|Home | About | Journals | Submit | Contact Us | Français|
Larger body size in childhood is correlated with earlier age at menarche; whether birth and infant body size changes are also associated with age at menarche is less clear. The authors contacted female participants enrolled in the New York site of the US National Collaborative Perinatal Project born between 1959 and 1963 (n =262). This racially and ethnically diverse cohort (38% white, 40% African American, and 22% Puerto Rican) was used to investigate whether maternal (body size, pregnancy weight gain, age at menarche, smoking) and birth (birth weight, birth length, placental weight) variables and early infant body size changes were associated with age at menarche even after considering later childhood body size. Higher percentile change in weight from ages 4 months to 1 year was associated with earlier age at menarche even after adjustment for later childhood growth (β =−0.15, 95% confidence interval: −0.27, −0.02 years per 10-percentile change in weight from ages 4 months to 1 year). The association was in the same direction for all 3 racial/ethnic groups but was largest for the white group. These New York Women's Birth Cohort Adult Follow-up data (2001−2006) suggest that infant weight gain, in addition to childhood weight gain, may be associated with earlier age at menarche.
Larger body size (higher weight, taller height, and higher body mass index) in childhood has been consistently associated with earlier age at menarche (1–14). Whether body size and changes earlier in life are also associated with age at menarche is less clear. About 40 years ago, Frisch and Revelle (4) first hypothesized that achieving a critical weight was necessary for timing of menses and that increases in childhood body weight over time may explain the secular decline in menarche (4). They also observed higher birth weights among cohorts with earlier age at menarche (4). More recent studies, however, have supported that smaller babies have an earlier age at menarche (15–21), although few studies have detailed measures of weight and height changes after birth. One study that did utilize early childhood growth data suggested that rapid childhood growth from ages 2 years to 8 years was the most important predictor of age at menarche (22). This study, however, lacked data on infant growth prior to age 2 years.
Incomplete assessment of growth throughout infancy and childhood can affect overall inferences relating birth weight and earlier growth measures to age at menarche. In particular, babies grow very rapidly during the first year of life, and rapid growth during one period of life is usually associated with slower growth during other periods. Thus, studies with prospective measures of growth during the first year of life are needed to determine whether the processes that relate growth to timing of menarche begin much earlier in life.
We undertook a study to investigate whether weight and height changes from birth through age 7 years were related to age at menarche. We examined these questions using information from the New York site of the National Collaborative Perinatal Project, which, unlike other birth cohorts, includes detailed measures of infant growth and is racially and ethnically diverse.
All women who were born at Columbia Presbyterian Medical Center in New York from 1959 to 1963 and participated in the US National Collaborative Perinatal Project until age 7 years were eligible for participation. Of the 841 eligible women, 44% were successfully traced, and 70% of the traced women (n=262) participated in the adult follow-up: 38% were white, 40% were African American and 22% were Puerto Rican. Of the women we traced, 18 refused to participate, 3 were too ill to participate, and 16 had died. The study was approved by the Internal Review Board at Columbia Medical Center. Refer to Broman (23) for further details on the overall National Collaborative Perinatal Project and to Terry et al. (24) regarding the New York Women's Birth Cohort Adult Follow-up Study.
Baseline data were available for all eligible women, and we compared those women who participated with the remaining cohort. Higher family socioeconomic status (SES) at age 7 years, availability of maternal social security number, and white race/ethnicity were statistically significantly associated with a higher probability of tracing. For those traced, race/ethnicity was associated with participation, with African Americans and Puerto Ricans less likely to participate than whites. In addition, higher weight at age 7 years was associated with being less likely to participate (odds ratio (OR)=0.95, 95% confidence interval (CI): 0.92, 0.99). None of the other maternal characteristics, including age at menarche as well as infant and early childhood growth measures, were associated with participation or with tracing, either overall or within each racial/ethnic subgroup (24).
Mothers were enrolled in the National Collaborative Perinatal Project during their second or third trimester. Information on maternal age at menarche and maternal smoking behavior was obtained through self-report at the initial prenatal visit. Information on pregnancy conditions was recorded prospectively at the clinic where mothers received prenatal care, following a uniform protocol. Prospective growth measures for the children in the National Collaborative Perinatal Project were assessed by trained clinical researchers using a standard protocol. Measures included birth weight; birth length; placental weight; and weight and height at ages 4 months, 1 year, and 7 years (23). Because all subjects did not attend the clinic at exactly these time points, and to reduce any potential bias arising from the different measurement times, we interpolated these measurements at the target times using individual cubic interpolation splines (25). No interpolation was needed for birth measurements. Maternal weight gain was calculated from weight measured just prior to birth and reported weight prior to pregnancy. Maternal body mass index was calculated by reported height and weight prior to pregnancy. In addition to the physical measurements, SES was determined from data on maternal and paternal education, occupation, and income at enrollment and when the child was age 7 years (23). Information on income, education, and occupation for the head of household or the main wage earner (most frequently the father) was combined into a continuous SES index, with higher scores indicating higher or more privileged SES (26, 27).
From 2001 to 2006, we sent follow-up questionnaires to women who were successfully traced to obtain information on age at menarche, adult health, and reproductive events. Participants were asked to recall age at menarche by responding to the following question: How old were you when you had your first menstrual period? The average age of women during adult data collection was 41.8 years (range, 38–46).
We examined associations between maternal, infant, and childhood variables and age at menarche, with age at menarche described as both a continuous variable and a binary variable, dichotomized as earlier (≤12 years) and later (>12 years). We first examined univariate associations using correlation coefficients for continuous variables, chi-square tests, and analysis of variance to compare averages between subgroups. We then examined multivariable models using age at menarche as a continuous variable. We also constructed supplemental models that considered age at menarche as a categorized variable using logistic regression modeling the probability of earlier age at menarche (≤12 years) versus later age at menarche (>12 years) (28) and quantile regression methods (29).
Covariates considered in the multivariable models included maternal variables (age at pregnancy, age at menarche, prepregnancy body mass index, weight gain during pregnancy, smoking status, education, occupation), family variables (family SES at age 7 years), and birth variables (gestational age, birth weight, birth order, birth length, placental weight, race/ethnicity). Postnatal growth was characterized as percentile-rank changes over 3 consecutive periods, from birth to age 4 months, from ages 4 months to 1 year, and from ages 1 year to 7 years. Changes in the percentile rank provide a convenient and intuitive way to assess the growth rate while avoiding additional adjustment for age-dependent measurement scales. We developed parsimonius models based on the 10% change in parameter estimate criterion for the growth variables that were statistically significant in the saturated model. In this paper, we present these final models alongside the fully adjusted multivariable model, which includes all prenatal and early-life variables, so the reader can determine the empirical impact of excluding the other prenatal and postnatal variables. The final models were then stratified by race/ethnicity (white, African American, and Puerto Rican as classified at birth) to evaluate whether the magnitude of the effect estimates differed between these groups.
Overall, the mean age at menarche for the cohort was 12.5 years (standard deviation, 1.72), and the median age was 12 years. By race/ethnicity, the means were as follows: whites, 12.53 (standard deviation, 1.79); African Americans, 12.70 (standard deviation, 1.85); and Puerto Ricans, 12.14 (standard deviation, 1.29) (P from analysis of variance= 0.14). Table 1 summarizes descriptive statistics for women who reported earlier age (≤12 years) compared with later age (>12 years) at menarche. Women who reported earlier age at menarche were more likely to have a mother with an earlier age at menarche (P=0.05). The average age at menarche for the mothers was 12.92 (standard deviation, 1.56) years, and the median age was 13.0 years. The correlation between maternal age at menarche and daughter age at menarche was 0.18 (P = 0.005). Figure 1 shows the difference between maternal age at menarche and daughter age at menarche. On average, age of the participants at menarche was 5.0 months earlier than that of their mothers. Higher family SES at age 7 years was associated with later age at menarche (P=0.01). In addition, women who reported an earlier age at menarche were more likely to be heavier at age 7 years (average weight =24.7 kg) compared with those who reported a later age at menarche (average weight =23.2 kg) (P=0.01). The correlation between weight at age 7 years and age at menarche was −0.17 (P=0.009). Other characteristics, including birth weight, were not univariably different between the 2 groups.
Table 2 reports the average age at menarche by quartile of birth weight and weight at age 7 years. Within each category of birth weight, with the exception of the second quartile, the average age at menarche was lower among girls in the highest quartile of weight at age 7 years compared with girls in the lowest quartile of weight at age 7 years. The average age at menarche was lower for higher-birth-weight babies only among the girls of lower weight at age 7 years.
Table 3 reports the findings from the linear regression models. Rapid postnatal weight gain from ages 4 months to 1 year and ages 1 year to 7 years was negatively associated with age at menarche (β=−0.03, 95% CI: −0.13, 0.07 years per 10-percentile increase in weight from birth to age 4 months; β=−0.13, 95% CI: −0.24, −0.02 years per 10-percentile increase in weight from ages 4 months to 1 year; β=−0.10, 95% CI: −0.19, −0.01 years per 10-percentile increase in weight from ages 1 year to 7 years). After adjustment for other maternal and family variables and for postnatal changes in height, these associations remained fairly consistent, but only rapid weight gain from ages 4 months to 1 year was statistically significant (β=−0.15, 95% CI: −0.27, −0.02 years per 10-percentile increase in weight from ages 4 months to 1 year). Later maternal age at menarche and higher SES were positively associated with later age at menarche.
Percentile weight changes in each of the 3 time periods were negatively correlated with each other (r=−0.16, P=0.01; r=−0.26, P<0.0001; and r=−0.19, P=0.002 for percentile weight change from birth to age 4 months and ages 4 months to 1 year, from birth to age 4 months and ages 1 year to 7 years, and from ages 4 months to 1 year and ages 1 year to 7 years, respectively). However, there were no statistically significant interaction effects on age at menarche between percentile weight change in any of the 3 time periods and birth weight (P=0.35). These negative correlations, however, generally resulted in stronger parameter estimates for earlier-period weight percentile measures as later weight measures were added to the model (refer to Table 3). Apart from the growth variables in other time periods, most other variables that we assessed, including placental weight, maternal prepregnancy body mass index, and maternal pregnancy weight gain, did not affect the parameter estimates of the association between weight and height percentile changes and age at menarche. Exclusion of data for 29 babies born preterm did not affect inferences.
We also constructed supplemental models using quantile regression analysis to estimate whether the effects reported in Table 3 were similar across quantiles of age at menarche (data not shown). Overall, we found that the negative association (larger size, earlier menarche) for all measures (birth weight and weight change in each of the postnatal periods) was similar across quantiles. Findings from multivariable logistic regression models for earlier age at menarche (≤12 years) relative to later age at menarche (>12 years) also revealed similar inferences: the association between birth weight and age at menarche was positive; however, it was not statistically significant (OR=2.59, 95% CI: 0.79, 8.53). Postnatal weight gain during the 3 time periods was also positively associated with earlier age at menarche (OR = 1.10, 95% CI: 0.98, 1.24; OR=1.16, 95% CI: 1.02, 1.33; and OR=1.17, 95% CI: 1.05, 1.31 for a 10-percentile increase in weight from birth to age 4 months, ages 4 months to 1 year, and ages 1 year to 7 years, respectively).
As a secondary analysis, we also examined the associations between absolute measures of weight and height at ages 4 months, 1 year, and 7 years and age at menarche (data not shown). These models supported negative associations of birth weight, weight at age 1 year, and weight at age 7 years with age at menarche after adjusting for birth length, postnatal height changes, maternal age at menarche, and family SES at age 7 years. Weight at age 4 months, however, was positively and statistically significantly associated with later age at menarche (β=0.49, 95% CI: 0.07, 0.92 years per kilogram of weight at age 4 months). There were no statistically significant interactions in the linear model between birth weight and absolute measures of weight at ages 4 months, 1 year, and 7 years (P=0.64).
Figure 2 presents the parsimonious linear model stratified by race/ethnicity. Although the stratified models support negative associations between percentile weight gain and age at menarche for the latter 2 time periods—4 months to 1 year and 1 year to 7 years—for all 3 groups, the point estimates were stronger for the white group regarding weight gain from ages 4 months to 1 year and were stronger for the Puerto Rican group regarding weight gain from ages 1 year to 7 years. There were no statistically significant interactions in the linear model between race/ethnicity and any of the growth measures, measured as either percentile difference (P=0.99) or absolute measures (P=0.95).
Table 4 reports the overall final model (the last column of Table 3) stratified by weight at age 7 years (defined by median weight at age 7 years). These results suggest that the associations of percentile weight change from ages 4 months to 1 year and from ages 1 year to 7 years with earlier age at menarche are observed primarily among those whose weight at age 7 years is below the median. Among girls of lower weight at age 7 years, rapid weight change (as measured by percentile change) from ages 4 months to 1 year and from ages 1 year to 7 years was associated with earlier age at menarche, although these associations were not statistically significant. Rapid increases in height, particularly from ages 4 months to 1 year, were associated with later age at menarche among girls of lower weight at age 7 years.
Overall, in this racially and ethnically diverse cohort, we observed rapid weight gain from ages 4 months to 1 year and from ages 1 year to 7 years to be associated with earlier age at menarche. Higher birth weight was also associated with earlier age at menarche, but the association was not statistically significant. Although limited by our overall small sample size, our study suggested that, even among girls of lower weight at age 7 years, rapid infant weight gain may be associated with earlier age at menarche, and rapid height changes in infancy may be related to later age at menarche. These data suggest complex pathways that need to be replicated in larger studies in which infant growth measures are available. Rapid infant weight gain may be associated with earlier age at menarche in addition to body size later in childhood because early growth may be associated with changes in leptin, insulin-like growth factors, and other growth and steroid hormones (30).
Although most studies of menarche support that a larger body size in childhood (usually measured at ages 7–8 years) is associated with earlier age at menarche (1–14), data on growth during earlier life periods have been less consistent. Despite Frisch and Revelle's observation (4) that populations with an earlier age at menarche had higher birth weights, many other studies have observed that babies of smaller birth weight and/or length have an earlier age at menarche (15–21). For example, in studies by Adair (21) and Tam et al. (15), babies who were long and light at birth had a 6–12-month earlier age at menarche. Our distributions of birth weight and birth length (average of 3.15 kg and 50 cm, respectively) were slightly lower than those of other cohorts (15, 17, 19), but, when we performed secondary analyses with the birth-size cutpoints used in these cohorts, our overall inferences did not materially change. For example, using the same cutpoints for birth weight and weight at age 7 years as Cooper et al. (20), we found that those girls in the highest fifth of the birth-weight distribution experienced menarche 6 months earlier than girls in the lowest fifth of the birth-weight distribution. We further assessed the effect of low birth weight on age at menarche. In our cohort, low-birth-weight babies (<2,500 g) were more likely to experience menarche 7.2 months later than babies born weighing 2,500 g or more, although our study included very few low-birth-weight babies (n=25).
Our mean and median ages at menarche (12.5 years (standard deviation, 1.7) and 12.0 years, respectively) were lower than those in other studies that reported a mean age ranging from 12.6 years to 13.2 years (15, 17–19, 22, 31) and a median of 13.0–13.1 years (16, 20, 21); however, applying the cutpoints for age at menarche used by others did not alter our findings. Except for one study (31), the other studies of birth size and age at menarche were conducted outside of the United States and were more racially and ethnically homogeneous than our cohort.
A more likely explanation for the difference in findings between our studies and the others is that most studies did not adjust for postnatal growth. For example, in 2 separate analyses of the 1946 British Birth Cohort, high birth weight was associated with later and then earlier age at menarche, depending on whether postnatal growth was considered simultaneously in the model (20, 22). However, no information on weight and height changes between birth and age 2 years was available for the 1946 birth cohort. In addition to having prospectively collected information on weight at birth and ages 4 months, 1 year, and 7 years, we were able to examine whether further adjustment for height changes during earlier life; maternal variables, including maternal age at menarche, body size, and pregnancy weight gain; placental weight; gestation length; and family SES confounded our results. Our findings remained after further adjustment and were also robust to alternative statistical models.
Although we had detailed prospective data on all of the independent variables, we were limited to retrospective assessment of age at menarche because the cohort was followed prospectively until age 7 years only. However, age at menarche has been shown to be reliably reported even 17–40 years later (range of Pearson correlation coefficients r=0.60–0.79) (32–38), although reliability was lower when measured by the kappa statistic (33). The reliability of our retrospective assessment was likely enhanced by asking women to recall age at menarche in only discrete years. In our study, recalled age at menarche was positively correlated with maternal age at menarche as well as inversely correlated with weight and height at age 7 years, consistent with many other studies (15, 16, 19–22, 31, 39) and supporting the overall validity of our measure. Nevertheless, the measurement error likely influenced our ability to detect more subtle associations with age at menarche. However, there is no reason to expect that retrospective assessment of age at menarche would be differential with respect to exposure; therefore, we would expect the true differences to be even larger because of this nondifferential misclassification of the outcome. In addition, in terms of maternal age at menarche, maternal pregnancy variables, birth size, or any postnatal height or weight changes, women who participated in the adult follow-up and therefore had age-at-menarche data did not differ from those who were not traced and/or who were traced and did not participate (24).
The main limitation of our study was the sample size, which limited our overall statistical power. For example, we were able to detect differences in age at menarche of 10 months or more between racial/ethnic groups. However, for most of the continuous growth measures, and interactions between these continuous measures, we had greater statistical power, particularly in the linear regression model.
Between the late 1960s and 1990, there was a 3-month decline in age at menarche for US white girls and a 5.5-month decline for US African American girls (40). Comparison of mothers in our cohort, who experienced menarche from 1928 to 1960, with their daughters, who experienced menarche from 1967 to 1979, supports the accumulating data suggesting a secular decline in age at menarche (41–44). Specifically, 50 girls were the same age at menarche as their mothers, 118 had an earlier age at menarche than their mothers, and 84 had a later age at menarche than their mothers (average difference =5.0 months).
Some have suggested that the declines in age at menarche over time may be explained by the trends in increasing body weight (4, 45, 46). Birth weight has also been increasing in recent decades (47). Our data support that, in addition to child body size, early infant growth may be associated with earlier age at menarche. Taken together, these findings suggest that multiple measures of infant and early childhood growth are needed to understand the complex interplay between body size and timing of menarche. These findings add to a growing body of literature suggesting that body size and growth early in life may impact the timing of menses.
Author affiliations: Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York (Mary Beth Terry, Jennifer S. Ferris, Parisa Tehranifar, Julie D. Flom); and Department of Biostatistics, Mailman School of Public Health, Columbia University, New York, New York (Ying Wei).
This work was supported by the Department of Defense Breast Cancer Research Program (grant DAMD170210357) and the National Cancer Institute (grant K07CA90685).
The authors thank the following individuals for their contributions to the New York Women's Birth Cohort Adult Follow-up Study: Ezra Susser, Tara Kalra, Tamarra James, Lina Titievsky-Konikov, Dipal Shah, Shobana Ramachandran, Julia Meurling, Adey Tsega, Sujata Narayanan, and Summer Wright.
Conflict of interest: none declared.