Size at birth and risk of breast cancer: Prospective population-based study

Authors

  • Lars J. Vatten,

    Corresponding author
    1. Department of Public Health and General Practice, Norwegian University of Science and Technology, Trondheim, Norway
    • Department of Public Health and General Practice, Norwegian University of Science and Technology, NO-7489 Trondheim, Norway
    Search for more papers by this author
    • Fax: +47-73-59-75-77

  • Tom I. Lund Nilsen,

    1. Department of Public Health and General Practice, Norwegian University of Science and Technology, Trondheim, Norway
    Search for more papers by this author
  • Steinar Tretli,

    1. Department of Public Health and General Practice, Norwegian University of Science and Technology, Trondheim, Norway
    2. Norwegian Cancer Registry, Oslo, Norway
    Search for more papers by this author
  • Dimitrios Trichopoulos,

    1. Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA
    Search for more papers by this author
  • Pål R. Romundstad

    1. Department of Public Health and General Practice, Norwegian University of Science and Technology, Trondheim, Norway
    2. Norwegian Cancer Registry, Oslo, Norway
    Search for more papers by this author

Abstract

It has been hypothesized that birth size is positively associated with breast cancer risk in adulthood. We studied birth length, birth weight and head circumference at birth and subsequent risk for breast cancer in a cohort of 16,016 women in Norway. Birth length was positively associated with risk (p trend = 002), and women who were 53 cm or longer had a relative risk of 1.8 (CI = 1.2–2.6) compared with women who were shorter than 50 cm, after adjustment for birth year, length of gestation, birth order, maternal age, maternal marital status and socioeconomic status at childbearing. Mutual adjustment for birth weight did not influence the results, and further adjustment for maternal height and adult factors (age at first birth and parity) in a subset of the cohort did not change the results. For birth weight, women in the highest category (≥ 3,840 g) had an adjusted relative risk (RR) of 1.5 (CI = 1.0–2.2) compared to women in the lowest (< 3,040 g), but mutual adjustment for birth length attenuated this association (RR = 1.1; CI = 0.7–1.8). Head circumference at birth showed a similar association as birth weight, with attenuation after mutual adjustment for birth length. The positive association with birth length was stronger among women whose mothers were relatively tall (median or taller, p trend = 0.001) compared to women whose mothers were relatively short (below median, p trend = 0.67) at childbearing. The results provide evidence that intrauterine factors influence future breast cancer risk. The positive association related to birth length suggests that factors that stimulate intrauterine longitudinal growth are particularly important. © 2004 Wiley-Liss, Inc.

The hypothesis that breast cancer may originate in utero implies that factors that stimulate intrauterine growth in female offspring also increase the risk for breast cancer in adulthood.1 Most studies have used birth weight as a marker for fetal growth to test this hypothesis. However, the results of 3 cohort studies,2, 3, 4 1 nested case-control study5 and 11 case-control studies6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 are not conclusive, but suggest that measures of birth size and fetal growth may be positively associated with breast cancer risk. Information on other markers, such as birth length or head circumference, has not been available in most studies, and information on other factors of importance for birth size, such as length of gestation, or adult risk factors with potentially confounding effect, such as age at first birth or parity, is typically not available in these studies and could not be taken into account.

In this study of a relatively large cohort of Norwegian women, we have assessed whether birth length, birth weight and head circumference at birth are positively associated with subsequent risk for breast cancer.

Material and methods

Study population

We have collected information from birth records kept at St. Olav's University Hospital in Trondheim, Norway, for the period 1920–1958. We abstracted information from all births that took place during that period, yielding information from 21,147 female offspring. We excluded 537 twins, leaving a total of 20,610 singletons. The regional committee for ethics in medical research approved the study.

In 1960, Norwegian citizens were assigned an individual 11-digit person number. For women, the Central Person Registry of Statistics Norway maintains continuously updated records on their residential and childbearing history, including their mother's name. In the birth archive, the daughters were registered by their mother's name. By combining the mother's name and the daughter's unique 11-digit number, we could identify women who were alive in 1960 and born at St. Olav's Hospital between 1920 and 1958. However, for women whose mothers had died before being assigned the person number, we could not be absolutely certain about the identity of the daughter. Also, most daughters changed their last name when they married, and if their mother had died before 1960, their identity could not always be verified. Thus, among a total of 20,610 female singletons, we reliably identified 16,016 singletons that were eligible for breast cancer follow-up by linkage to the Norwegian Cancer Registry. For women born before 1941, follow-up time was calculated from 1 January 1961 until the date of the first diagnosis of breast cancer or another cancer, emigration, death from other causes than cancer, or to the end of follow-up (31 December 2001), whichever occurred first. For women born in 1941 or later, similar follow-up started at the age of 20 years. Breast cancer was registered according to the international classification of diseases (ICD-7, code 170).

From the birth records of the eligible women, we abstracted information on birth weight (g), birth length (cm), head circumference (cm) and length of gestation (weeks or months). Maternal information that was recorded from the birth records included height and marital status. For mothers who were married at birth, their husband's occupation was used as a measure of socioeconomic status; if the mother was unmarried, her own occupation was used.

In addition, we collected information from the Central Person Registry on childbearing history in adulthood for cohort members born in 1930 or later and still residing in Trondheim (81.6%). Thus, age at first birth and parity, which are established factors associated with breast cancer risk, were included in this subset of the population.

Statistical methods

We categorized the continuous variables (birth weight, birth length and head circumference) into approximate quintiles. We estimated the effect of each birth characteristic on risk of breast cancer as incidence rate ratios (RRs) using Cox regression analysis. We also studied breast cancer risk separately for women before and after the age of 50 to distinguish approximately between premenopausal and postmenopausal breast cancer. Precision of the estimated effects was assessed by 95% confidence intervals (CIs). In the Cox regression analysis, we adjusted for potential confounding by year of birth (1920–1958) in 5-year birth categories. Further, we adjusted for length of gestation, birth order, maternal age at childbearing, marital status and maternal socioeconomic status. In a subset of women (74.3%) with sufficient information, we also adjusted for maternal height and the adult factors age at first birth and parity. In a stratified analysis, we studied the association between birth length and breast cancer risk in women whose mother was relatively tall (median or higher) at childbearing and in women whose mother was relatively short (below median). We tested the statistical significance of a possible interaction between offspring birth length and maternal height by including a product term of these variables in the regression model.

Results

We have followed 16,016 women born between 1920 and 1958 for breast cancer incidence from 1961 until the end of 2001. During follow-up, 312 women were diagnosed with breast cancer; 167 were diagnosed before 50 years of age and 145 were diagnosed among women who were 50 years or older. The relatively large proportion of young cases reflects that the cohort as a whole is still relatively young. Median age at diagnosis was 49.1 years (range, 25–76 years). During follow-up (1961–2001), 798 women were diagnosed with other cancers, after which they were censored from further follow-up. Also, 484 women who died from other causes than cancer were censored at time of death, and 325 who emigrated were censored at the time of emigration.

Our results show a modest positive association between birth weight and breast cancer risk (Table I). The association was not consistent across birth weight categories (p for trend = 0.14) but showed moderate increase in risk associated with the highest category of birth weight. Women whose birth weight was 3,840 g (highest quintile) or higher had an RR of 1.5 (CI = 1.0–2.2) compared with women whose birth weight was less than 3,040 g (lowest quintile) after adjustment for birth year, length of gestation, birth order, maternal age, marital status and socioeconomic circumstances at childbearing. Further adjustment for maternal height and the adult risk factors age at first birth and parity did not influence these results. However, mutual adjustment for birth length attenuated the association between birth weight and breast cancer risk. Thus, the RR comparing the highest and lowest quintile of birth weight was reduced to 1.1 (CI = 0.7–1.8) after additional adjustment for birth length (data not shown).

Table I. Relative Risk and 95% Confidence Interval for Breast Cancer Associated with Birth Size Characteristics Among 16,016 Women and 312 Cases of Breast-Cancer
 Number of casesNumber of womenAge-adjusted RR1Multivariable RR (95% CI)2p-trend3
  • 1

    Adjusted for year of birth (1920–1924, 1925–1929, …, 1955–1958).

  • 2

    Adjusted for year of birth (1920–1924, 1925–1929, …, 1955–1958), length of gestation (preterm, term, postterm), marital status (married, unmarried, widow, divorced), socioeconomic status (low, high), maternal age (<25, 25–29, 30–34, ≥35), birth order (1st, 2nd, 3rd, ≥4th).

  • 3

    Two-sided p-values for trend in Cox regression.

  • 4

    Women in the highest category of both birth weight and birth length defined as large, and those in the lowest category of both variables defined as small.

Birth weight (g)     
 <3,040502,9641.01.0 
 3,040–3,310743,1861.41.4 (1.0–2.1) 
 3,320–3,550513,2781.01.0 (0.6–1.5) 
 3,560–3,830673,2831.31.3 (0.9–1.9) 
 ≥3,840693,3001.41.5 (1.0–2.2)0.14
Birth length (cm)     
 <50874,0951.01.0 
 50783,7641.01.0 (0.7–1.4) 
 51553,0021.11.2 (0.8–1.7) 
 52412,7941.01.0 (0.7–1.5) 
 ≥53512,3581.81.8 (1.2–2.6)0.02
Head circumference (cm)     
 <34522,6601.01.0 
 34723,7681.01.0 (0.7–1.4) 
 35854,5201.01.0 (0.7–1.4) 
 36573,1841.01.0 (0.7–1.5) 
 ≥37451,8201.51.5 (1.0–2.2)0.14
Combined birth weight and birth length4     
 Small422,2671.01.0 
 Large371,5652.22.1 (1.2–3.6) 

For birth length, there was a significant positive trend across the 5 categories (p for trend = 0.02; Table I), but the estimates of relative risk suggest that the risk increase was restricted to the highest category. Women in the highest (≥ 53 cm) category of birth length had an RR of 1.8 (CI = 1.2–2.6) compared to the lowest (< 50 cm) category after adjustment for the potentially confounding factors mentioned above. For birth length, mutual adjustment for birth weight and head circumference did not alter the positive association with breast cancer risk (data not shown).

We found similar positive associations for head circumference, with higher risk restricted to the highest category of head circumference (Table I). Women in the highest quintile (≥ 37 cm) at birth had a multivariable adjusted RR of 1.5 (CI = 1.0–2.2) compared to women with a head circumference less than 34 cm. After mutual adjustment for birth length, this association was slightly attenuated (RR = 1.3; data not shown).

We also combined birth length and birth weight in order to compare women who were in the highest category for both factors and women who were in the lowest category for both birth length and weight (Table I). The results showed that women who were in the larger group at birth (37 cases among 1,565 women) had 2-fold higher risk (multivariable RR = 2.1; CI = 1.2–3.6) for breast cancer in adulthood compared with women in the smaller group (42 cases among 2267 women).

We also analyzed the data according to whether breast cancer was diagnosed before or after the age of 50 years to approximate pre- and postmenopausal disease (Table II). The results did not substantially differ between these categories and displayed similar patterns both in relation to birth weight, length and head circumference. Furthermore, mutual adjustment for each of the birth characteristics resulted in attenuation of the association for birth weight and head circumference, whereas the positive association with birth length remained unchanged both in pre- and postmenopausal women.

Table II. Relative Risk and 95% Confidence Interval for Breast Cancer Diagnosed Before (167 Cases) and After (145 Cases) the Age of 50 Years Associated with Birth Size Characteristics Among 16,016 Women
 < 50 years (premenopausal) multivariable RR (95% CI)1≥ 50 years (postmenopausal) multivariable RR (95% CI)1
  • 1

    Adjusted for year of birth (1920–1924, 1925–1929, …, 1955–1958), length of gestation (preterm, term, postterm), marital status (married, unmarried, widow, divorced), socioeconomic status (low, high), maternal age (<25, 25–29, 30–34, ≥35), birth order (1st, 2nd, 3rd, ≥4th) and each of the other variables.

Birth weight (g)  
 <3,0401.01.0
 3,040–3,3102.0 (1.1–3.5)0.9 (0.5–1.7)
 3,320–3,5501.0 (0.5–2.0)0.9 (0.5–1.8)
 3,560–3,8301.2 (0.6–2.5)1.2 (0.6–2.4)
 ≥3,8401.1 (0.5–2.4)1.1 (0.5–2.5)
Birth length (cm)  
 <501.01.0
 500.9 (0.5–1.4)1.2 (0.7–2.0)
 510.9 (0.5–1.6)1.4 (0.8–2.7)
 520.9 (0.5–1.7)1.0 (0.5–2.2)
 ≥531.5 (0.8–2.9)2.1 (0.9–5.0)
Head circumference (cm)  
 <341.01.0
 340.7 (0.4–1.2)1.2 (0.7–2.1)
 350.8 (0.5–1.4)0.9 (0.5–1.7)
 361.0 (0.6–1.7)0.7 (0.3–1.4)
 ≥371.1 (0.6–2.1)1.3 (0.6–2.8)

In supplementary analysis, we studied whether the association between birth length and breast cancer risk was modified by maternal height (Table III). Among women whose mother was relatively tall (163 cm and taller), the multivariable RR for those in the highest quintile of birth length (≥ 53 cm) was 2.8 (CI = 1.5–5.2). In this group, the association across categories of birth length displayed a strong positive trend (p for trend = 0.001). For women whose mother was shorter (< 163 cm), the association with birth length was weaker (p for trend = 0.67), and women in the highest quintile of birth length had a multivariable RR of 1.4 (CI = 0.8–2.7). The test for interaction between maternal height and offspring birth length in relation to breast cancer risk was of marginal statistical significance (p = 0.07).

Table III. Relative Risk for Breast Cancer Associated with Birth Length Stratified by the Median Maternal Height Among 14,109 Women and 247 Cases of Breast Cancer
Birth length (cm)Maternal height < 163 cmMaternal height ≥ 163 cm
Number of casesNumber of womenMultivariable RR (95% CI)1p trend2Number of casesNumber of womenMultivariable RR (95% CI)1p trend2
  • 1

    Adjusted for year of birth (1920–1924, 1925–1929, …, 1955–1958), length of gestation (preterm, term, postterm), marital status (married, unmarried, widow, divorced), socioeconomic status (low, high), maternal age (<25, 25–29, 30–34, ≥35), birth order (1st, 2nd, 3rd, ≥4th).

  • 2

    Two-sided p-values for test of trend in Cox regression. Test for interaction between birth length and maternal height (p = 0.07).

<50442,1921.0 181,2911.0 
50391,8381.0 (0.6–1.6) 211,4821.1 (0.6–2.1) 
51221,3421.0 (0.6–1.7) 251,4411.7 (0.9–3.1) 
52161,1540.8 (0.5–1.6) 211,4621.4 (0.7–2.8) 
≥53178181.4 (0.8–2.7)0.67351,4362.8 (1.5–5.2)0.001

Discussion

In this cohort of Norwegian women, we found that characteristics of large birth size were associated with increased risk for breast cancer in adulthood. We found a particularly striking association with birth length that remained strong after adjustment for potentially confounding variables, including adjustment for birth weight and head circumference.

Our findings did not substantially differ between women diagnosed before or after the age of 50 years. Thus, our results differ from those of a recent cohort study based on Swedish data that reported clear positive associations between markers for fetal growth and premenopausal breast cancer risk, but no evidence for any association for postmenopausal disease.3 On the other hand, a large Danish cohort study recently showed a positive association between birth weight and breast cancer risk, both for pre- and postmenopausal women,4 which is in accordance with our results.

The design of our study was similar to the Swedish3 study in employing information from a cohort of women born in one hospital and using the national cancer registry for cancer follow-up. Also, both studies showed that adjustment for variables that might influence risk (for example, length of gestation, maternal age at childbearing, age at first birth, parity and indicators of socioeconomic status) did not alter the associations between birth characteristics and breast cancer risk. The study using Swedish data was based on follow-up of 5,358 women born between 1915 and 1929, yielding 63 cases of premenopausal breast cancer and 295 postmenopausal cases. Our cohort consisted of 16,016 women born between 1920 and 1958 and yielded 167 cases diagnosed before the age of 50 and 145 diagnosed after that age. Both studies had virtually complete cancer follow-up due to linkage between the birth data and mandatory nationwide cancer registration using a unique national identity number. Factors that distinguish the studies include the larger and younger Norwegian cohort, with nearly 3 times more breast cancers diagnosed before the age of 50 years compared to the Swedish cohort.

Previous studies of the effect of birth size on breast cancer risk have reached varying results.2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 Initially, 2 other Swedish studies7, 9 also used birth size characteristics from population-based birth records but employed a case-control design. The first showed a positive but weak association between birth weight and breast cancer risk,7 but the second study,9 including data from 4 additional birth units, failed to confirm the initial results. In the prospective Nurses' Health Study,5 there was a positive association between birth weight and breast cancer risk that displayed a linear trend across birth weight categories. In that study, the mothers of the nurses provided information on birth size either from memory or from personal files, both being sources of error. In addition, a large proportion of the nurses' mothers were unable to respond to the request, resulting in missing information and loss to follow-up for this particular study question. In comparison to these studies, the major strengths of our study include the accurate measurements of birth size, adjustment for adult risk factors such as age at first birth and parity, as well as the complete follow-up for breast cancer incidence of a large unselected population.

The recent cohort study based on Swedish data also found a strong positive association between birth length and risk for premenopausal breast cancer, but not for postmenopausal disease.3 In most7, 9, 12 but not all16 previous studies, the positive association between birth length and breast cancer risk has been weak and not statistically significant. In our study, the association with birth length was not attenuated after adjustment for birth weight and head circumference, whereas the initial positive association with birth weight became weaker after adjustment for birth length. The consistency of this finding both before and after the age of 50 may strengthen the hypothesis that factors that stimulate intrauterine longitudinal growth may also influence future risk for breast cancer.

The positive association between birth length and overall risk for breast cancer may also reflect the general importance of growth, as indicated by the positive association between adult height and risk for breast cancer that has been reported in many studies,17 but also by the positive association between birth length and adolescent height.18 The less consistent association with birth weight, however, may reflect that birth weight combines the influence of longitudinal growth and weight gain.19 It has been repeatedly shown that body mass in adulthood exerts different effects on breast cancer risk, depending on the woman's menopausal status.17 Possibly, the dual association between adult body mass and breast cancer risk may be reflected in the less consistent association with breast cancer that we observed for birth weight than for birth length. Nonetheless, by combining birth length and birth weight in the analysis, we found that the positive association with large birth size was further strengthened.19

We also assessed whether maternal height could modify the association with birth length by analyzing birth length and breast cancer risk separately for offspring whose mothers were either taller or shorter than the median. The results showed a much stronger association with birth length if the mother was taller than the median height. The interpretation of this finding is not obvious, but it does suggest that a combination of genetic potential that stimulates longitudinal growth, as indicated by maternal height, and intrauterine growth of the offspring, as indicated by birth length, could be an important determinant for breast cancer risk in adulthood. This interpretation may be in accordance with the results of a British cohort study.2 That study also reported a positive association between birth size and breast cancer risk, but further analysis showed that the risk increase was restricted to women who were tall at the age of 7 years. Thus, it was suggested that the association with intrauterine growth could be modulated by childhood growth.2

Birth size measures are likely to be influenced by pregnancy steroids, insulin-like growth factors and other hormones that themselves may be influenced by external factors, including dietary or other factors. It is essential to identify the underlying biologic mechanisms for the positive association between birth size and breast cancer risk. A simple hypothesis points to the positive association between hormonal influences (pregnancy steroids or insulin-like growth factors) and total number of stem cells.19, 20 To the extent that number of stem cells is associated with the total number of cells, one might expect a positive association between breast tissue mass and breast cancer risk.

This study provides evidence that intrauterine factors may influence future breast cancer risk. In particular, the consistent positive association related to birth length suggests that factors that stimulate intrauterine longitudinal growth may be important for the risk of breast cancer.

Acknowledgements

The St. Olav Birth Cohort study was established by grants provided by the Norwegian Cancer Society and by the Norwegian Medical Research Council.

Ancillary