Objective. Although several studies have suggested that early menarche is associated with the development of adult overweight, few have accounted for childhood overweight before menarche.
Study Design. A 30-year follow-up of the original participants in the Newton Girls Study, a prospective study of development in a cohort of girls followed through menarche, provided data on premenarcheal relative weight and overweight (BMI >85th percentile), prospectively obtained age at menarche, self-reported adult BMI, overweight (BMI > 25), obesity (BMI > 30) and, for a subset of participants, percentage body fat by dual-energy x-ray absorptiometry.
Results. Of the 448 women who participated in the adult follow-up at a mean age of 42.1 years (SD: 0.76 years), 307 had childhood data with which to characterize premenarcheal and menarcheal weight status and age at menarche. After a follow-up of 30.1 years (SD: 1.4 years), reported BMI was 23.4 (4.8), 28% were overweight, and 9% were obese. In multivariate linear and logistic-regression analyses, almost all of the influence on adult weight status was a result of premenarcheal weight status (model R2 = 0.199). Inclusion of a variable to reflect menarcheal timing provided very little additional information (model R2 = 0.208). Girls who were overweight before menarche were 7.7 times more likely to be overweight as adults (95% confidence interval: 2.3, 25.8), whereas early menarche (at ≤12 years of age) did not elevate risk (odds ratio: 1.3, 95% confidence interval: 0.66, 2.43). A similar pattern of results was observed when percentage body fat in adulthood was evaluated.
Conclusions. The apparent influence of early maturation on adult female overweight is largely a result of the influence of elevated relative weight on early maturation. Interventions to prevent and treat overweight should focus on girls before they begin puberty.
Current estimates suggest that almost 2 in 3 adult Americans are overweight, and 1 in 4 is obese.1 Worldwide, it is estimated that as many as 300 million people are obese, with the number expected to continue to rise.2 The extent and severity of the epidemic of obesity has refocused interest on the natural history of the disease. The existence of 2 critical periods in childhood for the development of obesity is posited: early childhood and adolescence. Common to both periods is that each is characterized by rapid growth. A possible role for childhood growth as antecedent to adult obesity represents an important line of inquiry inasmuch as the developmental period of childhood may be more amenable to intervention.
An elevated likelihood of overweight persistence from childhood into adulthood has been suggested by a number of long-term studies, including the Bogalusa Heart Study3 and the Fels Longitudinal Study4 in the United States, as well as longitudinal studies from the United Kingdom.5,6 As would be expected, this tracking of BMI occurs over the range of relative weights and is strongest at higher weights. Tracking decreases with the time interval between early and late measurements. For example, in the Fels Longitudinal Study, probabilities of obesity at age 35 (BMI > 30) ranged from 0.23 (males) and 0.43 (females) at age 6 to 0.52 (males) and 0.77 (females) at age 17.4 Although not all studies demonstrate a gender difference in the likelihood of persistence of overweight,7 in studies that have demonstrated a difference, tracking of relative weight is stronger for females than for males.4–6,8
Maturational timing represents another established consequence of excess weight before adolescence. The observation that early menarche is associated with excess adiposity reflects the more general phenomenon that childhood obesity is auxogenic: some of the excess calories responsible for increased fatness are also reflected in accelerated growth and development. For this reason, obese children exhibit advanced bone age and tend to reach sexual maturity earlier. Whether this phenomenon is the result of a direct effect of adipose tissue, as proposed by Frisch and Revelle,9 or is secondary to overall skeletal maturity10 remains controversial. Nonetheless, earlier puberty and earlier menarche have been observed consistently among young girls who are overweight.
Several studies have also suggested that overweight or obesity are consequences of maturational timing. Some of the reports are cross-sectional,11 and most that are longitudinal have been initiated late in childhood12 or have a limited duration of follow-up.13 In studies with adequate follow-up that have attempted to look simultaneously at early fatness, maturational timing, and later fatness,6,12 the impact of early fatness and menarcheal timing has not been separated in a manner that allows one to assess their relative impact. Additionally, much of the previous work has relied on a retrospective measure of age at menarche,14–16 which, although generally quite accurate,17 is subject to random and nonrandom misclassification error.
In 1998, we undertook a retrospective follow-up of the original participants in the Newton Girls Study (NGS), a prospective study of physical growth and sexual maturation initiated in 1965.18 The contemporaneously obtained anthropometric data, together with the historical information of premenarcheal BMI and prospectively obtained age at menarche, allowed us to examine the relative influence of early weight status and maturational timing on subsequent weight status and, for a subset of women, percentage body fatness. Our goal was to establish the relative contributions of early fatness and menarcheal timing to adult weight status and adiposity at midlife.
SUBJECTS AND METHODS
This retrospective cohort study uses data from the NGS, a 10-year community-based prospective study of physical growth and sexual maturation initiated in 1965.18–20 In the original study, families of all female children attending public school in Newton, Massachusetts, whose 9th or 10th birthday would fall between October 1, 1965, and September 30, 1966, were invited to participate. At that time, the population of Newton, a small city located 7 miles from downtown Boston, was largely middle class and predominately white (98%). Of the 1258 families invited, 793 (63%) agreed to participate. Girls were followed through their first menstrual period (menarche) and for ∼2 years postmenarche. Of the 793 girls who were enrolled, 708 were followed through menarche. The girls’ mothers were sent monthly questionnaires that requested them to report the height, weight, exact date of each period, amount of flow, and the presence of discomfort and/or pain for the most recent menstrual period.
In 1998, original study participants were recontacted to update the health status of this cohort and to assess the accuracy of recall of early menstrual characteristics. Efforts to locate NGS participants included the use of the Massachusetts Department of Public Health Vital Statistics Office, alumni books for the Newton high schools, voting books from Newton City Hall, Internet address locators, and credit bureau searches. Because this cohort is comprised entirely of women, identifying married names was essential to location efforts. An initial contact letter and response card were sent through the mail to the subject or, if no address for the subject was available, to 1 of her parents or siblings. Up to 3 repeat mailings were sent to confirmed addresses. The letter described the study as an investigation of the relation between early growth and women’s later health outcomes. Subjects were invited to participate in a self-administered questionnaire to be completed by mail and/or to enroll in a study of bone health (dual-energy x-ray absorptiometry [DXA] substudy) that required an in-person visit to the Jean Mayer US Department of Agriculture Human Nutrition Research Center at Tufts University. The protocol was approved by the Tufts-New England Medical Center Institutional Review Board, and signed consent forms were obtained from participants.
Of the original 793 participants, 1% (n = 9) were deceased at the time of follow-up or were unable to respond because of poor health. Of the remaining 784 women, 57% (n = 448) completed the mailed questionnaire, 6% (n = 44) declined participation, and the remaining women could not be located (n = 50) or did not respond to repeated mailings to a verified address (n = 242). The follow-up questionnaire asked women to report their current height and weight. Two hundred thirty-six women were screened for the DXA substudy, which required a visit to the US Department of Agriculture Human Nutrition Research Center for additional direct measurements of bone density, body composition, and hormonal status.21
After excluding those subjects who had experienced natural or surgical menopause (n = 5), pregnancy or breastfeeding within the last 4 months (n = 6), or use of oral glucocorticoids for >4 consecutive months (n = 2) or had a history of a medical condition affecting bone metabolism (n = 6), 153 of the remaining 217 women were eligible and available to come into the center for DXA measurements. Subjects underwent a total body scan (DXA, model DPX-L scanner; Lunar Radiation Corp, Madison, WI) for measurement of lean tissue mass. In addition, weight and height were measured according to a standard protocol.21
Age at menarche was calculated based on the subject’s date of birth and the date reported by the girl or her mother within 1 month of the event. The report of the menarcheal event was made by the subject herself or, in some cases, by the subject’s mother. Early menarche was defined as menarche before age 12.
Premenarcheal and menarcheal BMI were calculated from height and weight. To provide a gender- and age-specific measure of relative weight, a BMI z score was calculated for each BMI measure with the reference parameters provided as part of the Centers for Disease Control and Prevention 2000 growth reference.22 The premenarcheal time interval was defined operationally as 0.5 to 1.5 years before menarche; the menarcheal time period was defined operationally as 0.5 years before to 0.5 years after menarche. For participants who had multiple measurements within these intervals, the mean of measurements was used; for the premenarcheal time period, girls had a mean (SD) of 4.3 (1.6) measurements and, for the menarcheal period, girls had a mean (SD) of 5.1 (1.8) measurements. Premenarcheal overweight was considered present when the mean premenarcheal BMI z score was >1.036 (85th percentile, Centers for Disease Control and Prevention growth reference).
Adult BMI was calculated from weight and height as reported on the questionnaire completed at the adult follow-up. Adult overweight and obesity were defined as BMI of >25 and >30, respectively, in keeping with current recommendations.23 Among the subjects in the DXA substudy, the correlation of BMI based on self-reported height and weight with measured BMI was 0.97, with the average measured BMI exceeding BMI based on self-report by 0.72 (SE: 0.11) BMI units.
For subjects who participated in the DXA substudy, the percentage of total weight that was fat tissue (percentage body fat) was calculated from fat and lean tissue mass estimated by DXA. The coefficient of variation in the laboratory performing the measurements is 2.2 for fat mass and 1.1 for lean mass.24
For inclusion in the current series of analyses, subjects had to provide at least 1 premenarcheal measurement and 1 menarcheal measurement and have an adult follow-up report of weight and height. The α level was set at .05. Comparisons between subjects followed and those lost to follow-up and between those followed and those in the DXA substudy were made by using a 2-sample t test. Adult BMI was very positively skewed; the variable was log-transformed to better approximate normality. Bivariate associations were assessed with Pearson correlation. The association between premenarcheal BMI and age at menarche with adult measures of BMI and percentage body fat was assessed in multivariate linear models. Because the subjects were nearly the same age when both premenarcheal and adult BMI were measured, we did not control for age at either time point. Explanatory capability of the models was assessed by using R2. The interactive effects of premenarcheal BMI and age at menarche on adult BMI and percentage body fat were visualized with colorimage plots. An interaction between 2 continuous variables is difficult to visualize; there are no standard visualization techniques available to epidemiologists to aid in understanding this fundamental construct. Therefore, we introduced a color-image plot to capture the 3 variables in a single display using the vertical and horizontal axes for predictor variables (premenarcheal BMI and age at menarche, respectively) and a diverging 2-hue (blue-to-pink) color scheme for either of the 2 outcome variables: log-transformed adult BMI or adult percentage body fat. The image plots use linear interpolation in neighboring data points. This type of interpolation works well because of the distributions of the predictor variables: premenarcheal BMI and age at menarche had substantial variability and were essentially normal. Image plots were produced by using S-Plus 6.2 (Insightful, Inc, Seattle, WA). Multiple logistic-regression analysis was used to evaluate the impact of premenarcheal overweight and early menarche on adult overweight and obesity. Data were analyzed by using SAS 8.2 (SAS Institute, Cary, NC).
Table 1 provides descriptive data for the 793 subjects studied in childhood and compares those who completed the adult follow-up questionnaire with those who did not. Age, weight, height, and BMI at menarche did not differ between those subjects who participated in the adult follow-up and those who did not. Also tabulated are descriptive statistics for the subset of subjects who participated in the DXA substudy. Childhood anthropometric measures were similar among subjects followed in adulthood in both the main study and DXA substudy.
Of the 448 subjects who responded to the adult questionnaire, 307 subjects (69%) had at least 1 premenarcheal and 1 menarcheal height and weight measurement, age at menarche, and an adult follow-up report of weight and height documented (Table 2). Of the 153 women who participated in the DXA substudy, 101 (66%) had recorded at least 1 premenarcheal and menarcheal height and weight measurement, age at menarche, and percentage body fat by DXA. As suggested by Table 2, the subset of subjects who participated in the DXA substudy were representative of the larger sample subjects who participated in the adult follow-up.
The subjects in the main study were 12 years of age for the premenarcheal measures, 13 years of age at the menarcheal time period, and 42 years of age at the adult follow-up (Table 2). The interquartile range for age at menarche was 12.2, 13.7, with 20% of girls experiencing menarche before age 12. The prevalence of premenarcheal overweight was 4.2%. Based on self-reported height and weight, the average adult BMI was 23.43; 27.8% of subjects were overweight (BMI > 25), and 8.7% were obese (BMI > 30). Similar estimates of these parameters were observed among participants in the DXA substudy.
Age at menarche was weakly negatively correlated with premenarcheal BMI (r = −0.10; P = .08) and with log-transformed BMI at the adult follow-up (r = −0.15; P = .01). Premenarcheal BMI was moderately correlated with log-transformed adult BMI (r = 0.45; P = .0001). In a multivariate linear-regression analysis, adult BMI on the log scale was best explained by a regression model that included premenarcheal BMI, age at menarche, and their interaction (Table 3, Model 4). The total R2 for this model was 0.22. Although age at menarche was statistically significant (P = .042), inclusion of this variable provided very little explanatory power.
The interactive effect of premenarcheal BMI and age at menarche on log-transformed adult BMI are visualized in Fig 1 A. Among the leanest girls, none experienced early menarche. The highest level of predicted adult BMI (pink) is observed when age at menarche is between 12 and 14 years and premenarcheal BMI between 21 and 22 kg/m2. When scanning vertically along the range of premenarcheal BMI, the blue-to-pink color gradient is distinct, with the deep-pink color indicating elevated adult BMI seen among girls with higher premenarcheal BMI. When scanning horizontally along the menarcheal age range, the color gradient is less apparent. The image plot indicates a strong relationship of adult BMI with premenarcheal BMI.
Logistic regression was used to evaluate the impact of premenarcheal BMI and age at menarche on the likelihood of adult overweight and adult obesity. Girls who were overweight before menarche were 7.7 times more likely to be overweight as adults (95% confidence interval: 2.3, 25.9). Neither early menarche nor the interaction of early menarche with premenarcheal overweight were statistically significant predictors of adult overweight. Inclusion of early menarche (before age 12) in the model did not influence the estimate of risk associated with overweight before menarche. A similar pattern of results was observed for adult obesity (Table 4).
These analyses were repeated for the 101 participants in the DXA substudy with percentage body fat as the outcome of interest. Age at menarche was weakly negatively correlated with premenarcheal BMI (r = −0.17; P = .10) and with percentage body fat at the adult follow-up (r = −0.17; P = .09). Premenarcheal BMI was moderately correlated with adult percentage body fat (r = 0.53; P = .0001). Multivariate linear-regression modeling revealed that models 3 and 4 (Table 5) have essentially the same and highest R2 value (R2 = 0.28). The interaction term for premenarcheal BMI and age at menarche was not significant, and its inclusion has no impact on the model R2 value. Figure 1B graphically displays these relations. The effect of premenarcheal BMI and age at menarche on percentage body fat is much sharper than that for adult BMI (Fig 1A). The highest level of predicted adult percentage body fat is observed over essentially the entire range of age at menarche and when premenarcheal BMI is >19 kg/m2. When scanning vertically along the range of premenarcheal BMI, the color gradient is very distinct, with a strong gradient from low to high adult percentage body fat as premenarcheal BMI increases. When scanning horizontally across the menarcheal age range, a color gradient is not apparent. The higher adult percentage body fat is evident across the full range of menarcheal ages. This image plot and modeling results indicate that adult percentage body fat is accounted for by premenarcheal BMI, not by age at menarche.
The explosion of obesity in both the pediatric and adult populations has spawned a renewed interest in the relation of childhood growth factors, such as early weight and maturational timing, to the subsequent development of obesity. Our prospectively collected childhood data provide the natural, and therefore appropriate, temporal sequence for an investigation of the relative impact of maturational timing and childhood weight status on adult weight and fatness. We find that weight status before menarche is far more influential than maturational timing on midlife weight status, overweight, and body fatness. The issue of the temporal sequence is critical to establishing the relative contribution of these childhood growth factors; much of the previous research in this area has been limited by failure to establish a clear time sequence.
In a review of age at menarche across the last century, Tanner25 estimated that the decline in menarcheal age was 4 months per decade. Improved nutritional status, reflected by “better” growth, is assumed to be responsible. In 1971, Frisch and coworkers9,26 proposed that menarche was triggered when a critical level of fatness was reached. The strongest evidence for the critical fat hypothesis came 30 years after it was advanced, with the observation that reproductive cycles were initiated in immature rodents injected with leptin, a hormone tightly correlated with body fat.27 An alternative explanation was advanced by Ellison,10 who suggested that skeletal maturity accounted for the induction of hormones that trigger menstrual cycling. Regardless of the physiologic mechanism involved, the inverse relation between body size, as expressed by BMI or fatness, and age at menarche has been consistently observed in human populations.11,28,29
In their comprehensive reviews of the persistence of childhood overweight into adulthood, Serdula et al30 and Power et al31 conclude that the tracking of weight status is moderate, with a two- to sixfold increase in the likelihood of adult obesity given elevated weight in childhood. Our estimates of relative risks of 7.7 and 4.3 for adult overweight and obesity, respectively, are consistent with these prior estimates.
Several investigations have reported that early menarche is associated with adult overweight or elevated BMI.6,11–16,32–39 Many are cross-sectional studies of adolescents or adults in which current weight status or fatness is related to recalled age at menarche.11,15,16,38,39 For example, a recent report from the second wave of the National Longitudinal Study of Adolescent Health reported an almost twofold increase in the likelihood of the prevalence of overweight (>85th percentile BMI) for early maturing girls.11 No conclusions about the direction of these relations or their implications can be drawn from these cross-sectional investigations, however. Few longitudinal studies have been conducted, and fewer relate childhood measures to relative weight in adulthood.6,12,13,36 In the Amsterdam Growth and Health Study, van Lenthe et al12 demonstrated significantly higher BMI levels at age 27 for early maturing girls compared with late maturing girls, based on either age at menarche or skeletal age. Examination of their data, however, indicates that the differences in BMI were already apparent at the initial age-13 measurement, as expected given the relation between elevated relative weight and early menarche. The large and comprehensive 1958 British born cohort data also indicate that BMI at ages 7 and 11 are moderately correlated with adult levels and that earlier menarche is associated with higher adult BMI; their relative contributions, however, were not evaluated.6 Only 1 previous report has explicitly accounted for childhood obesity in the relation of menarcheal age to adult obesity. In the Bogalusa Heart Study, Freedman et al13 estimated that 60% to 75% of the apparent effect of recalled menarcheal age on obesity at age 26 among participants in the Bogalusa Heart Study was a result of childhood obesity. Consistent with our findings, Freedman et al concluded that most of the apparent effect of menarcheal age on adult BMI could be attributed to its association with childhood obesity.
Part of the complexity of these relations may derive from the crudeness of the measures used; BMI z score (used to estimate overweight) and BMI provide only indirect measures of childhood and adult overweight. We find a complex relationship between menarcheal age, childhood BMI, and adult BMI, with evidence of a weak interaction between premenarcheal BMI and menarcheal timing. As demonstrated visually in the image plot, girls who had higher BMI before menarche and experienced menarche between the ages of 12 and 14 years, had the highest BMI as adults. In contrast, the analysis of percentage body fat, a “direct” measure of fatness, is more straightforward. These analyses show clearly that adult body fatness echoes childhood body size. Once premenarcheal BMI has been accounted for, age at menarche provides no additional information. It is possible that the interaction observed in the analysis that relies on childhood BMI is an artifact that arises because BMI does not directly reflect fatness.
The strengths of the current undertaking are substantial. Prospectively collected growth data, with multiple measures during childhood and adolescence, including age at menarche are exceedingly rare. Thirty years is an extremely long duration of follow-up and covers the adult years in which women tend to gain weight. Additionally, the DXA substudy provides a criterion measure of fatness as well as validation of our adult BMI measure based on self-report. Finally, prospectively collected age at menarche reduces measurement error that is likely present in studies that rely on recalled age at menarche.17
Several limitations are noteworthy. Our characterization of childhood weight status is limited to BMI. Additional clarity in the relations observed would be anticipated if a direct fatness measure before menarche were also available. Although self-reported height and weight in adults is fairly accurate,40 it is subject to some differential bias. To the extent that BMI is systematically underestimated, the effects estimated in the logistic models that predict adult overweight and obesity would be biased toward 1, the null value for the odds ratio. The study population was predominantly white and of middle or upper socioeconomic status. The prevalence of adult overweight and obesity of 28% and 9%, respectively, estimated in the current study is substantially lower than national estimates but is as expected given the demographic characteristics of the study sample. A large longitudinal study of an ethnically and socioeconomically diverse sample with direct measures of fatness and follow-up from before puberty well into adulthood is needed but would be extremely expensive, in terms of both cost and time.
Our attempt to separate the joint effects of elevated premenarcheal weight status and maturational timing is hampered to some extent by the use of BMI, which reflects adiposity best at the upper extremes of BMI. Additionally, because BMI increases over the childhood ages studied, an early-maturing girl will have a higher BMI than a late-maturing girl of the same age. Thus, some of the maturational timing effects are likely to be reflected in the BMI level itself. An improved noninvasive measure of childhood relative weight status that is independent of maturation represents a pressing methodologic research challenge.
The relation between early maturation and later obesity is largely explained by its association with elevated premenarcheal weight status. Stated simply, early menarche per se is, at best, a weak risk factor for later overweight. Much of the extant literature has overstated the importance of maturational timing in the natural history of obesity. In contrast, the long-term tracking of overweight from childhood is demonstrable and underscores the importance of both preventive and treatment efforts beginning well before adolescence.
Financial support was provided by the Massachusetts Department of Public Health Breast Cancer Research Program.
We thank the Epidemiology Core of the Boston Obesity Research Center under the direction of Graham Colditz for assistance in the development of the questionnaire for the study; Diane Reidy for assistance with location of study subjects; Dr Susan Harris for contributions to the dual energy x-ray absorptiometry substudy; and staff of the Metabolic Research Unit at the US Department of Agriculture’s Jean Mayer Human Nutrition Research Center on Aging at Tufts University.
- Accepted December 21, 2004.
- Address correspondence to Aviva Must, PhD, Department of Public Health and Family Medicine, Tufts University School of Medicine, 136 Harrison Ave, Boston, MA 02111. E-mail:
No conflict of interest declared.
- ↵World Health Organization. Global strategy on diet, physical activity and health: obesity and overweight. Available at: www.who.int/dietphysicalactivity/publications/facts/obesity/en. Accessed December 1, 2004
- ↵Power C, Lake JK, Cole TJ. Body mass index and height from childhood to adulthood in the 1959 British born cohort. Am J Clin Nutr.1997;66 :1094– 1101
- ↵Frisch RE, Revelle R. Height and weight at menarche and a hypothesis of menarche. Arch Dis Child.1971;48 :695– 701
- ↵van Lenthe FJ, Kemper HCG, van Mechelen W. Rapid maturation in adolescence results in greater obesity in adulthood: the Amsterdam Growth and Health Study. Am J Clin Nutr.1996;64 :18– 24
- ↵Must A, Phillips SM, Naumova EN, et al. Recall of early menstrual history and menarcheal body size: after 30 years, how well do women remember? Am J Epidemiol.2002;155 :672– 679
- ↵Blum M, Must A, Harris SS, Rand WM, Phillips SM, Dawson-Hughes B. Association of documented weight at menarche with premenopausal bone mineral density. J Bone Miner Res.1999;14 :S379
- ↵Centers for Disease Control and Prevention. CDC growth charts: United States. Available at: www.cdc.gov/growthcharts. Accessed February 14, 2001
- ↵Tanner JM. Growth at Adolescence: With a General Consideration of the Effects of Hereditary and Environmental Factors Upon Growth and Maturation From Birth to Maturity. Second ed. Oxford, United Kingdom: Blackwell Scientific Publishers; 1962
- ↵Frisch RE, McArthur JW. Menstrual cycles: fatness as a determinant of minimum weight for height necessary for their maintenance or onset. Science.1974;185 :949– 951
- Garn SM, LaVelle M, Rosenberg KR, Hawthorne VM. Maturational timing as a factor in female fatness and obesity. Am J Clin Nutr.1986;43 :879– 883
- ↵Laitinen J, Power C, Marjo-Riitta J. Family social class, maternal body mass index, childhood body mass index, and age at menarche as predictors of adult obesity. Am J Clin Nutr.2001;74 :287– 294
- Copyright © 2005 by the American Academy of Pediatrics