Introduction
There is considerable evidence that childhood anthropometric measurements are associated with postmenopausal breast cancer risk. It has been consistently shown that variables that approximate body shape and size early in life are inversely associated with breast cancer risk in adulthood. For example, a study conducted in 1998 on the same data set as used in the current study [
1] reported that a larger somatotype at age seven years was associated with a lower postmenopausal breast cancer risk. Likewise, Hilakivi-Clarke and colleagues [
2] found that a shorter height and higher body mass in girls from age 7 to 15 years were associated with a decreased incidence of breast cancer. Berkey and colleagues [
3] also found extremely lean body mass at age 10 years to be associated with elevated breast cancer risk. In another study performed in 141,393 Danish girls, a high childhood body mass index (BMI) at age 14 years was shown to be protective against breast cancer later on in life [
4]. In addition, a study performed on the large Nurses' Health Study dataset concluded that average body fatness between the ages of 5 and 10 years are inversely associated with mammographic density [
5], which is generally considered to be an intermediate phenotype of breast cancer [
6].
Although a role of childhood body size in adult breast cancer risk has been established, less is known about its influence on tumour characteristics. One study by Bardia and colleagues [
7] looked into the risk of developing postmenopausal breast cancer stratified by estrogen receptor (ER) and progesterone receptor (PR) subtypes and reported that an increase in weight at age 12 years was associated with a decrease in adult breast cancer risk, with the most pronounced effects exhibited by ER-positive/PR-negative tumours. No significant heterogeneity, however, was observed between the tumour subtypes studied. To our knowledge, no other study has been conducted to assess whether pre-/peri-pubertal measurements of body size can also influence tumour characteristics. We thus followed up on the work of Bardia and colleagues and in the present study examined the relations between childhood body size to address if the far-reaching effects of childhood body size have any influence on tumour characteristics in adult cancers.
Results
Table S1 in Additional file
1 describes the characteristics of study subjects with respect to several breast cancer risk factors. Age of menarche was weakly but positively associated with the disease (OR per year increase in age of menarche = 0.96, 95% CI = 0.93 to 1.00,
P = 0.057), a result consistent with the literature [
4]. Family history, age at menopause, parity, age of first birth, benign breast disease, mammographic density, recent BMI and use of HRT were strongly significant for breast cancer risk with effects in a direction consistent with those estimated in other epidemiological studies. The first association analyses we performed between somatotypes at different ages and breast cancer risk were adjusted for age at enrolment only. Among the different measurements of somatotypes, only the time point at age seven years was found to affect breast cancer risk (OR per increase in somatotype class = 0.87, 95% CI = 0.8 to 0.95,
P = 0.001). A larger proportion of cases than controls had a leaner body shape at age seven years. Despite somatotype one year prior to enrolment having a high correlation to recent BMI (Spearman correlation coefficient: 0.760, data not shown), it was not found to be significantly associated with breast cancer (OR per increase in somatotype class = 1.04, 95% CI = 0.94 to 1.15,
P = 0.160).
To identify potential confounders of the association between somatotype at age seven years and breast cancer risk, we assessed whether other established risk factors for breast cancer are associated with somatotype at age seven years. An increase in childhood body size was found to exhibit strong inverse associations with age of menarche (OR comparing large to lean somatotype at age seven years = 0.61, 95% CI = 0.50 to 0.76,
P trend < 0.0001), benign breast disease (0.47, 95% CI = 0.25 to 0.89,
P trend = 0.006), and mammographic density (0.61, 95% CI = 0.41 to 0.90,
P trend = 0.001; Table
1). Associations in the opposite direction were found for proxy measures of physique at other time points, such as birth weight (OR comparing birthweight >4000 g to ≤2500 g = 1.89, 95% CI = 0.95 to 3.76,
P trend = 0.014), somatotype one year prior to enrolment (OR comparing large to lean somatotype at age seven years = 2.33, 95% CI = 1.70 to 3.18,
P trend < 0.0001) and recent BMI (2.66, 95% CI = 1.47 to 4.83,
P trend < 0.0001). No evidence of association was found between age of menopause and somatotype at age seven years or between family history and somatotype at age seven years. Parity and HRT were found to be independent of somatotype at age seven years (0.93, 95% CI = 0.76 to 1.13,
P trend = 0.217 and 0.98, 95% CI = 0.73 to 1.32,
P trend = 0.868, respectively).
Table 1
Associations of somatotype at age seven years with other breast cancer risk factors (controls only)
Age of menarche (years) | Lean | 1456 | 1.00 | reference | <0.0001 |
| Medium | 669 | 0.72 | 0.64 | 0.82 | |
| Large | 187 | 0.61 | 0.50 | 0.76 | |
Age of menopause (years) | Lean | 1572 | 1.00 | reference | 0.697 |
| Medium | 736 | 1.19 | 0.85 | 1.68 | |
| Large | 204 | 0.93 | 0.53 | 1.65 | |
Parity (Number of live births) | Lean | 1578 | 1.00 | reference | 0.217 |
| Medium | 745 | 0.93 | 0.83 | 1.05 | |
| Large | 207 | 0.93 | 0.76 | 1.13 | |
Benign breast disease | Lean | 1578 | 1.00 | reference | 0.006 |
| Medium | 745 | 0.76 | 0.56 | 1.03 | |
| Large | 207 | 0.47 | 0.25 | 0.89 | |
Somatotype one year prior to enrolment | Lean | 1571 | 1.00 | reference | <0.0001 |
| Medium | 739 | 1.72 | 1.44 | 2.05 | |
| Large | 206 | 2.33 | 1.70 | 3.18 | |
BMI (kg/m2) | Lean | 1562 | 1.00 | reference | <0.0001 |
| Medium | 742 | 1.85 | 1.30 | 2.65 | |
| Large | 205 | 2.66 | 1.47 | 4.83 | |
Percent mammographic density (%)† | Lean | 862 | 1.00 | reference | 0.001 |
| Medium | 428 | 0.72 | 0.58 | 0.91 | |
| Large | 108 | 0.61 | 0.41 | 0.90 | |
HRT | Lean | 1569 | 1.00 | reference | 0.868 |
| Medium | 739 | 0.99 | 0.83 | 1.18 | |
| Large | 206 | 0.98 | 0.73 | 1.32 | |
| Other independent variables | | | | | |
Birthweight (g) on somatotype at age 7 | ≤2500 | 49 | 1.00 | reference | 0.014 |
| 2500-3000 | 229 | 1.18 | 0.61 | 2.29 | |
| 3000-3500 | 470 | 1.29 | 0.68 | 2.43 | |
| 3500-4000 | 397 | 1.44 | 0.76 | 2.73 | |
| >4000 | 135 | 1.89 | 0.95 | 3.76 | |
Family history on somatotype at age 7 | No | 2258 | 1.00 | reference | 0.485 |
| Yes | 227 | 1.10 | 0.84 | 1.44 | |
After adjustment of known breast cancer predictors and other associated risk factors, the inverse association of somatotype at age seven years with breast cancer remained highly significant (Table
2; OR comparing large to lean somatotype at age seven years = 0.73, 95% CI = 0.58 to 0.91,
P trend = 0.004). The protective effect of a larger somatotype was found to be significant (
P trend < 0.05) for ER-negative, PR-positive and PR-negative subtypes and marginally significant (
P trend = 0.062) for the ER-positive subtype. Within the group consisting of large somatotypes, the most prominent effects were shown in ER-negative (OR comparing large to lean somatotype at age seven years = 0.40, 95% CI = 0.21 to 0.75,
P trend = 0.002) and PR-negative (0.63, 95% CI = 0.40 to 0.99,
P trend = 0.028) tumours. The point estimates changed very little before and after additional adjustment for mammographic density as a continuous variable [see Table S2 in Additional file
2], using a subset of the data with this information available (n = 3232).
Table 2
Multivariate-adjusted OR estimates and corresponding 95% CIs of postmenopausal breast cancer for somatotype at age seven years, overall and stratified by breast cancer tumour subtype based on ER and PR status
All data | Lean | 1784 | 1.00 | reference | 0.004 |
| Medium | 757 | 0.90 | 0.79 | 1.02 | |
| Large | 173 | 0.73 | 0.58 | 0.91 | |
ER positive | Lean | 963 | 1.00 | reference | 0.062 |
| Medium | 408 | 0.91 | 0.78 | 1.06 | |
| Large | 98 | 0.80 | 0.62 | 1.05 | |
ER negative | Lean | 219 | 1.00 | reference | 0.002 |
| Medium | 81 | 0.77 | 0.58 | 1.03 | |
| Large | 14 | 0.40 | 0.21 | 0.75 | |
PR positive | Lean | 841 | 1.00 | reference | 0.027 |
| Medium | 354 | 0.89 | 0.75 | 1.04 | |
| Large | 83 | 0.76 | 0.57 | 1.00 | |
PR negative | Lean | 320 | 1.00 | reference | 0.028 |
| Medium | 126 | 0.86 | 0.68 | 1.08 | |
| Large | 25 | 0.63 | 0.40 | 0.99 | |
We next assessed the effects of childhood body size on tumour characteristics (ER status, PR status, tumour size, grade, histology, and absence/presence of metastatic nodes) by fitting binary/ordinal logistic regression models, adjusting for age at diagnosis in years as a confounder. We established that the protective effect of somatotype at age seven years was significantly stronger for ER-negative disease than for ER-positive disease (
P heterogeneity = 0.046; Table
3). When comparing between two extreme groups, women with a larger body size at age seven years were 1.71 times (95% CI = 0.96 to 3.06) more likely to get ER-positive than ER-negative disease after menopause. Although the estimated trend suggests that women with the same physique are more likely to get the PR-positive disease in adulthood, the difference between the two tumour subtypes was not significant (
P heterogeneity = 0.283). The point estimates for tumour size, histology, grade, or the presence or absence of metastatic nodes did not vary much before and after adjustment for age of diagnosis as a continuous variable.
Table 3
Relation of somatotype at age seven years to tumour-defined characteristics of breast cancer
Tumour size (cm)* | <1 | 300 | 138 | 39 | |
| 1-2 | 752 | 299 | 70 | |
| 2-3 | 366 | 152 | 31 | |
| 3-4 | 116 | 66 | 14 | |
| 4-5 | 52 | 16 | 3 | |
| >=5 | 65 | 26 | 4 | |
Cumulative OR (95% CI) | | 1.00 (ref.) | 1.00 (0.85-1.17) | 0.78 (0.58-1.05) | 0.255 |
Cumulative OR (95% CI) § | | 1.00 (ref.) | 1.00 (0.85-1.18) | 0.78 (0.58-1.06) | 0.266 |
Grade* | Low | 159 | 69 | 20 | |
| Medium | 479 | 186 | 46 | |
| High | 463 | 222 | 51 | |
Cumulative OR (95% CI) | | 1.00 (ref.) | 1.15 (0.94-1.41) | 0.99 (0.69-1.43) | 0.443 |
Cumulative OR (95% CI) § | | 1.00 (ref.) | 1.15 (0.93-1.41) | 0.99 (0.69-1.42) | 0.463 |
Histology* | Ductal | 1350 | 570 | 137 | |
| Lobular | 206 | 77 | 16 | |
| All other | 92 | 37 | 7 | |
Cumulative OR (95% CI) | | 1.00 (ref.) | 0.91 (0.72-1.15) | 0.76 (0.48-1.20) | 0.192 |
Cumulative OR (95% CI) § | | 1.00 (ref.) | 0.92 (0.73-1.17) | 0.79 (0.50-1.25) | 0.265 |
Metastatic nodes† | Absent | 1159 | 473 | 116 | |
| Present | 513 | 227 | 46 | |
OR (95% CI) | | 1.00 (ref.) | 1.08 (0.90-1.31) | 0.90 (0.63-1.28) | 0.923 |
OR (95% CI) § | | 1.00 (ref.) | 1.07 (0.88-1.29) | 0.86 (0.60-1.23) | 0.878 |
ER status† | Negative | 219 | 81 | 14 | |
| Positive | 963 | 408 | 98 | |
OR (95% CI) | | 1.00 (ref.) | 1.15 (0.87-1.52) | 1.59 (0.89-2.84) | 0.089 |
OR (95% CI) § | | 1.00 (ref.) | 1.18 (0.89-1.56) | 1.71 (0.96-3.06) |
0.046
|
PR status† | Negative | 320 | 126 | 25 | |
| Positive | 841 | 354 | 83 | |
OR (95% CI) | | 1.00 (ref.) | 1.07 (0.84-1.36) | 1.26 (0.79-2.01) | 0.307 |
OR (95% CI) § | | 1.00 (ref.) | 1.07 (0.84-1.37) | 1.28 (0.80-2.03) | 0.283 |
Discussion
Our first main finding was that a large somatotype at age seven years was associated with a decreased risk of postmenopausal breast cancer. Although strongly associated with other risk factors such as age of menarche, adult BMI and mammographic density, somatotype at age seven years remained a significant protective factor (OR comparing large to lean somatotype at age seven years = 0.73, 95% CI = 0.58 to 0.91, P trend = 0.004) after adjustment for these other risk factors. Our second and most novel finding was of a significant protective effect of somatotype at age seven years regardless of receptor status, but with a stronger effect for ER-negative (0.40, 95% CI = 0.21 to 0.75, P trend = 0.002), than for ER-positive (0.80, 95% CI = 0.62 to 1.05, P trend = 0.062), tumours (P heterogeneity = 0.046).
Our findings regarding the protective effects of childhood body size for adult breast cancer are consistent with previous studies [
3‐
5]. Associations with other breast cancer risk factors were also in the same direction as found in other epidemiological studies. Several studies have found birth weight and gain in BMI in early childhood to predict adult lean mass, while adult adiposity has been attributed to weight gain in late childhood and adolescence [
15‐
19]. Similarly, anthropometric measurements at other time points (birth weight, and somatotype one year prior to enrolment) in our data were found to be positively associated with somatotype at age seven years. The adverse effects of birth weight and adult body mass on postmenopausal breast cancer risk may be explained by a surplus of estrogen exposure from either the uterine environment or excess adipose tissue [
4,
20]. However, studies performed on children have not consistently found an association between obesity and circulating estradiol levels [
21,
22], thus it is unclear what mechanisms drive the associated decrease in risk during the pre-/peri-puberty window.
Strong inverse relationships found between childhood body size, age of menarche, benign breast disease, and mammographic density were in line with other reports in the literature. Baer and colleagues [
23] found a large childhood body size to be associated with a decrease in risk of benign breast disease. Age of menarche is often considered along with age of menopause and other hormonal risk factors for a woman's cumulative exposure to estrogen [
24,
25]. An earlier age of menarche is associated with an increased risk of breast cancer. On the other hand, a larger childhood somatotype, which is associated with decreased breast cancer risk, is also associated with an earlier age of menarche. As age of menarche is an established but weak predictor of breast cancer risk, its pronounced inverse relationship with childhood body size when seen in the context of breast cancer risk seems to be counterintuitive [
26,
27].
Mammographic density has also been found by others to be associated with childhood body mass [
5]. Estrogen is produced by adipose tissue in the body. A higher BMI is thus correlated with higher endogenous estrogen levels. In a murine study, exposure to estrogen prior to puberty led to a decrease in radiologically dense tissue and an increase in the number of radiolucent structures [
28], which may be analogous to a lower mammographic density in humans. In agreement, McCormack and colleagues [
29] showed that high childhood BMI was associated with a lower Wolfe grade, and Samimi and colleagues [
5] found that a rounder pre-pubertal body shape was predictive of lower mammographic density later in life.
The age-adjusted case-only comparison of our data reflected a significant difference in the effects of childhood body size on the two ER subtypes (
P trend = 0.046), but not the PR subtypes. However, in lieu of the fact that PR is an estrogen-induced target gene, and that its presence could serve to indicate ER functional capacity and tumour differentiation state [
30], we also conducted stratified analyses on PR subtypes. We found that the protective trend conferred by a larger childhood somatotype on postmenopausal breast cancer applies to all ER and PR tumour subtypes. Overall our results were consistent with Bardia and colleagues [
7], although in that study the effects were only significant for ER-positive (0.80, 95% CI = 0.67 to 0.96) and PR-negative (0.62, 95% CI = 0.43 to 0.89) tumours (comparing women with above average weight at age 12 years to women with average weight at age 12 years). Although Bardia and colleagues observed a stronger protective effect in ER-negative tumours than in their ER-positive counterparts (in agreement with our finding) when comparing women with above average weight at age 12 years to women with average weight at age 12 years, the association they observed in this subgroup was not statistically significant (0.77, 95% CI = 0.5 to 1.19).
Hormonal exposure and mammographic density are established risk factors of breast cancer that have been suggested to be independent, operating through different pathways [
31]. Adjustment for these factors and other traditional risk factors did not attenuate the negative association of childhood body size on breast cancer risk (OR comparing large to lean somatotype at age seven years = 0.73, 95% CI = 0.58 to 0.91,
P trend = 0.004, for association, after adjustment), thus suggesting an independent underlying mechanism. We speculate that a possible mechanism driving the negative association with breast cancer risk could be epigenetic changes that occur during mammary development. Hilakivi-Clarke [
32] summarised in a review several perspectives on special windows of mammary development. Mammary tissue is postulated to undergo epigenetic extensive modelling or re-modelling during different stages in life such as fetal development, puberty or pregnancy. Such epigenetic modification can persist into adulthood if taken place in mammary stem cells, uncommitted mammary myoepithelial or luminal progenitor cells and inherited by subsequent daughter cells [
33]. Prepubertal exposure to estrogen has been shown to upregulate the expression of BRCA1, a well-known DNA repair gene [
28]. Liu and colleagues [
34] also demonstrated that BRCA1 is responsible for differentiating ER-negative stem/progenitor cells into ER-positive luminal cells. They also proposed that loss of expression of the DNA repair gene (BRCA1) may result in an accumulation of ER-negative stem cells with multiple genetic defects. Incidentally, loss of BRCA1 is frequently associated with ER-negative breast cancers [
35]. The evidence for altered gene expression possibly caused by childhood body size helps to explain the general reduction in breast cancer risk overall. The apparent differential protection conferred to the ER-negative subtype could possibly be driven by the same underlying mechanism that operates through epigenetic modifications.
The strengths of our study include it being a population-based study, its large sample size and detailed information on many variables: anthropometric measures at different time points throughout life, mammographic density, reproductive and hormonal risk factors, and tumour characteristics. To our knowledge, this is the first study to consider the effects of somatotype at age seven years on adult breast cancer with the consideration of mammographic density, and also the first to examine its effects on tumour characteristics other than ER status.
A limitation of our study is that risk factor data were self-reported, and could thus be measured with error. Although two studies have demonstrated the validity of using the nine-level somatotype diagram for the long-term recall of childhood body size via high correlations with BMI at the same ages [
8,
36], it is noteworthy that in those studies no woman recalled their figure as larger than level seven in these studies, and that women with large body size were more likely to misreport their childhood somatotypes than women who were lean. However, any such measurement error is most likely to attenuate any association between childhood body size and breast cancer risk [
37]. In addition, as the questionnaire study was conducted post-diagnosis of breast cancer, recall bias could have been introduced. Although the nine-level somatotype measure has not been validated specifically in a group of breast cancer cases, it is unlikely that childhood body size was differentially recalled by breast cancer cases and by controls.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
JLi participated in the study design, carried out the analyses and drafted the manuscript. LE digitised and obtained readings for the mammograms. KH, KC, JLiu and PH participated in study design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.