Height and Body Size in Childhood, Adolescence, and Young Adulthood and Breast Cancer Risk According to Molecular Subtype in the Nurses' Health Studies

Body size in childhood and adolescence, body mass index (BMI) in young adulthood, and height are established risk factors for pre- and postmenopausal breast cancer (1–3). However, breast cancer is a heterogeneous disease with multiple subtypes of differing prognosis defined by tumor characteristics such as estrogen receptor (ER) and progesterone receptor (PR) status (4, 5). Inverse associations between childhood and adolescent body fatness, BMI in young adulthood and breast cancer have been observed for estrogen receptor–positive (ER⁺; refs. 1, 2) and negative (ER⁻; ref. 1) breast cancer. Height is positively associated with ER⁺ tumors, whereas associations with ER⁻ tumors have been inconsistent, with both positive (6), and null associations (7) observed. Beyond ER and PR, molecular subtypes defined by gene expression or immunohistochemical markers have been explored with respect to breast cancer etiology and prognosis (8–10) and their relationships with risk factors differ (11). Molecular subtypes include luminal A and B, human epidermal growth factor receptor 2 (HER2)-enriched, basal-like and unclassified cancers (4, 12, 13). Few studies have examined associations between body fatness in childhood and adolescence, BMI in young adulthood, and height, and molecular subtypes (7, 14).

Animal data, epidemiologic studies including examinations of effects of radiation exposure on breast cancer risk (15, 16), and risk prediction models (17, 18), have shown that breast tissue is particularly susceptible to exposures in early life, with the period between menarche and first birth most vulnerable (19). This is because the mammary gland goes through extensive morphological changes during early life. Ducts that were developed before birth grow and branch rapidly as a result of hormonal stimulation with final differentiation achieved during pregnancy and lactation (20, 21). The consistency of association between early life and young adulthood body fatness and breast cancer risk across strata of age and menopausal status suggests that greater body fatness at young ages may be associated with permanent changes to breast tissue during this important development period that results in a long-term reduction in breast cancer risk. Similarly, adult height is attained during adolescence or young adulthood and may reflect the concentration of growth factors during that phase of life and beyond (22).

We prospectively examined the association between body fatness in childhood and adolescence, BMI at the age of 18, height, and incidence of breast cancer according to molecular subtype in a pooled analysis of women in the Nurses' Health Study (NHS) and the Nurses' Health Study II (NHSII).

NHS and NHSII are two ongoing prospective cohort studies of primarily white (>95%) female registered nurses across the United States. NHS began in 1976 with 121,701 female registered between the ages of 30 and 55 years at baseline. NHSII began in 1989 with 116,430 female registered nurses ages 25 to 42 years. Women are sent follow-up questionnaires every 2 years to obtain information about health behaviors, disease status, medical care, and treatment. Cumulative follow-up rates are high (>90%) in both cohorts.

Women self-reported height in 1976 (NHS) or 1989 (NHSII) and weight at age 18 in 1980 (NHS) or 1989 (NHSII). Using records from physical examinations conducted at college or nursing school entrance from 118 NHSII participants, the validity of recalled weight at age 18 and self-reported current height was previously assessed (23). On average, participants underreported weight at age 18 by 1.4 kg. The correlation between recalled and measured weight at age 18 was 0.87.

In 1988 (NHS) and 1989 (NHSII), women were asked to select the figure from a validated 9-figure drawing which best corresponded to their body fatness at age 5, 10, and 20 (24, 25). The Third Harvard Growth Study compared women's recalled figure at 15 with measured BMI at age 15. They found that BMI at age 15 increased linearly (19 to 32 kg/m²) with each level of the figure (levels 1–7; no participants selected figure 8 or 9; ref. 26). We averaged somatotypes across 2 ages to create measures for childhood (ages 5 and 10) and adolescent (ages 10 and 20) body fatness. Because of sparse data, we collapsed the top 5 categories of each measure creating a top category that included anyone that selected figure 5 through 9.

Incident breast cancer diagnoses on each biennial questionnaire are, with participant or next of kin permission, confirmed through medical record review. From each woman, we request pathology reports and abstract information on tumor characteristics including grade, histologic type, metastases, and hormone receptor status. Pathology reports were available for >95% of the breast cancer cases in this study. Cases among deceased nonrespondents are identified via review of death certificates and medical records. Nearly all (99%) of self-reported breast cancers are confirmed after medical record review. In-situ cases were censored at date of diagnosis and only invasive cases were included in the analysis.

Tissue block collection and tissue microarray (TMA) construction were described in detail previously (27, 28). In brief, we obtained archived formalin-fixed, paraffin-embedded tissue blocks for approximately 70% of incident primary breast cancer cases from 1976 to 2006. Women with available tissue blocks were similar with respect to breast cancer risk factors and tumor characteristics compared with the women without available tissue blocks (27, 29). In brief, hematoxylin and eosin sections from cases with pathology samples were reviewed to confirm the diagnosis, classify the histologic type and grade of the breast cancer, and identify the area from which the TMA cores would be taken. TMAs contained three 0.6-mm diameter cores from each breast cancer sample. Immunostaining was performed on 5-μm paraffin sections cut from TMA blocks. Antibodies used for ER, PR, HER2, cytokeratin 5/6 (CK5/6), and EGFR staining are described elsewhere (11). A pathologist manually assessed ER, PR, HER2, CK5/6, and EGFR expression on each available core. We used the grade assigned by study pathologists except when not available, in which case we used grade collected from the participant's pathology report. A case was considered ER-positive (ER+) or PR-positive (PR+) if any of their tissue cores showed any nuclear staining for ER or PR, respectively. A case was deemed ER⁻ and/or PR⁻ if there was complete absence of staining for ER and/or PR in all tissue cores. HER2 protein overexpression was defined as moderate or strong membrane staining (2+ or 3+) in more than 10% of the cells in any of the tissue cores. Cases were considered CK5/6⁺ or EGFR⁺ if any cytoplasmic and/or membranous staining was detected in the tumor cells in any of the cores. Luminal A tumors were defined as ER⁺ and/or PR⁺, with no HER2 overexpression and grade 1 (low) or grade 2 (intermediate). Luminal B tumors were either: (i) ER⁺ and/or PR⁺ and HER2-overexpressed or (ii) ER⁺ and/or PR⁺ and did not overexpress HER2 and grade 3 (high grade). In secondary analyses, we looked separately at luminal B tumors that over expressed HER2 and those that were high grade. Cases that were ER⁻, PR⁻, HER2-overexpressed were classified as HER2-enriched. Basal-like cases were ER⁻, PR⁻, did not overexpress HER2, and were positive for CK5/6 and/or EGFR. Unclassified tumors lacked expression of all 5 markers. As many studies lack information on CK5/6 and EGFR needed to define the basal-like subtype, we present estimates for triple-negative (ER⁻, PR⁻, and HER2⁻) breast cancer as well (Supplementary Table S3).

As each exposure was assessed in different survey cycles, and we excluded women missing data on the primary exposure from each analysis, the analytic sample and length of follow-up differ for each exposure. Sample sizes are as follows: height (n = 233,214); BMI at age 18 (n = 207,490); childhood and adolescent somatotype (n = 188,689). Women stopped contributing person-time when they reported a breast cancer diagnosis, reported a diagnosis of any other cancer (excluding non-melanoma skin cancer), date of death or the study cutoff date June 1, 2006, whichever occurred first. Follow-up continued through 2006, the most recent year tissue data were available. Our analyses included a maximum of 233,214 women contributing 6,129,327 person-years of follow-up. Through the study cutoff date, there were the following numbers of cases: luminal A (n = 2,983), luminal B (n = 1,281), HER2-enriched (n = 316), basal-like (n = 408), and unclassified (n = 128).

Cox proportional hazards regression models were used to estimate incidence rate ratios and 95% confidence intervals (CI) for breast cancer subtypes associated with adult height (≤62, 63, 64, 65, 66, ≥67 inches), BMI at age 18 (<18.5, 18.5–20, 20–21.9, 22–24.9, ≥25 kg/m²), and childhood and adolescent somatotype (1, 1.5–2, 2.5–3, 3.5–4.5, ≥5). Age-adjusted and multivariate adjusted models are presented. Multivariate models include covariates selected on the basis of their association with breast cancer in previous analyses. In secondary analyses (Supplementary Tables S1 and S2), we present estimates for each exposure stratified by menopausal status (pre- or postmenopausal). To evaluate the consistency of risk estimates across molecular subtypes, we performed a competing risks analysis to estimate separate associations of each exposure with the relative hazard of each subtype (30–32). Analyses were performed using SAS version 9.3 (SAS). All statistical tests were 2-sided, and 0.05 was the threshold for statistical significance. The Institutional Review Board at Brigham and Women's Hospital approved this investigation.

Women who were taller had higher alcohol intake, lower BMI at age 18, lower current BMI, greater weight gain since age 18, older age at menarche, and higher birth weight (Table 1). Greater adolescent somatotype was associated with earlier age at menarche, higher BMI at age 18 and current BMI, and lower prevalence of benign breast disease (Table 2).

Height was positively associated with all molecular subtypes (P_{heterogeneity} = 0.78; Table 3). Women who were ≥67 inches tall, compared to those ≤62 inches tall, were 52% more likely to develop luminal A tumors (HR, 1.52; 95% CI, 1.34–1.73; P_trend: <0.0001), 48% more likely to develop HER2-enriched tumors (HR, 1.48; 95% CI, 1.02–2.15; P_trend: 0.004), and 2 times more likely to develop unclassified tumors (HR, 2.00; 95% CI, 1.08–3.70; P_trend: 0.02). Height was positively associated with all subtypes in pre- and postmenopausal women, although associations were generally stronger in premenopausal women (Supplementary Table S1). For example, for luminal A tumors, each one inch increase in height was associated with a 8% increase in risk for premenopausal women and a 5% increase among postmenopausal women.

BMI at age 18 was inversely associated with all molecular subtypes, although associations were strongest for HER2-enriched and basal-like (P_{heterogeneity} = 0.001; Table 3). Compared with women with a BMI at age 18 or 20 to 21.9 kg/m², those with a BMI at age 18 of ≥25 kg/m² were between 52% (HER2-enriched; P_trend < 0.0001) and 21% (luminal A; P_trend < 0.0001) less likely to develop breast cancer. Each one unit increase in BMI at age 18 was associated with a 2% decrease in risk of luminal B tumors overall; however, this varied between those that overexpressed HER2 (3% per kg/m²) versus those that were high grade (no association; data not shown). Associations were generally stronger for pre- versus postmenopausal women, although we observed inverse associations in both groups (Supplementary Table S2).

Overall, we observed inverse associations between childhood (P_{heterogeneity} = 0.51) and adolescent (P_{heterogeneity} = 0.046) somatotype and most subtypes (Table 4). Compared with lowest category of childhood somatotype, risk reductions for women in the highest category ranged from 23% (luminal A) to 65% (basal-like). For adolescent somatotype, the same contrast was associated with from a 20% reduced risk for luminal A tumors up to a 71% decrease for HER2-enriched tumors. There was no association with unclassified tumors for either childhood or adolescent somatotype. Inverse associations were generally stronger for premenopausal women than postmenopausal women; however, there were small numbers of cases particularly for ER⁻ subtypes among premenopausal women (Supplementary Table S2).

In this prospective pooled analysis of two large cohorts, we observed significant positive associations between adult height and risk of all breast cancer subtypes and inverse associations between BMI at age 18 and childhood and adolescent somatotype and risk of most subtypes. We observed heterogeneity in associations across subtypes for BMI at age 18 and adolescent somatotype. For BMI at age 18, there was a strong inverse association for HER2-enriched and basal-like tumors. The strongest inverse associations with body fatness in adolescence were observed among HER2-enriched, basal-like, and, to a lesser extent, luminal B tumors. In addition to differences by subtype, we also found that compared with postmenopausal women, associations were similar, but generally stronger among premenopausal women.

Consistent with previous studies, we found that greater adult height was associated with an increased risk of luminal tumors, but we also observed strong positive associations between height and risk of HER2-enriched, basal-like, and unclassified tumors, which has not been consistently seen in other studies (6, 7). A recent meta-analysis of 159 prospective cohorts that included 5,216,302 women and 9,732 breast cancer cases with known ER status, found a positive association between height and risk of ER⁺ tumors (HR, 1.18; 95% CI, 1.13–1.23 per 10-cm increase), but no association with ER⁻ tumors (HR, 1.00; 95% CI, 0.86–1.14). Interestingly, when tumors were classified jointly by ER and PR status, there was a borderline significant association with ER⁻/PR⁻ tumors (HR, 1.08; 95% CI, 0.99–1.18; ref. 33). They also conducted a Mendelian randomization analysis using a genetic risk score derived from 168 height-associated variants and found strong positive associations with ER⁺ and ER⁺/PR⁺ tumors and no association with ER⁻ or ER⁻/PR⁻ tumors. However, this study comes with several caveats. There was significant heterogeneity across studies overall (P_{heterogeneity} < .001, I² = 61%) and ER status was only available on 8.7% (9,792 of 113,178) of cases. Importantly, determination of ER and PR status was based on pathology reports and cases were diagnosed over a wide period (late 1960s through 2000s) in many different countries. A major strength of our study is uniform staining and scoring of immunohistochemical markers. In addition, the genetic risk score used Zhang and colleagues' Mendelian randomization analysis only explained 10% of the variation in height among people of European ancestry. The authors do not draw conclusions regarding the association of height with breast cancer risk according to ER/PR status. Instead, they conclude that their results demonstrate that height is causally implicated in breast cancer etiology and that height and breast cancer share genetic risk variants (33).

We found stronger inverse associations between our adolescent and young adulthood body size measures and risk of HER2-enriched and basal-like tumors than with luminal tumors. We also found stronger inverse associations with luminal B tumors that overexpressed HER2 compared with those that were high grade. This builds upon previous work in our cohorts that found stronger inverse associations with ER⁻ tumors and a suggestion of stronger associations with tumors that overexpressed HER2 (1). Two other studies also found that the inverse association between childhood body size and breast cancer was strongest for triple-negative tumors and those that overexpressed HER2 (34). While associations have also been observed only among ER⁺/PR⁺ (2) or ER⁺/PR⁻ tumors (35); overall, the body of literature suggests that body size in early life may reduce risk of breast cancer in adulthood through sex hormone–dependent and -independent mechanisms. It was previously postulated that higher body fatness in early life may be associated with higher levels of estrogen which could induce early breast differentiation thereby making breast cells less susceptible to malignant transformation (36). However, in girls age 8–10, higher BMI was associated with higher levels of dehydroepiandrosterone sulfate and sex hormone–binding globulin; there was no difference in circulating levels of estrogen or progesterone (37). It is possible; however, that early life body size affects breast differentiation through a different mechanism.

There are several possible mechanisms through which height and early life body size may impact breast cancer risk. Greater body fatness in childhood, adolescence, and young adulthood may reflect slower growth and lower IGF1 levels, which are associated with lower breast cancer risk (38, 39). IGF1 regulates growth beginning in utero and throughout childhood and adolescence (40). It also induces endothelial growth factor, promotes tumor growth, and inhibits apoptosis (41). Genome-wide association studies have demonstrated that genes in the IGF signaling pathway are also associated with adult height (42, 43). Rapid growth in childhood and adolescence is associated with increased risk of breast cancer (39, 44), independent of adult height. In our data, childhood and adolescent somatotype were not associated with adult height, whereas there was a weak inverse association between height and BMI at age 18. Age at attained height was also not associated with breast cancer risk (45) in NHS. Women with higher birth weight and higher BMI at age 18 had lower circulating IGF1 levels in adulthood compared with women who were leaner at early ages (46). Body fatness during childhood and adolescence has also been associated with lower premenopausal breast density (47). Thus, height and body size in early life and young adulthood may reflect the concentration of growth factors during that phase of life, which then has long-term health effects via growth factors in adulthood and breast tissue composition.

This study's strengths include its large size, prospective design, detailed and repeated assessment of known breast cancer risk factors, long follow-up, and comprehensive case ascertainment. Although the somatotype pictogram and BMI at age 18 have been validated (26), the measures rely on participant recall, and there is potential for misclassification. The somatotype figure has also been criticized because, while it queries about body size in childhood and adolescence, the pictures presented are of adult women (48). Yet, because the analysis is prospective, any potential misclassification is likely to be nondifferential with respect to disease status and if anything leads to an underestimation of association. In addition, the somatotype figure has been consistently associated with breast cancer and other outcomes (49–51). It is also important to note that studies with measured height and weight in adolescence have also observed significant inverse associations, albeit of lesser magnitude (44). Small case numbers among the rare subtypes (e.g., HER2-enriched, basal-like, and unclassified) limited our power and introduced some instability in effect estimates. In the future, pooled studies should assess associations with these less common breast cancer subtypes. Our study population was predominately lean in childhood and young adulthood and white. The prevalence of overweight and obesity has increased over time, and the distribution of body size in this study is not representative of today's population (52). In addition, it is possible that the mechanisms for larger body size are different today than they were for earlier birth cohorts included in this analysis. Finally, it is important to study these associations in non-White populations where the distribution of both early life body size and molecular subtype is different than observed here (52, 53).

In conclusion, we found that height was positively associated with risk of all breast cancer molecular subtypes, whereas childhood and adolescent somatotype and BMI at age 18 were inversely associated with risk of most subtypes. Adolescent and young adult body sizes were more strongly associated with HER2-enriched and, to a lesser extent, basal-like tumors, than other subtypes. Despite the consistent inverse associations observed between early life body fatness, BMI at age 18, and breast cancer risk, there are many well-known negative health consequences of overweight and obesity throughout the life course. However, these results provide additional evidence of the importance of early life exposures in breast cancer etiology. The observed associations across molecular subtypes suggest that these factors act either through multiple pathways or through a common nonhormonal or HER2-driven pathway to influence breast cancer risk in adulthood.

No potential conflicts of interest were disclosed.

Conception and design: E.T. Warner, L.C. Collins, W.C. Willett, R.M. Tamimi

Development of methodology: E.T. Warner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.H. Beck, S.J. Schnitt, K.B. Michels, W.C. Willett, R.M. Tamimi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.T. Warner, R. Hu, L.C. Collins, S.J. Schnitt, B. Rosner, A.H. Eliassen, W.C. Willett, R.M. Tamimi

Writing, review, and/or revision of the manuscript: E.T. Warner, L.C. Collins, A.H. Beck, S.J. Schnitt, B. Rosner, A.H. Eliassen, K.B. Michels, W.C. Willett, R.M. Tamimi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.T. Warner, R. Hu, A.H. Eliassen, R.M. Tamimi

Study supervision: E.T. Warner, R.M. Tamimi

We would like to thank the participants and staff of the Nurses' Health Study for their valuable contributions as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. The authors assume full responsibility for analyses and interpretation of these data.

This project was supported National Cancer Institute (NCI) grants P01 CA87969 and UM1 CA186107 and UM1 CA176726. E.T. Warner was supported by National Cancer Institute grant K01 CA188075.

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