesting the need for an updated systematic search and meta-analysis to summarize these findings. As such, we aimed to conduct a systematic review and meta-analysis on the associations between adolescent dietary intake and risk of breast cancer, BBD, and mammographic breast density in adulthood.

Methods

This systematic review and meta-analysis followed the PRISMA reporting guidelines [22]. The review protocol was registered in PROSPERO (CRD 42024532597).

Search strategy

A systematic search was conducted in Web of Science, Ovid MEDLINE, Cochrane CENTRAL, and Embase for articles published through 4 March, 2024, with no restrictions on language or time. The search was updated on 16 October, 2024 to identify any newly published relevant articles. The search strategy combined keywords “adolescent,” “youth,” “diet,” “breast cancer,” “breast neoplasm,” “breast density,” and “benign breast disease.” Reference lists of the selected publications were also hand-searched for identification of additional relevant studies. The full search strategy is detailed in Supplemental Table 1.

Selection criteria

We used the Population, Exposure, Comparator, Outcomes, Study design criteria to identify eligible studies. Studies were included if the following criteria were fulfilled: 1) epidemiological studies in preadolescent (ages 10–12) or adolescent (ages 13–18) girls, 2) studies that evaluated the associations between dietary pattern, food groups, or nutrient intake during adolescence as exposure of interest, and 3) breast cancer, BBD or mammographic breast density in adulthood as outcome of interest. Studies were excluded if they were: 1) populations outside the specified age range, breast cancer survivors, or populations with underlying medical conditions, 2) breast cancer, BBD, or mammographic breast density that was examined at any other unspecified time points, and 3) literature or narrative reviews, or nonhuman studies.

Three authors (CKYL, GHL, and NK) independently screened the titles and abstracts of articles for eligibility, and full texts of the potentially relevant studies were retrieved and evaluated. Disagreements were resolved by discussion with 2 senior investigators (AC and MF-FC).

Data extraction and synthesis

The following data were extracted using a standardized template which consisted of the following: 1) study characteristics (study design, geographic region, sample size, follow-up duration), 2) population characteristics (adolescent age range, inclusion and exclusion criteria), 3) types of dietary exposure, level of intake and method of assessment, 4) study outcome, number of cases and ascertainment method, and 5) effect estimates, which included relative risks (RR), odds ratios (OR), hazard ratios (HR), beta-coefficients, and mean percent density values, along with their corresponding 95% confidence intervals (CI) for the most adjusted model and covariates included. When necessary, we contacted the respective authors to obtain additional information that would be relevant. Data extraction was carried out by 2 pairs of independent authors (CKYL and NK; GHL and EL). Any disagreements were resolved by discussion with a senior investigator (AC).

Risk of bias assessment

We evaluated the quality of studies using the Newcastle-Ottawa Scale (NOS) [23]. The NOS comprises 3 domains: 1) selection, 2) comparability, and 3) exposure for case-control studies, or outcome for cohort studies. A predefined set of confounders was selected a priori to assess the domain on comparability, where 1 star was awarded to studies that adjusted for age, energy intake, and BMI in their statistical analyses, and an additional star was awarded for adjustments in age at menarche or family history of breast cancer. Each study received a score ranging from 0 (low quality) to 9 (high quality), and scores of ≥7 were appraised as “high quality.” Critical appraisal of the included studies was conducted independently by 3 authors (CKYL, NK, and EL). Differences in the overall judgment between authors were resolved by discussion with a senior investigator (AC).

Data synthesis and statistical methods

Dietary variables, that is, food, food groups, or nutrients, were categorized and meta-analyzed. When there were multiple dietary variables from a single publication grouped under the same category, for example, red meat and poultry, we pooled the individual effect estimates using a fixed-effects model to obtain an overall estimate for use in the meta-analysis of studies examining “meat and poultry.” If similar dietary variables were identified within the same study cohort, results from the publication with a longer follow-up and/or larger number of cases were included in the main analysis, whereas the other studies on the same cohort were considered in sensitivity analysis by alternate inclusion to determine the robustness of the result.

As studies included in this meta-analysis reported measures of association based on a mixture of equal-sized groups (i.e., tertile/quartile/quintile) and unequal-sized groups (i.e., servings/level or frequency of intake), we transformed the effect estimates from each study to a standard scale of effect to facilitate comparison between the top and bottom tertile using methods that were previously described [[24], [25], [26], [27]]. Under a log-normal distribution, the difference in means between the top and bottom tertile, quartile, and quintile is set apart by 2.18, 2.54, and 2.80 SDs, respectively. The log effect estimates and corresponding SE were then multiplied by the following: 1) 2.18/2.54, for conversion of quartile to tertile; 2) 2.18/2.80, for conversion quintile to tertile; and 3) 2.18/x, for conversion of any unequal-sized group to tertile, where x is the difference in means between the top and bottom category expressed in SD units.

RR, OR, and HR were eligible for inclusion in the meta-analysis. To account for between-study variations, random-effects models were used to calculate the pooled effect estimates, which were expressed in RR with its corresponding 95% CI for the association between adolescent dietary variables and breast cancer [28,29]. For studies reporting mammographic breast density, standardized mean differences were calculated by obtaining the difference between mean percent density values of the highest and lowest categories, followed by pooling using random-effects models. When there were ≥5 or more studies included in the meta-analysis of each dietary variable, we investigated publication bias by visual inspection of the funnel plot for asymmetry and Egger’s regression test [28].

We also used Cochran’s Q test (P-heterogeneity) and I² statistics to evaluate heterogeneity. Because of the small number of studies for each dietary variable (n < 10), we were unable to perform meta-regression analyses to explore potential sources of heterogeneity [28]. Instead, we used the sequential exclusion strategy for dietary variables with >2 studies when significant heterogeneity was observed (I² > 50%) [28,30]. Studies contributing to the largest reduction in heterogeneity were identified. We then further examined the pooled estimates before and after exclusion for consistency.

Further sensitivity analyses were also conducted to examine the robustness of our findings: 1) excluding 1 study at a time and recalculating the pooled estimates for the remaining studies, 2) restricting the analysis to high-quality studies, and 3) calculating the pooled estimates using unconverted data. A cut-off of 10% was set for the identification of a substantial difference from the original estimate. As studies examining breast cancer included both case-control and cohort studies, we combined both study designs (case-control and cohort studies) in the meta-analyses after verifying that the incidence of breast cancer was <10% in the cohort studies. Thereafter, we also conducted a subgroup analysis based on the types of study design in the sensitivity analysis.

Results of the meta-analysis were presented using forest plots, illustrating the pooled and individual study effect estimates with their corresponding 95% CI, weightage of each study, and I² statistics. All statistical analyses were performed using Stata 18 (StataCorp LLC), and 2-tailed P values of <0.05 were considered statistically significant.

Dietary variables (e.g., dietary patterns) were summarized narratively if there were fewer than 2 studies to meta-analyze [[31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50]], distribution of participants in each level of intake was not available [[51], [52], [53]], P values and CIs were not reported [54], differences in statistical unit, for instance, adjusted means and beta-coefficients [42,43,55], or dietary patterns sharing different constituent foods [43,44,47,[56], [57], [58]].

Results

Study selection

The flow chart of the study selection process is shown in Figure 1. Of the 3209 articles identified in the initial search, 2397 were screened by title and abstract after removal of duplicates. The full texts of 124 articles were evaluated, and 75 articles were excluded. Two additional eligible articles were identified through handsearching of reference lists. Finally, 51 studies were selected for inclusion in this systematic review, all of which were published in English.

FIGURE 1.

PRISMA flow diagram of the study selection process.

Study characteristics

The characteristics of the included studies are presented in Table 1 [[3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81]]. There were 15 case-control [31,32,52,57,[64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74]], 1 nested case-control [54], and 16 cohort studies on breast cancer [[33], [34], [35], [36], [37],[45], [46], [47],51,56,[58], [59], [60], [61], [62], [63]]; 11 cohort studies on BBD [[38], [39], [40], [41],48,49,53,[75], [76], [77], [78]]; and 8 cohort studies on mammographic breast density [[42], [43], [44],50,55,[79], [80], [81]]. Of these, 40 studies were conducted in the United States [[35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45],[47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57],[60], [61], [62], [63], [64], [65],67,70,[73], [74], [75], [76], [77], [78], [79], [80], [81]], 5 in Canada [31,32,68,69,71], 3 in Europe [58,59,72], and 3 in China [33,46,66]. Studies were published between 1989 and 2024, and sample sizes ranged from 201 to 75,826 participants. Except for the 5 prospective studies from the Growing Up Today Study (GUTS) cohort [38,48,[75], [76], [77]], which assessed dietary intake directly from adolescents using self-administered food frequency questionnaires (FFQ) delivered by mail or internet, the remaining studies relied on adults’ retrospective recall of their adolescent intake. These were predominantly conducted in person using self-administered [58,59] or interviewer-administered FFQ [33,46,64,66,70,72], or FFQ completed by mail [31,32,[34], [35], [36], [37],[39], [40], [41],[43], [44], [45],47,[49], [50], [51],[53], [54], [55], [56], [57],[60], [61], [62], [63],71,73,80]. Other methods included dietary questionnaires (with unspecified format), which were either mailed [78,79,81] or completed in person [42,67], and dietary interviews conducted by telephone [68,69], in person [52,65] or via computer-assisted mode [74]. Breast cancer diagnoses were obtained from medical records or national registries in 16 studies [[31], [32], [33], [34],46,52,54,[57], [58], [59], [60],[66], [67], [68], [69],71], histologically or cytologically confirmed in 5 studies [73,65,70,72,74], and self-reported in 8 studies, with 99% of the cases verified by pathology reports [[35], [36], [37],45,47,51,62,63]. Three additional studies included self-reported cases, but the proportion of verified cases was not reported [56,60,61]. BBD cases were identified through biopsy-verified self-reports in 6 studies [[39], [40], [41],49,53,78], whereas the remaining 5 studies relied on self-reported diagnoses with self-reported biopsy confirmation [38,48,[75], [76], [77]]. Finally, breast density was assessed through mammograms in all 8 studies [[42], [43], [44],50,55,[79], [80], [81]].

TABLE 1.

Characteristics of studies included in systematic review and meta-analysis of adolescent dietary intake and breast cancer in adulthood.

Lead author	Publication year	Country	Study cohort	Sample size	Cases	Adolescent age range (y)	Dietary variable	Study design1
Breast cancer
Haraldsdottir [59]	2017	Iceland	AGES-Reykjavik	2882	91	14–19	Fish	Cohort
Haraldsdottir [58]	2018	Iceland		3130	97	14–19	Meat, salted or smoked meat, milk and milk products, rye bread, oatmeal, 4 dietary patterns	Cohort
Frazier [54]	2003	United States	NHS	NR2	843	12–18	Protein, animal fat, vegetable fat, MUFA, saturated fat, total fat, folate, vitamin A, vitamin C, vitamin D, vitamin E, dietary fiber	Nested case-control
Riseberg [60]	2024	United States		63,847	5733	13–18	Total dairy, total milk	Cohort
Linos [51]	2008	United States	NHS II3	39,268	455	13–18	Red meat, processed meat	Cohort
Linos [35]	2010	United States		39,268	455	13–18	Total fat, saturated fat, MUFA, trans fat, animal fat, vegetable fat, total milk, total dairy	Cohort
Frazier [34]	2004	United States	NHS II	47,355	361	13–18	Total dairy, total red meat, total vegetables, total fruit, total chicken, total fish, total bread, total calories, total fat, animal fat, vegetable fat, saturated fat, MUFA, PUFA, folate, vitamin A, vitamin C, vitamin D, vitamin E, total fiber	Cohort4
Farvid [36]	2015	United States		44,263	1135	13–18	Carbohydrate, glycemic index, glycemic load, dietary insulin index, insulin load	Cohort
Farvid [61]	2015	United States		44,231	1132	13–18	Total red meat, poultry, fish, egg, legumes, nuts	Cohort
Farvid [62]	2016	United States		44,263	1347	13–18	Whole grain, refined grain	Cohort
Farvid [63]	2016	United States		44,263	1118	13–18	Dietary fiber	Cohort
Farvid [45]	2016	United States		44,223	1347	13–18	Fruits, vegetables, total fruit and vegetables, fruit juice	Cohort
Harris [56]	2016	United States		45,204	1477	13–18	Dietary patterns: prudent, western, fast food, AHEI	Cohort4
Harris [47]	2017	United States		45,204	1477	13–18	Inflammatory dietary pattern	Cohort5
Farvid [37]	2018	United States		44,264	1318	13–18	Total dairy products, milk, total calcium, vitamin D	Cohort
Lee [33]	2009	China	SWHS	73,225	594	13–15	Soy protein, isoflavones	Cohort
Baglia [46]	2016	China		70,578	1034	13–15	Soy protein	Cohort
Pryor [64]	1989	United States	—	362	172	12	Fat, total fiber	Case control
Freudenheim [65]	1995	United States	—	1549	740	16	Total alcohol	Case control
Potischman [57]	1998	United States	—	3148	1647	12–13	Dairy, animal fat, high-fat foods, fruits and vegetables	Case control
Shu [66]	2001	China	SBCS	3015	1459	13–15	Total soy foods, fresh legumes, dried beans, total vegetables/fruits, vegetables, fruits, rice/wheat products, total meat, eggs, seafood, milk, preserved meat/fish	Case control
Wu [67]	2002	United States6	—	1095	501	12–18	Tofu	Case control
Terry [52]	2006	United States	LIBCSP	3064	1508	< 20	Alcohol	Case control
Thanos [31]	2006	Canada	—	6444	3024	10–15	Isoflavone, lignan, total phytoestrogen	Case control
Knight [68]	2007	Canada	—	2107	972	10–19	Milk, salmon/tuna	Case control
Blackmore [69]	2008	Canada	—	1894	759	10–19	Milk, salmon/tuna	Case control
Korde [70]	2009	United States6	—	1563	597	12–19	Soy	Case control
Anderson [71]	2013	Canada	OWDHS	5690	2353	10–15	Isoflavone, lignan, total phytoestrogen	Case control
Liu [32]	2014	Canada		6164	2865	10–15	Dietary fiber, vegetable protein, vegetable fat, total nuts	Case control
Donat-Vargas [72]	2021	Spain	EpiGEICAM	1578	770	12–19	Alcohol	Case control
Pathak [73]	2021	United States	PWHS	415	131	12–13	Raw/short-cooked cabbage and sauerkraut	Case control
Hirko [74]	2024	United States	YWHHS	3173	1796	< 20	Alcohol	Case control
Benign breast diseases
Berkey [48]	2013	United States	GUTS	6611	108	< 20	Beans, lentils and soybeans	Cohort
Boeke [75]	2014	United States		6593	122	9–16	a-carotene, b-carotene, b-cryptoxanthin, lutein/zeaxanthin, lycopene, vitamin A	Cohort
Boeke [76]	2015	United States		6593	122	9–15	Vitamin D	Cohort
Berkey [38]	2019	United States		5677	173	9–15	Dairy milk, yogurt, cheese, eggs, total fruit, fruit juice, total vegetables, legumes, peanut butter and nuts, fish, white potatoes, bread and fries, olive oil, alcohol, animal (nondairy) protein, dairy protein, vegetable protein, animal (nondairy) fat, dairy fat, vegetable fat, total carbohydrates, total fiber, total calories	Cohort
Berkey [77]	2020	United States		6554	173	14–17	Alcohol, peanut butter and nut, dietary fiber	Cohort
Byrne [78]	2002	United States	NHS II	75,826	905	15–17	Alcohol	Cohort
Baer [53]	2003	United States		29,494	470	13–18	Total fat, animal fat, vegetable fat, MUFA, vitamin C, vitamin A, total carotenoids, lycopene, lutein and zeaxanthin, fiber, fruits and vegetables	Cohort5
Su [49]	2010	United States		29,480	682	13–18	Fiber, total nuts, fruits, vegetables, fruits and vegetables, beans/lentils, peas or lima beans, cold breakfast cereal	Cohort4
Su [40]	2012	United States		29,480	682	13–18	Vitamin D, calcium, dairy protein, total dairy, total milk	Cohort4
Liu [41]	2012	United States		29,117	659	13–18	Dietary folate	Cohort5
Su [39]	2015	United States		29,480	682	13–18	Total fat, animal fat, vegetable fat, saturated fat, MUFA, PUFA, trans-unsaturated fat, retinol activity equivalents from food, vitamin E, a-carotene, b-carotene, b-cryptoxanthin, lycopene, lutein and zeaxanthin	Cohort4
Mammographic breast density
Tseng [42]	2011	United States6	—	201	—	12–17	Red meat, tofu, green vegetable, fruit	Cohort
Vachon [79]	2005	United States	MBCF	1893	—	< 18	Alcohol	Cohort
Sellers [43]	2007	United States		1552	—	12–13	High-fat meats, dairy products, animal fat, high-fat foods, fish and chicken, fruits, vegetables	Cohort
Bertrand [80]	2016	United States	NHS II3,7	687	—	13–18	Animal fat, red meat	Cohort
Yaghjyan [55]	2016	United States		743	—	13–18	Total fiber, total nuts, fruit, vegetables	Cohort
Liu [81]	2018	United States		1211	—	15–17	Alcohol	Cohort
Yaghjyan [50]	2020	United States		751	—	13–18	Caffeine	Cohort
Garzia [44]	2021	United States		709	—	13–18	Proinflammatory dietary pattern, AHEI	Cohort

Abbreviations: AGES, age, gene/environment susceptibility; AHEI, Alternative Healthy Eating Index; BBD, benign breast disease; Epi-GEICAM, Epidemiological study of the Spanish Breast Cancer Research Group (Grupo Español de Investigación en Cáncer de Mama); GUTS, Growing Up Today Study; LIBCSP, Long Island Breast Cancer Study Project; MBCF, Minnesota Breast Cancer Family; NHS, Nurses’ Health Study; NR, not reported; OWDHS, Ontario Women’s Diet and Health Study; PWHS, Polish Women's Health Study; SBCS, Shanghai Breast Cancer Study; SWHS, Shanghai Women’s Health Study; YWHHS, Young Women’s Health History Study.

¹Person-time for cohort studies examining breast cancer and BBD was defined as the period from assessment of adolescent dietary intake to diagnosis, death, lost to follow-up, or end of study, unless otherwise stated.

²The sample size was not reported; the authors only stated that controls were matched to cases at a 10:1 ratio.

³Analyses were conducted on premenopausal women only.

⁴Statistical analysis was inclusive of cases diagnosed before and after adolescent dietary intake was assessed, and further sensitivity analyses were conducted by restricting to cases that were diagnosed after assessment of adolescent intake.

⁵Statistical analysis was inclusive of cases diagnosed before and after adolescent dietary intake was assessed.

⁶Participants were of Asian ethnicity.

⁷Participants were controls from a case-control study nested within the NHS II cohort.

Study quality assessment

Of the 51 studies, only 20 were assessed as high-quality [33,[35], [36], [37],[44], [45], [46], [47],50,51,55,[58], [59], [60], [61], [62], [63],72,73,80]. Among the 16 case-control studies (Supplemental Table 2), only 4 adjusted for pre-identified key covariates (i.e., age, BMI, energy intake, and age at menarche or family history of breast cancer) [65,67,72,73], and just 1 study used a validated dietary assessment tool [72]. Of the 35 cohort studies (Supplemental Table 3), only half adjusted for key covariates [[35], [36], [37],44,45,47,50,51,53,55,[60], [61], [62], [63],75,76,80], and just 9 studies achieved a follow-up rate of 80% or more [33,38,42,46,48,58,59,77,81].

Associations between adolescent dietary intake and breast cancer in adulthood

The associations between adolescent dietary intake and breast cancer, BBD, and mammographic breast density are summarized in Table 2, and the forest plots are illustrated in Supplemental Figures 1, 2, and 3.

TABLE 2.

Pooled results of studies examining associations between adolescent dietary intake and breast cancer in adulthood.

Dietary variable	Breast cancer				BBD				Mammographic breast density
	n	RR (95% CI)	P value	I2 (%)	n	RR (95% CI)	P value	I2 (%)	n	SMD (95% CI)	P value	I2 (%)
Food groups
Fruits and vegetables1	3	0.90 (0.82, 0.99)	0.036	0.0	2	1.00 (0.85, 1.18)	0.999	0.0
Fruits	2	0.88 (0.72, 1.09)	0.243	67.9	2	0.89 (0.73, 1.08)	0.230	10.2
Vegetables1	3	0.75 (0.55, 1.01)	0.057	65.9	2	1.00 (0.83, 1.21)	0.965	0.0
Nuts and legumes	5	0.84 (0.75, 0.94)	0.003	58.7#	2	0.74 (0.48, 1.14)	0.172	72.3
Excluding soy	3	0.90 (0.82, 0.99)	0.023	33.4
Soy	3	0.67 (0.55, 0.82)	< 0.001	32.3
Red meat									2	0.13 (–2.97, 3.23)	0.936	76.5#
Meat and poultry	3	0.97 (0.88, 1.08)	0.611	0.0
Fish	4	0.93 (0.83, 1.03)	0.170	0.0
Processed meat/fish	2	1.16 (0.91, 1.49)	0.228	17.1
Eggs	2	0.97 (0.82, 1.14)	0.699	29.9
Dairy	6	0.90 (0.77, 1.06)	0.217	73.2#	2	0.93 (0.80, 1.09)	0.372	0.0
Milk	4	0.89 (0.72, 1.11)	0.302	82.1#	2	1.02 (0.63, 1.66)	0.931	71.0
Grains2	3	0.89 (0.73, 1.09)	0.263	51.9
Alcohol	3	1.07 (0.58, 1.96)	0.832	66.6	2	1.31 (0.65, 2.66)	0.454	89.0#	2	0.66 (–0.70, 2.02)	0.340	0.0
Nutrients
Dietary fiber	3	0.78 (0.67, 0.92)	0.003	49.4	2	0.64 (0.50, 0.82)	< 0.001	0.0
Vitamin D3					2	0.77 (0.62, 0.95)	0.013	0.0
Total fat	2	0.91 (0.72, 1.14)	0.415	0.0	2	0.98 (0.79, 1.22)	0.876	0.0
Vegetable fat	2	0.76 (0.66, 0.88)	<0.001	15.4	2	0.82 (0.59, 1.15)	0.251	0.0
Animal fat	2	1.02 (0.88, 1.18)	0.811	0.0	2	1.06 (0.81, 1.39)	0.663	0.0	2	1.54 (–1.45, 4.53)	0.313	77.1#
MUFA					2	0.74 (0.45, 1.24)	0.257	10.2
Animal protein					2	1.00 (0.86, 1.18)	0.968	0.0
Vitamin A					2	0.94 (0.69, 1.29)	0.722	0.0
Lycopene					2	1.08 (0.80, 1.46)	0.616	0.0
Carotenoids					2	0.90 (0.69, 1.17)	0.420	48.3
Lutein and zeaxanthin					2	0.91 (0.65, 1.29)	0.604	0.0
Isoflavone	2	0.97 (0.71, 1.32)	0.827	85.0#

Abbreviations: BBD, benign breast disease; CI, confidence interval; RR, relative risk; SMD, standardized mean difference.

^#denotes P-heterogeneity < 0.05.

¹Studies did not differentiate between starchy and nonstarchy vegetables.

²Consists of both whole and refined grains.

³Vitamin D intake includes supplement use.

Breast cancer

When data on various food or food groups were pooled, higher intakes of fruits and vegetables, nuts and legumes, and soy consumed during adolescence (ages 10–19 y) were associated with a 10%, 16%, and 33% lower risk of breast cancer, respectively, when comparing the top and bottom tertiles (Fruits and vegetables—RR: 0.90; 95% CI: 0.82, 0.99; I² = 0.0; nuts and legumes—RR: 0.84; 95% CI: 0.75, 0.94; I² = 58.7; Soy—RR: 0.67; 95% CI: 0.55, 0.82; I² = 32.3; Supplemental Figure 1A, D, and F). The results for nuts and legumes weakened after excluding studies on soy intake (RR: 0.90; 95% CI: 0.82, 0.99), suggesting that the association was primarily driven by soy.

When data on nutrient intake were pooled, adolescents in the top tertile of dietary fiber and vegetable fat intakes had a 22% and 24% lower risk of breast cancer, respectively, compared with those in the bottom tertile (dietary fiber—RR: 0.78; 95% CI: 0.67, 0.92; I² = 49.4; vegetable fat—RR: 0.76; 95% CI: 0.66, 0.88; I² = 15.4; Supplemental Figure 1O and Q). No other statistically significant associations were observed when fruits and vegetables were analyzed separately, nor for meat and poultry, fish, processed meat/fish, eggs, dairy, milk, grains, alcohol, total fat, animal fat, and isoflavone.

We sequentially excluded studies to identify sources of heterogeneity in dietary variables with significant heterogeneity (I² > 50%). For nuts and legumes intake, the I² value decreased from 58.2% to 0% after excluding Farvid et al. [61] (RR: 0.79; 95% CI: 0.73, 0.86). Similarly, stratifying by study design resulted in the same outcome, with Farvid et al. [61] being the only cohort study, whereas the remaining studies were case-control studies (RR: 0.79; 95% CI: 0.73, 0.86; I² = 0%) (Supplemental Table 4). For dairy intake, substantial heterogeneity was still observed regardless of the studies excluded. For milk intake, stratification by study design revealed a significant direct association with breast cancer in cohort studies (RR: 1.08; 95% CI: 1.01, 1.17; I² = 0%), and an inverse association in case-control studies (RR: 0.67; 95% CI: 0.53, 0.83; I² = 0%; Supplemental Table 4). However, further exploration of heterogeneity was not possible due to the lack of detailed reporting on food groups in some studies (e.g., omission of specific information on the types of milk, such as flavored compared with plain milk) [60,66,68].

When evaluating publication bias, the funnel plot for nuts and legumes intake suggests potential bias, indicated by the missing studies in the bottom right of the funnel plot (Egger’s test P = 0.583; Supplemental Figure 4).

Among the sensitivity analyses conducted, none of the pooled estimates were substantially influenced when we omitted 1 study at a time, restricted our analysis to high-quality studies (NOS ≥ 7), and using unconverted data. When we included similar dietary variables from the same study cohort alternately, similar estimates were observed, with the only exception observed in vegetable fat where the significant inverse association was no longer observed following the alternate inclusion of study by Linos et al. [35] (RR: 0.89; 95% CI: 0.69, 1.16) (Supplemental Table 5).

Benign breast disease

When data on adolescent nutrient intake were pooled, higher intakes of dietary fiber and vitamin D (from food and supplement use) were associated with a 36% and 23% lower risk of BBD, respectively, when comparing the top and bottom tertiles (dietary fiber—RR: 0.64; 95% CI: 0.50, 0.82; I² = 0.0; vitamin D—RR: 0.77; 95% CI: 0.62, 0.95; I² = 0.0; Supplemental Figure 2H, I).

There were no statistically significant associations between any of the remaining nutrients (i.e., total fat, vegetable fat, animal fat, monounsaturated fat, animal protein, vitamin A, lycopene, carotenoids, lutein and zeaxanthin) nor food groups (i.e., fruits and vegetables, nuts and legumes, dairy, milk and alcohol) with BBD risk. Similar estimates were observed in the sensitivity analyses when we alternately included similar dietary variables from the same study cohort and performed the analyses using unconverted data (Supplemental Table 5).

Mammographic breast density

Existing studies have only examined adolescent dietary intake of red meat, alcohol, and animal fat in relation to mammographic breast density in adulthood [43,[79], [80], [81]], but no statistically significant associations were observed for any of these dietary variables.

Because of the limited number of studies on mammographic breast density, we were only able to perform sensitivity analysis using unconverted data, and the results were similar to the original estimates (Supplemental Table 5).

Results of other adolescent dietary variables with breast cancer in adulthood

The results of adolescent dietary variables that could not be evaluated in the meta-analysis are summarized in Supplemental Table 6.

In studies evaluating breast cancer as the outcome measure [[31], [32], [33], [34], [35], [36], [37],46,51,52,54,[56], [57], [58]], higher intakes of phytoestrogen [31], lignan [31], and vegetable protein [32] during adolescence were associated with lower risks of breast cancer. Conversely, there were no associations reported among any of the dietary variables in studies evaluating BBD as the outcome measure [[38], [39], [40], [41],43,53]. In the study by Tseng et al. [42], a higher intake of red meat in adolescence was associated with a higher mammographic breast density. None of the other studies reported any significant associations between the remaining adolescent dietary variables and mammographic breast density [[42], [43], [44],50,55]. Lastly, results of the studies evaluating healthy and unhealthy dietary patterns with breast cancer and its markers were not found to be statistically significant [43,44,47,[56], [57], [58]].

Discussion

This systematic review and meta-analysis evaluated the associations between different aspects of dietary intakes during adolescence and subsequent risk of breast cancer, BBD and mammographic breast density in adult females. Greater intakes of fruits and vegetables, soy, dietary fiber, and vegetable fat during adolescence were associated with lower risks of breast cancer later in adulthood. Additionally, greater intakes of dietary fiber and vitamin D during adolescence were associated with lower risks of BBD. We did not find any significant associations between any dietary variables and mammographic breast density. To our knowledge, this review is the first to comprehensively quantify the associations between adolescent intake of various dietary components and breast cancer in later adulthood.

Comparison with existing meta-analyses on adult diet and breast cancer

Breast cancer

Existing meta-analyses on adult diet and breast cancer risk found a 9%–11% lower risk of breast cancer for fruits and vegetables intake [82,83], and an 8%–17% reduced risk for fiber intake [84,85]. Consistent with these findings, our study observed a 10% lower risk for fruits and vegetables, and a 22% reduced risk for fiber intake. However, no significant associations were found when fruits and vegetables were analyzed separately. This may be due to the smaller individual protective effects, as reported in previous adult studies [4,82]. For fiber intake, the meta-analysis by Farvid et al. [84] found that adult females consumed ∼34.5 g/d in the highest category, whereas adolescents in our meta-analysis consumed a relatively lower intake of 27.5 g/d in the top tertile [32,63,64]. Although this is slightly lower than the adult intake, it still falls within the widely recommended range of 25–30 g/d for adults [86]. Our findings suggest a potential reduction in breast cancer risk, highlighting the value of promoting adequate fiber intake early in life. Our meta-analysis also found that higher intake of soy foods was associated with a 33% lower breast cancer risk, whereas no association was observed for soy isoflavones. This is consistent with existing meta-analyses, which reported a 3.5%–13% lower risk of breast cancer for adult soy intake, with weak or no association for soy isoflavones [4,87,88]. The difference may be due to isoflavones being only 1 component of soy, whereas soy foods contain a complex mixture of nutrients such as protein, dietary fiber, B-vitamins, calcium, and other minerals, which may provide additional protective effects against breast cancer. Therefore, isoflavones alone may not capture the full benefits of consuming whole soy foods.

To our knowledge, only 1 meta-analysis has examined the association between adult vegetable fat intake and breast cancer, finding no significant association [89]. Another meta-analysis on adult dietary fat intake found an increased risk of breast cancer with higher intakes of total and saturated fat intake [90]. In contrast, our analysis of adolescent intake showed a 24% lower breast cancer risk, but the association was no longer significant in the sensitivity analysis when we alternated the inclusion of a study from the same cohort that focused only on premenopausal females [35]. More epidemiologic data are needed to strengthen these observations.

Benign breast disease

To date, no existing meta-analysis has quantified the association between dietary intake during adolescence or adulthood and BBD risk. We demonstrated that greater intakes of dietary fiber and dietary vitamin D were associated with a 36% and 23% decreased risk of BBD, respectively. However, it is important to highlight that our meta-analysis was based on data from 2 main cohorts, Nurses’ Health Study (NHS) II and GUTS, which comprised the offspring of females from NHS II. The NHS II specifically included proliferative BBD that were biopsy-verified (both atypical hyperplasia and nonatypical hyperplasia), but the GUTS cohort did not differentiate between proliferative or nonproliferative BBD. It is therefore unclear whether these findings truly reflect the beneficial effects of vitamin D and dietary fiber on BBD. Given that proliferative forms of BBD are associated with a significantly increased risk of subsequent breast cancer, it is important for future studies to ensure clear classification of BBD subtypes to better clarify these associations.

Mammographic breast density

Along with proliferative BBD, having a high mammographic breast density (i.e., dense breast tissue) is also recognized as an independent risk factor for breast cancer [10]. A systematic review reported suggestive evidence that adult females who adhered to a healthy dietary pattern—characterized by higher intakes of cereals, vegetables, fruits, and vegetable oils, and lower intakes of saturated and trans fat, red meat, processed foods, and alcohol—tended to have lower mammographic breast density [91]. Although it has been suggested that certain dietary exposures may affect circulating estrogen levels, which may increase breast density, we did not detect any significant associations between red meat, alcohol, or animal fat with mammographic breast density. Further investigation is needed considering the small number of studies available, as well as the difficulty in pooling studies with differing statistical units, for instance, red meat reported in adjusted means [43,80] and OR [42].

Biological mechanisms of adolescent diet and breast cancer

As abundant sources of fiber and antioxidants, fruits, vegetables, and soy are thought to ameliorate breast cancer risk by suppressing angiogenesis, decreasing oxidative stress and DNA damage, inhibiting cell proliferation, and inducing apoptosis in cancer cells [83]. Additionally, fiber also lowers circulating levels of estrogen by inhibiting intestinal reabsorption and increasing fecal excretion [4], whereas exposure to phytoestrogens from soy (i.e., genistein and daidzein) during breast development may promote earlier differentiation of the terminal cells, thereby reducing the risk of breast carcinogenesis [92]. Similarly, although the specific sources of vegetable fat were not specified in the studies examining vegetable fat [32,34], MUFAs and PUFAs found in vegetable fats may reduce breast cancer risk through alterations in oxidative stress, inflammation, and regulation of gene expression in cell signaling pathways [93]. However, the total amount and type of dietary fat also warrant consideration. Among adult females, higher intakes of total and saturated fat have been associated with increased breast cancer risk [90]. In terms of breast cancer outcomes among breast cancer survivors, the Women’s Intervention Nutrition Study demonstrated that reducing dietary fat intake improved relapse-free survival rate [94], and a meta-analysis reported higher breast cancer-specific mortality with greater adult saturated fat intake [95]. These associations may be partly explained by the role of high-fat intake in promoting weight gain and adiposity, which can lead to increased estrogen production, insulin resistance, and inflammatory markers, all of which are implicated in the development, recurrence, and progression of breast cancer, particularly in estrogen receptor-positive subtypes [7,94]. Therefore, not all fats exert the same biological effects, and both the type and timing of fat intake may influence long-term breast cancer risk.

Although the biological mechanisms for BBD remain less well understood, it is plausible that the protective mechanisms of dietary fiber and vitamin D may be mediated in a similar way as those in breast cancer.

Public health implications

With the growing body of work from primary studies that suggest potential involvement of dietary exposures during adolescence in the etiology of breast cancer, our review has taken a step forward by demonstrating that even as early as adolescence stage, intakes of fruits and vegetables, soy, dietary fiber and vegetable fat can be associated with lower risks of breast cancer later in life. We have also demonstrated that adolescent intakes of dietary fiber and vitamin D are associated with lower risks of BBD.

To date, dietary strategies for preventing breast cancer have been largely informed by epidemiological evidence from studies involving adult populations. Moreover, apart from alcohol, which has been well-documented to have a direct association with both pre- and postmenopausal breast cancer, the current evidence on diet and breast cancer prevention as reported in the 2017 Breast Cancer Report by the World Cancer Research Fund and American Institute for Cancer Research is mostly concluded to be limited [2]. Breast cancer incidence is known to peak at a younger age in Asian females compared with females in Western populations [96]. In addition, the global incidence of early-onset cancers has been rising over the years, suggesting a pathogenic role of risk factor exposures during early life and young adulthood, given the long latency of carcinogenesis [97].

In our review, we observed that findings are significant even after adjustment for both age at menarche and at first childbirth, suggesting that dietary factors in adolescence could be an independent contributor to breast cancers later in life. Fruit, vegetables, and soy are food groups that make up part of a high-quality diet. However, adolescence is a life stage where poor diet quality can be common, and this often results in compromised intake of nutrient-dense foods and consequently, essential nutrients [98]. On the other hand, soy foods such as tofu, edamame, soy milk, etc., may be an integral part of the Asian diet but not in Western countries, much less in the early years. Moreover, our findings revealed that despite a lower fiber intake than their adult counterparts, adolescents may still have a lower breast cancer risk. Similarly, this is also the case for soy consumption in adolescence, where a substantially lower risk was observed when compared with meta-analyses on adult females. These findings draw attention to the need for further research and the initiation of public health efforts early in life, sustained across the life course, particularly during key windows of susceptibility when the breast tissue is more vulnerable to carcinogenic exposures.

Strengths, limitations, and gaps in adolescent research

The present systematic review and meta-analysis is the first to have examined the associations between adolescent diet and breast cancer in adulthood. We comprehensively considered all aspects of dietary intake, including individual nutrients, foods, food groups, and dietary patterns, to provide an extensive and up-to-date review of current available literature. Although the transformation of study-specific effect estimates to a standard scale of tertile enabled fair comparisons across studies, variation in how dietary intake was assessed, whether by frequency, serving size, or grammage, limited our ability to recommend a specific recommended intake range based on existing studies. Nonetheless, we also performed various sensitivity analyses to assess the robustness of our findings.

There are limitations inherent to our review. First, our meta-analysis was limited by the small number of available studies, resulting in lower statistical power compared with existing meta-analyses on adult females and limiting our ability to conduct a dose–response analysis. Consequently, several dietary variables, including dietary patterns, could only be summarized narratively. Although dietary patterns offer insight into the synergy between foods and nutrients, those narratively summarized in our review did not show any significant associations with breast cancer. This may be attributed to adolescence as a transitional period, where dietary habits are constantly evolving, unlike in adulthood when dietary patterns tend to be more stable [99], supporting the continued focus on individual food groups and nutrients in adolescent research. Moreover, we were unable to perform subgroup and meta-regression analyses to identify the potential sources of heterogeneity or evaluate differences in associations by molecular subtypes of breast cancer and menopausal status due to the limited number of studies. Second, less than half of the studies were of high quality (≥7 of 9 stars), primarily due to not adjusting for key covariates that we identified a priori and low follow-up rates in most studies. Nevertheless, our results remained consistent in the sensitivity analysis when we restricted the comparison to high-quality studies. In addition, although we pooled estimates from the most fully adjusted models, residual confounding cannot be excluded. Third, as only observational studies were included, we cannot establish a causal association with these data. Fourth, the reliance on adult recall of adolescent diet may introduce recall errors as participants might have difficulty remembering their dietary intake from ≥1 to 5 decades ago. This issue is particularly concerning in case-control studies, where differential misclassification may have biased results either toward or away from the null. Although cohort studies are generally preferred as they minimize recall bias, most existing cohorts outside the United States focus on adult populations. Moreover, few longitudinal studies track participants from adolescence into adulthood, largely due to time and cost constraints. Initiating cohorts earlier in life or extending follow-up in existing birth cohorts beyond adolescence could offer valuable insights. Fifth, our pooled analysis of soy intake, which comprised 3 case-control studies among participants of Asian descent [66,67,70], also highlights a research gap in whether the observed associations can be consistently replicated in Western populations where soy consumption is typically lower [100]. It is also worth noting that most studies in this review were conducted in the United States, potentially limiting the generalizability of our findings and underscores the need for more research in Asian populations to better understand and address the potential disparities in breast cancer risk. Lastly, the biological mechanisms underlying the relationship between dietary factors and proliferative BBD remain an important research gap, given its role as an independent risk factor for breast cancer. Elucidating these mechanisms could provide valuable insights to inform future targeted risk reduction strategies.

In conclusion, evidence from this systematic review and meta-analysis demonstrated that high intakes of fruits and vegetables, soy, dietary fiber, and vegetable fat during adolescence were associated with lower risks of breast cancer. Additionally, high intakes of dietary fiber and vitamin D during adolescence were associated with lower risks of BBD. Dietary and lifestyle changes often occur during the transition from adolescence to adulthood; therefore, well-designed prospective life course epidemiological studies are essential to substantiate these findings on early life exposures and breast cancer risk. Given the rapidly rising burden of breast cancer, there is an urgent need to raise awareness on the role of adolescent diet and timing of exposures, and to support continued high-quality research during critical developmental periods. Strengthening this evidence is key to better inform effective and timely prevention strategies aimed at reducing the future risk of cancer. As risk factors may vary by molecular breast cancer subtype [35,37,45,51,61,69,71] and menopausal status [32,33,45,60,61], it will be meaningful to explore these associations in future research.

Author contributions

The authors’ responsibilities were as follows – AC, MF-FC: contributed to study conception and design; CKYL, GHL, EL, NK: conducted the literature search, data extraction, and risk of bias assessment; AC, CKYL, GHL, EL: performed statistical analysis and data interpretation; YCRST: contributed expert advice; GHL: drafted the manuscript; and all authors: read, critically revised, and approved the final manuscript.

Data availability

Data described in the manuscript, code book, and analytic code will be made available on request pending application and approval.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work the authors used ChatGPT (OpenAI) in order to improve readability. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Funding

This work was supported by the Population Health Metrics and Analytics Programme, National University of Singapore and National University Health System, Singapore.

Conflicts of interest

The authors report no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: