Early onset of puberty, which has often been approximated by young age at menarche, can have important implications for conditions diagnosed later in life, such as breast cancer and cardiovascular disease.2
Furthermore, animal studies have shown that prenatal and neonatal exposure to exogenous estrogens can alter pubertal timing.26-28
We found that several early-life exposures were associated with early menarche: low birth weight, young maternal age at the index birth, firstborn status, soy formula, and maternal exposures during the index pregnancy (pre-pregnancy diabetes, pregnancy-related hypertensive disorder, smoking, and DES use). Associations with low birth weight, pre-pregnancy diabetes, pregnancy-related hypertensive disorder, in utero
DES, firstborn status, and soy formula were stronger for very early menarche (≤10 years) than for menarche at 11 years. Soy formula, multiple birth, and having been born at least one month early were the only exposures associated with late menarche.
Age at menarche was based exclusively on self-report. Previous reports suggest that there is generally moderate agreement between age at menarche reported in adolescence and later reported by middle-aged women.36-38
Classifying women into three categories corresponding to early, typical (12-13 years), and late menarche has been reported to improve validity.37
When our data were categorized in this way, results were largely unchanged (results not shown) except for associations that were specific to very early or very late menarche.
Our large study size allowed examination of associations with several early-life exposures. By using polytomous logistic regression, we could evaluate exposures associated with both early and late menarche. In addition, we could evaluate associations with prenatal exposures that are specific to certain birth cohorts (i.e. DES) or that may have occurred at different rates over time (e.g. smoking). Effects of early-life exposures on age at menarche may be important because early menarche is related to breast cancer and other health conditions in adulthood.2
Historical exposures, such as maternal DES use, continue to be relevant for the majority of women at risk for breast cancer today.
By design, the Sister Study enrolled women with a family history of breast cancer in order to enrich the cohort for environmental and genetic risk factors for breast cancer. To the extent that early menarche–a breast cancer risk factor–was more common among our study participants than among women in the general population, the statistical power for estimating associations with early-life exposures would be enhanced.39
In particular, the frequency of early menarche (20%) in our cohort of predominantly white women was lower than in white girls from the Bogalusa Heart Study cohort (30%)40
but higher than in white girls from a large clinical study conducted in the 1990s.41
There is also concern that women with a family history of breast cancer may report age at menarche either more or less accurately than women without such family history. However, our estimates of mean age at menarche by birth decade for whites (data not shown) were similar to results from the National Health and Nutrition Examination Survey (NHANES).42
Furthermore, given that the entire cohort has a family history of breast cancer and all were breast-cancer free when interviewed at baseline, effects on reporting of age at menarche could be expected to be shared across the cohort and unlikely to bias the associations presented.
Misclassification of self-reported exposures is a potential study limitation. We provided phone cards to encourage participants to contact their mothers or other relatives, but we did not ask whether they contacted relatives or consulted records before reporting early-life events. Response categories for some early-life exposures allowed for further uncertainty with reporting. Although not shown, associations with exposures reported as “definite” were generally similar or stronger than those with exposures reported as probable. However, analyses of infancy feeding are limited by the lack of adequate data on duration. For example, the null association with ever having been breastfed may have been influenced by the inclusion of those breastfed for a very short duration.
Because of the large proportion of missing data for some early-life exposures, we repeated our main analyses and considered whether there was additional confounding among early-life exposures after multiple imputation of missing data. Results from these secondary analyses were largely similar to the main analyses and showed little confounding among the early-life exposures. Because maternal vital status at baseline was one of the strongest and most consistent factors that influenced whether data were missing for early-life exposures, we also repeated our main analyses after restricting to the approximately 60% of participants whose mothers were alive at baseline. Results from the restricted subset were less precise, but differences were generally minor except that pre-pregnancy diabetes was no longer associated with very early menarche (rRR = 1.06 [95% CI = 0.32-3.56]). However, because having diabetes influences lifespan, our restricted subset had fewer mothers with pre-pregnancy diabetes who were alive at baseline.
Although heavier childhood body weight is a strong predictor of age at menarche,1
we did not adjust analyses for childhood body weight because it may be on the causal pathway between the exposures and age at menarche. Having been heavier than peers at age 10 was associated with maternal pre-pregnancy diabetes and with prenatal exposure to smoking, while having been lighter than peers at age 10 was associated with having a twin (or triplet) sibling (data not shown), which is consistent with childhood weight being on the causal pathway for age at menarche. However, pregnancy-related hypertensive disorder was weakly associated with both having been lighter and having been heavier than peers at age 10, and low birth weight was strongly associated with having been lighter than peers at age 10.
Although age at menarche does not completely capture pubertal development, our finding that soy formula consumption was associated with both early and late menarche is consistent with animal data suggesting that the dose of genistein influences whether puberty will be advanced or delayed. Mice administered a higher dose of genistein had a delayed vaginal opening (marker of puberty) while mice given a lower dose had an accelerated vaginal opening.26
Neonatal administration of genistein to mice has also produced other alterations in reproductive characteristics, including changes in estrous cycles, early reproductive senescence, and decreased fertility.43
Soy formula delivers a high dose of estrogenic isoflavones to infants per unit body weight44
; high plasma and urinary concentrations of genistein have been reported in infants fed soy formula.44-46
There are inconsistent results from the two previous epidemiologic studies of soy versus cow milk formula during infancy: one reported no difference in age at menarche,47
and the other reported that those given soy formula early in life (≤4 months through ≥6 months of age) had earlier menarche.48
Soy formula composition changed in the early 1960s from soy flour to a more highly digestible soy protein isolate,49
and so we evaluated whether the association between soy formula and age at menarche differed for women born before and after this time period. The association with late menarche did not change across birth decades, while the association with very early menarche was seen only in those born in 1960 or later (rRR = 1.50 [95% CI = 1.03-2.19]).
DES has been associated with various reproductive abnormalities in women.50
In mice, prenatal administration of DES results in an earlier vaginal opening.27,28
We found an association of DES with very early menarche (≤10 years of age). Two studies based on documented prenatal DES exposure found no difference in mean age at menarche51
and no association with early or late age at menarche.52
However, these studies did not report on the frequency of very early menarche. Evaluating mean differences in age at menarche may not be sensitive to differences in the frequency of very early menarche because this group typically represents a small proportion of any population (7% in our study sample). Similar to our study findings, Hatch et al.53
reported an association with very early menarche in a cohort study of medically documented in utero
DES exposure. Information on DES dose and timing during pregnancy was not available in our study.
We found that maternal pre-pregnancy diabetes and pregnancy-related hypertensive disorder were associated with early menarche. Two studies reported no association between prenatal exposure to maternal preeclampsia and age at menarche.54,55
However, a large UK cohort study found an earlier mean age at menarche in daughters whose mothers had preeclampsia during their gestation.12
Prenatal exposure to maternal diabetes has been previously reported to be associated with childhood obesity,29-31
but we are the first to investigate associations with early menarche in daughters.
Findings from previous studies have been inconsistent for associations with birth order and maternal age at birth.10-12,24,56
While the association with firstborn status was specific to very early menarche, we also found an association with early menarche for having had a teenage mother. Even though we adjusted associations for childhood family income, residual confounding may remain because teenage birth is related to poor socioeconomic status. The association with young maternal age at birth may be confounded by the maternal age at menarche because the mother’s menarche necessarily preceded her teenage pregnancy with her daughter, and it is known that ages at menarche are correlated for mothers and daughters.7
However, we have no data on the age at menarche of the participant’s mother. Our finding of an association between having been born from a multiple birth and very late age at menarche is consistent with a large UK cohort study that reported later age at menarche for twins.12
Our results are consistent with previous studies that found an association between lower birth weight and early menarche.12,15,18,21
However, a few other studies reported associations only with having been small for gestational age16,17
or having had lower birth weight and longer birth length.19,23
A study of white and Asian girls reported that younger age at menarche was not associated with low birth weight or preterm status, although nonsignificant associations in that direction were present.25
We had a high proportion of women with missing data on gestational age and birth weight. However, after imputation of missing data, adjustment for preterm status did not affect the low birth weight association with early menarche. Associations with low birth weight may be influenced by later growth, but we had only crude data on pre-pubertal growth. In particular, one study reported that the association with birth weight varied by childhood body size.24
Three studies found that subsequent infancy or pre-pubertal growth, but not birth weight, was associated with age at menarche,11,20,22
although two of these studies noted nonsignificant associations between high birth weight and age at menarche after accounting for subsequent growth.20,22
Five previous studies found associations between prenatal exposure to maternal smoking and younger age at menarche,8,10-13
with two of these studies reporting associations specifically with smoking in the third trimester10
and heavy smoking of at least a pack of cigarettes a day.8
Two studies in predominantly black or multiethnic cohorts reported an association with later age at menarche and heavy maternal smoking.9,14
Although we had no information on smoking frequency or timing during pregnancy, our finding of an association between maternal smoking and early menarche was consistent with other studies in predominantly white cohorts.8,10-13
It is also possible that the association with prenatal exposure to smoking may be confounded by maternal age at menarche if mothers with early menarche were more likely to become smokers and have daughters with early menarche. However, associations between prenatal exposure to smoking and early menarche remained in three of the studies after adjustment for maternal age at menarche.8,10,11
Our study adds to prior evidence that early-life exposures influence age at menarche and is the first to report an association with maternal pre-pregnancy diabetes. Besides replication of findings within other study populations, future studies should examine a broader set of pubertal changes beyond age at menarche. Greater understanding of the possible biologic mechanisms that could explain our findings is also an important area for future research. Furthermore, our findings have methodological implications, as early-life factors may confound some of the associations between age at menarche and hormonally-dependent adult outcomes such as breast cancer.