We estimated an increased risk of early fibroid diagnosis in association with being fed soy formula during infancy, having a mother with prepregnancy diabetes, being born at least 1 month early, and reporting factors indicating low socioeconomic status during childhood (low household education, being poor, and not having enough to eat). We also noted associations with fibroids for having a mother with gestational diabetes and DES use during pregnancy, although these associations were restricted to reporting probable rather than definite exposure.
The association with soy formula is of interest given the estrogenic isoflavones found in soy products. Infants fed soy formula are exposed to isoflavone levels that are more than five times higher than typical levels for adults consuming soy-based foods (Setchell et al. 1997
). Genistin is the naturally occurring isoflavone predominantly contained in soy formula, which can easily be hydrolyzed in the gut to the estrogenically active form of the compound, genistein, based on reporting of high plasma (Setchell et al. 1997
) and urinary (Cao et al. 2009
; Hoey et al. 2004
) concentrations of genistein in infants fed soy formula. The increased risk of fibroids in association with being fed soy formula within the first 2 months of life, which may include the time period most sensitive to genistein exposure, was similar to the association with soy formula at any time during infancy.
Genistein has been extensively investigated in laboratory animals. In particular, neonatal treatment of mice with genistein has been associated with development of uterine adenocarcinoma (Newbold et al. 2001
), abnormalities with mammary gland development and differences with mammary gland levels of estrogen and progesterone receptors (Padilla-Banks et al. 2006
), alterations with estrous cycles and reduced fertility (Jefferson et al. 2005
), and early reproductive senescence (Jefferson et al. 2005
). However, studies with adult outcomes in humans have been lacking, except for one small study reporting longer duration and increased pain with menstrual bleeding for women who were fed soy formula (Strom et al. 2001
). This finding is interesting given that pelvic pain and heavy bleeding are the most common symptoms of fibroids.
Our strongest association with fibroids was for maternal prepregnancy diabetes. Prenatal exposure to diabetes has also been associated with increased risk of obesity, abnormal glucose tolerance, and type 2 diabetes in adulthood (Dabelea 2007
; Fetita et al. 2006
). However, this is the first study to evaluate whether in utero
exposure to maternal diabetes affects fibroid risk in early adulthood. One hypothetical mechanism by which in utero
exposure to diabetes would affect later fibroid pathogenesis is the alteration of methylation patterns in regions that affect expression of relevant genes. In particular, expression of imprinted genes is especially sensitive to changes in methylation patterns (Thompson et al. 2001
; Waterland and Jirtle 2004
). Two studies in mice have reported that exposure of fetuses to induced maternal diabetes and of embryos before implantation to in vitro
insulin affected expression of two imprinted neighboring genes, H19
, by altering their methylation status and resulted in changes in fetal development and birth weight (Shao et al. 2007
). Seven microarray studies reviewed by Arslan et al. (2005)
reported at least a 2-fold increase in IGF2
expression in fibroids relative to normal myometrium.
Using the animal model of fibroids, the Eker rat, Cook et al. (2007)
found an association between early exposure to DES and later fibroid development. Consistent with the animal model, Baird and Newbold (2005)
reported a positive association of self-reported in utero
DES exposure with fibroid diagnosis based on ultrasound assessment within the NIEHS Uterine Fibroid Study. However, Wise et al. (2005a)
reported no association between maternal DES use and fibroids based on medically documented DES exposure and surgical fibroid cases. Given our inconsistent associations for women who report definite versus probable in utero
DES exposure, conclusions from our study are unclear.
We found a consistent association of fibroids with three indicators of low socioeconomic status during childhood (low household education, food insecurity, and poverty). These factors may influence development of fibroids through changes in methylation patterns in childhood that persist and affect gene expression as adults. Plausibility of this hypothesis is based on animal and human data on early-life neglect or abuse. Different methylation patterns within the hippocampus of adult rats were detected based on whether they received maternal care early in life (Weaver et al. 2004
). In a small sample of adult men who committed suicide, methylation patterns in similar genes in the hippocampus varied based on whether they were abused during childhood (McGowan et al. 2009
). Whether being exposed to low socioeconomic conditions during childhood can affect methylation patterns in genes relevant to fibroid pathogenesis needs further investigation.
We also found a strong association with fibroids for being born at least 1 month before mother’s due date. A weak association with low birth weight was no longer present after excluding women who were born at least 1 month early. Because levels of estrogen and progesterone rise throughout pregnancy, one hypothesis is that women who are born early are deprived of the estrogen needed for full differentiation of their reproductive system (Trotter and Pohlandt 2000
). Our analyses were limited by having only 60% of women who reported gestational age at birth. Missing data may be related to the true values of gestational age or other birth-related variables, but results were similar when we repeated the analyses assuming women missing gestational age data were not born ≥ 2 weeks early (data not shown).
Selection bias is a potential limitation for other exposures as well, given that the proportion of missing values was as high as 20%. There were slight differences in the proportion with fibroids based on whether exposures were missing, with generally more women reporting fibroids among those with missing responses. Based on the assumption that women with missing data for rare exposures (preeclampsia, pregnancy-related hypertension, soy formula, DES use, prepregnancy diabetes, and gestational diabetes) were likely to be unexposed, we repeated analyses in which we considered women with missing values for these factors as unexposed. However, changes in RR estimates were minimal (data not shown).
There is also the potential for misclassification of exposures given that women were reporting exposures during infancy and related to their mother’s pregnancy. However, women were provided phone cards to encourage them to ask these questions directly of their mothers, and we excluded older women (> 59 years) who would be less likely to have living mothers to ask about these exposures. In addition, response categories for many of the exposures included options of “definite” and “probable” that allowed for uncertainty in reporting. Associations with fibroids were generally consistent for definite and probable exposure, except associations with in utero DES exposure and maternal gestational diabetes, for which associations were much stronger with reporting probable exposure. Because none of these exposures is known to be related to fibroids, exposure misclassification would likely be nondifferential, which would generally result in RR estimates biased toward the null.
Our assessment of fibroid diagnoses was based exclusively on self-report. However, because fibroid incidence increases strongly with age, we only considered diagnoses by 35 years of age to reduce misclassification in the noncase group. Our estimated risk of 8% for early diagnosis of fibroids was similar to the risk of 11% for self-reported fibroids by 35 years of age in white women (35–49 years of age) from the NIEHS Uterine Fibroid Study (Baird DD, unpublished observations). We also reported a substantially greater proportion of women with early diagnosis of fibroids having hysterectomies compared with women without early diagnosis, which suggests that many of the women with fibroid diagnoses included in our case definition had fibroid-related morbidity. However, we do not have information on fibroid-related symptoms at time of diagnosis. We also excluded older women from our analyses because of possible secular differences in use of ultrasounds for fibroid diagnoses. Because studied factors may also be related to fibroids diagnosed later in life, including women with fibroids diagnosed after 35 years of age as noncases may have resulted in an underestimation of RRs. However, repeating our analyses after exclusion of women with later diagnoses of fibroids (> 35 years) from the noncase group did not affect RR estimates with early diagnosis of fibroids for four of our main findings (maternal prepregnancy diabetes, soy formula, low childhood socioeconomic status, and being born at least 1 month early (data not shown).
Strengths of this study include a large sample size, which allowed us to examine associations with rare intrauterine and early-life exposures. We adjusted for factors that may affect recall of exposures, including participant’s age and education. In addition, despite the potential for misclassification bias from self-reported exposure information and fibroid diagnoses, we observed expected associations between specific factors, including a positive association between early age at menarche and fibroids and an increased reporting of maternal preeclampsia among firstborn women and those from a multiple birth (data not shown).