Content uploaded by Anette E Buyken
Author content
All content in this area was uploaded by Anette E Buyken
Content may be subject to copyright.
Birth and early life influences on the timing of puberty onset: results
from the DONALD (DOrtmund Nutritional and Anthropometric
Longitudinally Designed) Study
1–3
Nadina Karaolis-Danckert, Anette E Buyken, Antje Sonntag, and Anja Kroke
ABSTRACT
Background: Early age at puberty onset may predispose an indi-
vidual to many currently prevalent diseases, including cancer and
adiposity.
Objective: The objective was to investigate whether early life ex-
posures influence the timing of puberty, as defined by both early and
late markers, in healthy German girls and boys.
Design: Term participants (n= 215; 49.8% female) of the DONALD
(DOrtmund Nutritional and Anthropometric Longitudinally De-
signed) Study, with sufficient repeated anthropometric measurements
between 6 and 13 y to allow estimation of age at take-off of the
pubertal growth spurt (ATO) and information on a variety of early
life exposures, including birth weight, breastfeeding status, velocity
of weight gain, and parental characteristics, were studied. Age at
peak height velocity (APHV) and menarche were also considered.
Results: Children who weighed between 2500 and ,3000 g at birth
were ’7 mo younger at ATO than were the other children (b6SE:
20.56 60.20 y; P= 0.006). Children who had gained weight
rapidly between birth and 24 mo (increase in weight SD score
.0.67) experienced ATO 4 mo earlier than those who had gained
weight normally (20.34 60.15 y; P= 0.02). Rapid weight gain
was also associated with an earlier APHV (P= 0.0006) and, in girls,
with an earlier menarche (P= 0.002). Adjustment for body mass
index SD score or body fat percentage 1, 2, or 3 y before ATO did
not account for these effects.
Conclusion: In both boys and girls, intrauterine and early postnatal
growth factors appear to influence both early and later markers
of puberty onset independently of prepubertal body composi-
tion. Am J Clin Nutr 2009;90:1559–65.
INTRODUCTION
Early age at puberty onset may be an intermediary factor on the
life-course path to many currently prevalent diseases, including
both breast (1, 2) and testicular (3) cancer, insulin resistance (4),
and adiposity (5). To date, many studies have identified both pre-
and perinatal exposures, such as low birth weight (6) or rapid
growth velocity in infancy (7) as being potential determinants of
pubertal timing. However, many issues remain unresolved.
These studies have mainly focused on girls and, in particular,
on the timing of menarche—a relatively late milestone of re-
productive development that usually takes place after peak height
velocity (PHV) has been achieved and height growth has begun to
slow down. Nevertheless, it has been concluded that an earlier
menarche implies a more rapid progression through puberty (8).
It is, however, possible that early life factors actually influence
when puberty begins and not necessarily the duration. In addition,
influence of a particular early life exposure on both early and late
markers of pubertal timing would support its relevance, but this
has rarely been investigated. Previous studies have also proved
inconclusive about the role of size and body composition in later
childhood. Whereas some have shown no effect of birth size
once childhood growth patterns were accounted for (9, 10), others
have shown additive effects of pre- and postnatal growth (7, 11,
12). Finally, the relevance of early life risk factors for pubertal
timing in boys is unclear, mainly because of the lack of an easily
identifiable puberty marker.
In this study, we used the prospectively collected height
measurements of participants from the DONALD (DOrtmund
Nutritional and Anthropometric Longitudinally Designed) Study
to investigate the association between early life exposures and
both an early marker of puberty onset [ie, age at take-off of the
pubertal growth spurt (ATO)] and 2 later ones [ie, age at PHV
(APHV) and menarche (in girls)]. In addition, we investigated
whether these early life exposures exert their influence on pu-
bertal timing independently of or via a pathway related to pre-
pubertal body composition.
SUBJECTS AND METHODS
Study population
The DONALD Study is an ongoing, open cohort study con-
ducted by the Research Institute of Child Nutrition in Dortmund,
Germany. This study was previously described in detail (13).
Briefly, since recruitment began in 1985, detailed information on
diet, growth, development, and metabolism between infancy and
1
From the Research Institute of Child Nutrition, Rheinische Friedrich-
Wilhelms–Universita
¨t Bonn, Dortmund, Germany (NK-D, AEB, and AS),
and the Department of Nutritional, Food and Consumer Sciences, Fulda
University of Applied Sciences, Fulda, Germany (AK).
2
Supported by a research grant from the World Cancer Research Fund
UK (WCRF UK no.2006/39). The DONALD Study is funded by the Min-
istry of Science and Research of North Rhine Westphalia, Germany.
3
Address correspondence to A Kroke, Department of Nutritional, Food
and Consumer Sciences, Fulda University of Applied Sciences, Marquard-
strasse 35, 36039 Fulda, Germany. E-mail: anja.kroke@he.hs-fulda.de.
Received June 19, 2009. Accepted for publication September 17, 2009.
First published online October 14, 2009; doi: 10.3945/ajcn.2009.28259.
Am J Clin Nutr 2009;90:1559–65. Printed in USA. Ó2009 American Society for Nutrition 1559
by guest on June 22, 2013ajcn.nutrition.orgDownloaded from
early adulthood has been collected from .1100 children. Every
year, an average of 40 to 50 infants are newly recruited and first
examined at the ages of 3 or 6 mo. Each child returns for 3 more
visits in the first year, 2 in the second and then once annually
until early adulthood (except during adolescence when twice-
yearly visits were offered until 2004). The study was approved
by the Ethics Committee of the University of Bonn, and all
examinations are performed with parental consent.
The ages of the children who were initially recruited into the
DONALD Study when it began in 1985 were quite variable, ie,
information on the first few years of life was not always available.
In addition, many current participants have not yet reached
adolescence. Therefore, for this analysis, only term (37–42 wk of
gestation) singletons with a birth weight .2500 g who fulfilled
the following minimum requirement were selected (n= 411):
they had already provided height measurements at 6 and 13 y
and 5 measurements between these ages to allow estimation of
ATO by using Preece and Bains model 1 (PB1) (14) (see de-
scription below). The plausibility of each child’s ATO was de-
termined by visual inspection of each individual growth curve
and by using the cutoffs ATO 5 and ,13 y. Reasons for an
implausible ATO included no measurements after age 13 y, too
few measurements between age 13 y and young adulthood, or an
unusually “flat” growth curve. A plausible ATO was achieved
for 376 children. Of these, 215 also had anthropometric meas-
urements at 24 mo (for determination of rapid weight gain be-
tween birth and 24 mo) and complete information on
breastfeeding and on maternal characteristics (BMI and educa-
tional status). Hence, the subcohort analyzed here included 215
term singletons (49.8% female).
Anthropometric measures
DONALD Study participants are measured at each visit by
trained nurses according to standard procedures (15). They are
dressed in underwear only and are barefoot. From the age of 2
onward, standing height is measured to the nearest 0.1 cm with
a digital stadiometer. Weight is measured to the nearest 0.1 kg
with an electronic scale (model 753 E; Seca, Hamburg, Ger-
many). Skinfold thicknesses are measured from the age of 6 mo
onward on the right side of the body to the nearest 0.1 mm with
a Holtain caliper (Holtain Ltd, Crymych, United Kingdom).
On their child’s admission to the study, parents are interviewed
by the study pediatrician, and weighed and measured by the study
nurses using the same equipment as for children from 2 y onward.
Information on birth weight, length and head circumference at
birth, and gestational age are abstracted from a standardized
document given to all pregnant women in Germany.
For this analysis, sex- and age-independent SD scores (SDS)
were calculated by using the German reference curves for weight
and body mass index (BMI; in kg/m
2
) (16). To remove general
deviations of our sample from the reference data, these variables
were internally standardized (mean = 0, SD = 1; by age and sex).
Percentage body fat (BF%) was calculated by using Slaughter
equations for prepubertal children (17).
Puberty outcome variables
Height data were analyzed by using the parametric PB1 (14).
The parameters of each child’s growth curve were estimated by
using a nonlinear regression model (PROC NLIN in SAS; SAS
Institute Inc, Cary, NC). It was previously shown that the lower
limit of the age range offered to the model should not be ,2yof
age and that the fit of the adolescent growth curve is better if the
age range includes data from not more than a few years before
ATO (18). PB1 was therefore fitted on various sex-specific age
ranges of the height-for-age data, beginning with age 2 y, to
determine the optimal range for our data. ATO was defined as the
age at minimal height velocity (zero acceleration) at the onset of
the pubertal growth spurt (19). Goodness of fit was determined as
follows: 1) by graphic inspection of each child’s individual
growth curve, 2) by comparison of the residual SD (random error
had to be smaller than the expected measurement error for
height), 3) by consideration of the plausibility, and 4) on the basis
of the distribution of the pubertal parameters estimated (20).
Consequently, all available measurements from age 5 y onward
for girls and from age 6 y onward for boys were used. The mean
number of measurements per child was 17.7 (range: 9–21) in
girls and 15.7 (range: 8–20) in boys. The PB1 model also pro-
duced estimates of velocity at take-off (VTO), APHV, and PHV.
Two other puberty variables were also considered: age at
Tanner stage 2 for either breast or penis development in girls (n=
75) and boys (n= 66), respectively, and age at menarche in girls
only. Tanner stage is visually assessed by a study pediatrician. In
addition, girls or their parents are asked whether menarche has
occurred since the previous visit, and, if so, in which month and
year. This information was available for 87 of the 107 girls in
this analysis. Of the remaining 20, 2 had incomplete information
on menarche (ie year but not month in which menarche began),
and 18 had not provided any information.
Early life exposures
Birth weight was considered as both a continuous and cate-
gorical variable (ie, 2500 to ,3000 g compared with 3000
g). Birth size-for-gestational-age was defined by using the
German sex-specific birth weight and length-for-gestational-age
curves (21). Small-for-gestational-age (SGA) was defined as
birth weight and length ,10th percentile, and large-for-gesta-
tional-age (LGA) was defined as birth weight and length .90th
percentile. All other infants were classified as appropriate-for-
gestational age (AGA). Other birth variables included year and
season of birth, birth at early (37 or 38 wk) or late (41 or 42 wk)
gestation, first-born status (yes or no), and breastfeeding status,
defined as full breastfeeding for 4 mo. Because rapid early
weight gain has been linked to elevated insulin-like growth
factor I concentrations and insulin resistance, elevated adrenal
androgen concentrations, exaggerated adrenarche, obesity, and
consequently to concentrations of hormones such as leptin, it
has been suggested that these could all promote the activity of
the gonadotropin-releasing hormone pulse generator, thereby
influencing the timing of puberty (22). These effects and the
possible pathway are still unclear. On the basis of this rationale,
the possible pathway between early rapid weight gain, over-
weight in childhood and puberty onset, and sample size con-
siderations, it was decided to follow Monteiro and Victora’s
recommendation in their systematic review (23) and define rapid
early weight gain as an increase in weight SDS .0.67 between
birth and 24 mo. Because a considerable number of siblings
participate in the DONALD Study, we also took the presence of
1560 KARAOLIS-DANCKERT ET AL
by guest on June 22, 2013ajcn.nutrition.orgDownloaded from
siblings in our subcohort (yes or no) into account. Smoking
exposure in the household (yes or no) and the parental charac-
teristics overweight status (BMI 25) and educational status
(12 y of schooling) were also considered.
Statistical analysis
Three ATO groups—early (,25th percentile), middle (25th
percentile and 75th percentile), and late onset (.75th per-
centile)—were created by using the sex-specific distribution of
ATO. Differences between ATO groups were tested by using
analysis of variance or Kruskal-Wallis tests (continuous varia-
bles) and chi-square tests (categorical variables). There was no
interaction between sex and birth weight category (Pfor in-
teraction = 0.5) or sex and rapid weight gain (Pfor interaction =
0.2), so boys and girls were pooled together for all statistical
analyses. Linear mixed-effects regression models (using PROC
MIXED), including both fixed and random effects, were used to
construct longitudinal models of ATO, APHV, or age at men-
arche (in girls only). The random component of these models
accounts for the nested nature of our data (children within
families). Initial models included ATO, APHV, or age at men-
arche (in girls only) as the dependent continuous variable and
the various early life exposures as the independent fixed effects.
All models (except those with age at menarche as the outcome)
included sex, and each early life exposure was initially consid-
ered separately. Variables were only retained in the multivariable
models if they tended to be associated with the outcome (P,
0.1). Next, all multivariable models were adjusted for BMI SDS
or BF% 1, 2, and 3 y before ATO as an indication of whether
early life exposures exert their effect on the respective outcome
via a pathway that involves changes in prepubertal body com-
position (our so-called “pathway model”). In a final step, in-
teractions between variables found to influence the outcomes of
interest were investigated to identify particularly vulnerable
subgroups. A Pvalue ,0.05 was considered statistically sig-
nificant. All statistical analyses were carried out by using SAS
version 8.2 (SAS Institute).
RESULTS
Early life and familial characteristics of the 215 boys and girls
in this analysis are presented in Table 1 according to ATO group.
A significantly larger proportion of those children whose pu-
berty growth spurt began relatively early were lighter at birth
(between 2500 and ,3000 g) and had gained weight rapidly
between birth and 24 mo compared with those in the middle or
late ATO groups (P= 0.04 and P= 0.006, respectively).
The pubertal characteristics of the study sample, stratified by
both sex and ATO group, are shown in Table 2. In both boys and
girls, those children in the early ATO group displayed an in-
creased velocity at take-off and experienced PHV and menarche
TABLE 1
Early life and familial characteristics of 215 boys and girls from the DONALD (DOrtmund Nutritional and Anthropometric Longitudinally Designed) Study,
by sex-specific groups of age at take-off of the pubertal growth spurt (ATO)
ATO group
1
Variable No. of subjects Early (n= 53) Middle (n= 108) Late (n= 54) Pvalue
2
Female sex [n(%)] 215 26 (49.1) 54 (50.0) 27 (50.0) 0.9
Early life characteristics
Birth weight (g) 215 3402 6456
3
3532 6464 3596 6465 0.1
Birth weight ,3000 g [n(%)] 215 11 (20.8) 11 (10.2) 3 (5.6) 0.04
Birth length (cm) 215 52 (50, 53)
4
52 (50, 53) 52 (51, 54) 0.2
Gestational age (wk) 215 40 (39, 41) 40 (39, 41) 40 (39, 41) 0.9
Birth year [n(%)] 215 0.9
,1987 52 11 (20.8) 26 (24.1) 15 (27.8)
1987–1990 96 24 (45.3) 47 (43.5) 25 (46.3)
.1990 67 18 (34.0) 35 (32.4) 14 (25.9)
Winter/spring birth [n(%)] 215 23 (43.4) 54 (50) 22 (40.7) 0.5
Fully breastfed 4mo[n(%)] 215 25 (47.2) 50 (46.3) 32 (59.3) 0.1
Rapid weight gain [n(%)] 215 22 (41.5) 22 (20.4) 10 (18.5) 0.006
Family characteristics [n(%)]
Firstborn 206 31 (63.3) 58 (55.2) 30 (57.7) 0.6
Siblings in data set 215 13 (24.5) 32 (29.6) 14 (25.9) 0.8
Smoking exposure at home 215 18 (34.0) 32 (29.6) 17 (31.5) 0.9
Maternal characteristics [n(%)]
Overweight 215 18 (34.0) 35 (32.4) 11 (20.4) 0.2
12 y of schooling 215 29 (54.7) 58 (53.7) 35 (64.8) 0.4
Paternal characteristics [n(%)]
Overweight 177 22 (51.2) 49 (56.3) 30 (63.8) 0.5
12 y of schooling 214 33 (62.3) 62 (57.9) 35 (64.8) 0.7
1
ATO groups were defined as follows: early (ATO ,25th sex-specific percentile), middle (ATO 25th and 75th sex-specific percentile), and late (ATO
.75th sex-specific percentile).
2
Significant differences between groups were tested by ANOVA for normally distributed continuous variables, Kruskal-Wallis tests for nonnormally
distributed continuous variables, and a chi-square test for categorical variables.
3
Mean 6SD (all such values).
4
Median; quartiles 1 and 3 in parentheses (all such values).
EARLY LIFE EXPOSURES AND PUBERTY ONSET 1561
by guest on June 22, 2013ajcn.nutrition.orgDownloaded from
(girls) at a younger age than the children in the other 2 ATO
groups (P,0.0001 for all 3 variables). In girls only, those in the
early ATO group also reached Tanner stage 2 at a younger age
than did those girls in the other 2 ATO groups (P,0.0001).
To gain an understanding of the clinical relevance of the effect
of those early life exposures, which remained of relevance in the
multivariable models on puberty onset, the association between
ATO, APHV, or age at menarche and these exposures was ex-
amined in linear mixed-model regression analyses, and the results
are presented as the time difference (in y) in ATO, APHV, or age
at menarche between exposed and unexposed children. Table 3
(multivariate model) shows that children who weighed between
2500 and ,3000 g at birth were younger (b6SE: 0.56 60.20
y, ie, ’7 mo) at ATO than were those children whose birth
weight was 3000 g (P= 0.006), whereas those children who
had gained weight rapidly between birth and 24 mo experienced
ATO 0.34 60.15 y (ie, 4 mo) earlier than did those who had
gained weight normally in the first 2 y of life (P= 0.02). In
initial models it appeared that children who weighed between
2500 and ,3000 g at birth also experienced both APHV and
menarche earlier than did the other children, but these effects
were attenuated in the multivariable analyses (Tables 4 and 5).
This was also the case for the effect of having been fully
breastfed for 4 mo on APHV (these children tended toward
a delayed APHV by 0.24 60.13 y, or ’3 mo in initial models).
Rapid weight gain, however, was associated with both an earlier
APHV (P= 0.0006) and an earlier menarche (P= 0.002). Fi-
nally, the data in Table 5 also suggest that having an overweight
mother resulted in these girls experiencing menarche 0.52 6
0.24 y or 6 mo earlier than those girls without an overweight
mother (P= 0.03).
TABLE 2
Pubertal characteristics of 107 girls and 108 boys from the DONALD (DOrtmund Nutritional and Anthropometric Longitudinally Designed) Study,
Dortmund, Germany, by sex-specific groups of age at take-off of the pubertal growth (ATO)
ATO group
1
Variable No. of subjects Early (n= 53) Middle (n= 108) Late (n= 54) Pvalue
2
ATO (y)
Girls 107 7.5 (6.9, 7.8)
3
8.7 (8.3, 9.1) 9.8 (9.4, 10.0) ,0.0001
Boys 108 9.2 (8.9, 9.5) 10.3 (10.0, 10.5) 11.1 (11.0, 11.5) ,0.0001
Velocity at take-off (cm/y)
Girls 107 5.9 60.7
4
5.5 60.5 4.9 60.5 ,0.0001
Boys 108 5.5 60.5 5.1 60.5 4.6 60.5 ,0.0001
Age at peak height velocity (y)
Girls 104 10.3 60.8 11.5 60.6 12.8 60.6 ,0.0001
Boys 104 12.4 60.6 13.4 60.5 14.5 60.6 ,0.0001
Age at Tanner stage 2 (y)
Girls 75 9.5 (9.0, 10.3) 10.5 (9.5, 11.0) 11.5 (11.0, 12.1) ,0.0001
Boys 66 10.5 (10.0, 10.8) 10.6 (10.0, 11.2) 10.5 (10.0, 11.0) 0.4
Age at menarche (y)
Girls 87 11.5 60.8 12.7 60.7 13.9 60.9 ,0.0001
BMI SD score 1 y before ATO
Girls 107 20.05 60.84 0.12 61.00 20.33 61.10 0.2
Boys 108 0.07 61.05 0.05 61.04 20.36 60.90 0.2
Body fat 1 y before ATO (%)
5
Girls 107 18.0 (14.3, 19.7) 16.7 (14.6, 22.6) 16.3 (14.1, 20.5) 0.8
Boys 108 13.7 (12.3, 22.3) 15.7 (13.3, 23.2) 14.8 (12.9, 18.4) 0.4
1
ATO groups were defined as follows: early (ATO ,25th sex-specific percentile), middle (ATO 25th and 75th sex-specific percentile), and late (ATO
.75th sex-specific percentile).
2
Significant differences between groups were tested by ANOVA for normally distributed continuous variables and Kruskal-Wallis tests for nonnormally
distributed continuous variables.
3
Median; quartiles 1 and 3 in parentheses (all such values).
4
Mean 6SD (all such values).
5
Defined by using Slaughter’s equations for prepubertal children (17).
TABLE 3
Linear mixed models of the association between early life factors and age
at take-off of the pubertal growth spurt (ATO) in a DONALD (DOrtmund
Nutritional and Anthropometric Longitudinally Designed) Study sample
(n= 215)
ATO
1
Pvalue
Multivariate model
Sex (reference category: male) 21.57 60.13 ,0.0001
Birth weight category
(reference category: 3000 g)
20.56 60.20 0.006
Rapid weight gain (reference
category: normal weight gain)
20.34 60.15 0.02
Pathway model
Sex (reference category: male) 21.56 60.12 ,0.0001
Birth weight category
(reference category: 3000 g)
20.63 60.20 0.002
Rapid weight gain (reference
category: normal weight gain)
20.27 60.15 0.07
BMI SD score 1 y before ATO 20.13 60.06 0.03
1
Values are bs6SEs. To account for the clustered nature of our data,
“family” was modeled as a random effect. This set up a common correlation
between all observations and individuals who belonged to the same family.
1562 KARAOLIS-DANCKERT ET AL
by guest on June 22, 2013ajcn.nutrition.orgDownloaded from
The pathway model in Tables 3–5 shows the effect of in-
troducing BMI SDS 1 y before ATO to the multivariable models,
to investigate the role of prepubertal body composition. In the
case of ATO (Table 3), the inclusion of the BMI SDS variable did
not influence the effect of birth weight, but attenuated the effect
of rapid weight gain. In the APHV (Table 4) and menarche (Table
5) models, inclusion of the BMI SDS variable only marginally
influenced the effect of rapid weight gain. However, the effect of
having a lower birth weight reappeared; these children experi-
enced APHV 20.42 60.20 y (ie, ’5 mo) earlier (P= 0.04) and,
in girls, menarche 0.68 60.29 y (ie, ’8 mo) earlier (P= 0.02)
than did the other children. Similar results were achieved when
BMI SDS 2 or 3 y before ATO or BF% 1, 2, or 3 y before ATO
were used to represent prepubertal body composition. Adjust-
ment for length of gestation did not change the findings de-
scribed above (data not shown).
In a final step, we considered the possible interaction between
birth weight category and weight gain to identify any particularly
susceptible subgroups. As can be seen in Figure 1, those children
with a low birth weight and rapid weight gain in the first 2 y of
life experienced ATO 1.11 60.25 y, APHV 1.14 60.26 y, and
menarche 1.68 60.35 y earlier than did children with a birth
weight 3000 g and normal weight gain between birth and 24
mo (reference group).
DISCUSSION
With the use of prospectively collected growth and puberty
data from 215 boys and girls, we could show that early life
exposures, in particular a relatively low birth weight and rapid
weight gain between birth and 24 mo, were independently as-
sociated not only with an earlier age at PHV or menarche (in
girls), both of which indicate a more advanced stage of pubertal
maturation, but also with an earlier age at take-off of the pubertal
growth spurt—one of the earliest markers of puberty onset. This
was the case in both girls and boys. Furthermore, these associ-
ations were essentially independent of measures of prepubertal
body composition 1, 2, or 3 y before ATO.
The consistent and independent association of both a relatively
low birth weight and rapid weight gain between birth and 24 mo
with an earlier ATO, APHV, and menarche extends the findings of
other studies by showing that these factors are also of relevance
for a very early, relatively unstudied marker of pubertal onset, ie,
ATO. With respect to menarche, our results support the findings
of some, but not all, studies previously reported. Despite their
developing country population, the findings of Adair et al (7) are
the most similar to ours. They showed that those girls who were
long and thin at birth experienced an earlier menarche, and that
this effect of birth size was potentiated by rapid growth in the first
6 mo postnatally (itself an independent determinant of age at
menarche). In our study, investigation of the interaction between
birth weight and weight gain showed that the most vulnerable
group, regardless of the puberty outcome being considered, were
indeed those with a low birth weight who went on to gain weight
rapidly between birth and 24 mo. Two other studies also showed
a possible additive effect of birth weight combined with accel-
erated postnatal growth on age at menarche. The earliest age at
menarche was found in those girls with the lowest birth weight
and highest BMI at age 8 y (11, 12). In contrast, dos Santos Silva
et al (9) initially found opposing effects of prenatal and early
postnatal growth on the timing of menarche, but these effects
disappeared once growth in childhood (2–7 y) was adjusted for.
In this analysis, the effect of the early life exposures remained of
relevance for both early and later markers of puberty, even after
prepubertalBMI SDSor BF%was adjustedfor.The findingsof Adair
et al (7) also remained unmodified by the inclusion of BMI and
skinfold thicknesses at age 8 y. This somewhat contradicts the
suggestion that the timing of menarche, for example, may be set in
TABLE 4
Linear mixed models of the association between early life factors and age
at peak height velocity (APHV) in a DONALD (DOrtmund Nutritional and
Anthropometric Longitudinally Designed) Study sample
(n= 208)
APHV
1
Pvalue
Multivariate model
Sex (reference category: male) 21.85 60.13 ,0.0001
Birth weight category (reference
category: 3000 g)
20.32 60.21 0.1
Breastfed 4 mo (reference
category: no)
0.14 60.13 0.3
Rapid weight gain (reference
category: normal weight gain)
20.54 60.16 0.0006
Maternal BMI 20.03 60.02 0.1
Pathway model
Sex (reference category: male) 21.81 60.13 ,0.0001
Birth weight category
(reference category: 3000 g)
20.42 60.20 0.04
Breastfed 4 mo (reference
category: no)
0.09 60.13 0.5
Rapid weight gain (reference
category: normal weight gain)
20.45 60.15 0.004
Maternal BMI 20.01 60.02 0.5
BMI SD score 1 y before ATO 20.22 60.07 0.002
1
Values are bs6SEs. To account for the clustered nature of our data,
“family” was modeled as a random effect. This set up a common correlation
between all observations and individuals who belonged to the same family.
TABLE 5
Linear mixed models of the association between early life factors and age
at menarche in a DONALD (DOrtmund Nutritional and Anthropometric
Longitudinally Designed) Study subsample (n= 87)
Age at menarche
1
Pvalue
Multivariate model
Birth weight category
(reference category: 3000 g)
20.49 60.29 0.1
Rapid weight gain (reference
category: normal weight gain)
20.82 60.25 0.002
Maternal overweight (reference
category: no)
20.52 60.24 0.03
Pathway model
Birth weight category (reference
category: 3000 g)
20.68 60.29 0.02
Rapid weight gain (reference
category: normal weight gain)
20.60 60.26 0.02
Maternal overweight (reference
category: no)
20.42 60.23 0.07
BMI SD score 1 y before ATO 20.28 60.12 0.02
1
Values are bs6SEs. To account for the clustered nature of our data,
“family” was modeled as a random effect. This set up a common correlation
between all observations and individuals who belonged to the same family.
EARLY LIFE EXPOSURES AND PUBERTY ONSET 1563
by guest on June 22, 2013ajcn.nutrition.orgDownloaded from
utero but modified by changes in body size and composition in
childhood (9) and, instead, suggests that early life factors could
influence the timing of puberty via a pathway other than the one that
leads via increased fat mass (24). We should point out, however, that
we used BMI and BF% estimated from skinfold-thickness meas-
urements to represent prepubertal body composition. We cannot
thereforeexclude the possibility thatthese crude markers of fat mass
concealed a true role of prepubertal body composition.
It has been suggested that lower birth weight results in an
earlier, more rapid progression through puberty because those
with the lowest birth weights reached menarche, which com-
monly occurs after PHV, earlier (8). However, by considering the
effect of early life exposures on both early and late pubertal
markers, we showed that one of the earliest indicators of pubertal
onset, ATO, occurs earlier too. This suggests that these children
do not necessarily experience a more rapid progression through
puberty, but rather a general shift to the left of pubertal de-
velopment as a whole, so that everything occurs at a younger age.
Lazar et al (25) showed a similar phenomenon but used the
association between Tanner stage, as a marker of puberty onset,
and menarche.
A significant degree (50–80%) of variability in pubertal timing
has been attributed to genetic differences between individuals
(26). Not being able to directly adjust for such factors (eg,
mother’s age at menarche) may therefore be seen as a limitation
of this analysis. However, the secular trend that has been observed
over the past 200 y, ie, a decrease in the age at menarche both in
the United States (27) and in Europe (28), suggests that factors
other than genetics must also have a role to play because the gene
pool alone has not really had a sufficient time to respond (24).
The strengths of this study include the availability of detailed,
prospectively collected height measurements from birth until
early adulthood in a sample of both boys and girls. Consequently,
both early and later markers of puberty could be estimated on the
basis of the pubertal growth curve, and we were not reliant on
either recall of pubertal events or more subjective markers of
puberty such as Tanner stage (29). Also, the consistency in the
results shown between the different pubertal variables clearly
supports their accuracy. The fact that we included birth variables,
as well as growth in infancy and prepubertal body composition,
meant that we could more closely investigate the role of each of
these components in pubertal timing. Finally, we could also
adjust for many potential confounders related to parental char-
acteristics and socioeconomic status.
The public health relevance of a difference in age at pubertal
onset of 4 to 7 mo remains to be addressed. Early age at puberty is
an established risk factor for many hormone-related cancers,
including breast (1, 2) and testicular cancer (30, 31), and has also
been linked to insulin resistance (4), higher insulin-like growth
factor I concentrations, and other hormonal changes associated
with cancer risk, obesity (32), and, more recently, mortality (33).
If one takes the example of breast cancer, the most common
cancer in women, a meta-analysis of 26 epidemiologic studies
showed a risk reduction of 9% for every additional year at
menarche (34). Breast cancer was diagnosed in more than
a million women worldwide in 2002 (35), so a difference in age at
menarche of ’6 mo could result in a risk reduction of 4.5% or
40,000 fewer cases, provided steps were taken to prevent low
birth weight or to monitor growth in infancy.
In this study we have identified early life factors that increase
a child’s risk of beginning puberty early. Furthermore, we have
shown that these factors act independently of prepubertal body
composition. These findings have important implications,
FIGURE 1. Least-squares mean (6SE) differences in age at take-off of
the pubertal growth spurt (A; n= 215), age at peak height velocity (B; n=
208), and age at menarche in girls only (C; n= 87) between subgroups of
growth velocity and birth weight. Differences are between the reference
group (normal weight gain, birth weight 3000 g; NN) and the other
subgroups: normal weight gain, birth weight between 2500 and ,3000 g
(NL); rapid weight gain, birth weight 3000 g (RN); and rapid weight gain,
birth weight between 2500 and ,3000 g (RL). See Tables 3–5 for
information on the multivariate models used for prediction. Pfor
interaction = 0.1 (A), 0.04 (B), and 0.01 (C). *Significantly different from
the reference group, P,0.05.
1564 KARAOLIS-DANCKERT ET AL
by guest on June 22, 2013ajcn.nutrition.orgDownloaded from
especially in view of the secular trend in pubertal onset. More
studies are needed to identify the mechanisms by which these
factors might operate.
The participation of all children and their families in the DONALD Study is
gratefully acknowledged. We also thank the staff of the Research Institute of
Child Nutrition for carrying out the anthropometric measurements and med-
ical examinations.
The authors’ responsibilities were as follows—NK-D and AK: conceived
the research project and acquired the funding; NK-D and AS: performed the
initial statistical analyses; NK-D: performed further analyses and drafted the
manuscript; and AEB and AK: supervised the project. All authors contributed
to the interpretation of the results and to critical revision of the manuscript.
None of the authors had any personal or financial conflicts of interest.
REFERENCES
1. Berkey CS, Frazier AL, Gardner JD, Colditz GA. Adolescence and
breast carcinoma risk. Cancer 1999;85:2400–9.
2. dos Santos Silva I, De Stavola BL, Hardy RJ, Kuh DJ, McCormack VA,
Wadsworth ME. Is the association of birth weight with premenopausal
breast cancer risk mediated through childhood growth? Br J Cancer
2004;91:519–24.
3. Richiardi L, Askling J, Granath F, Akre O. Body size at birth and
adulthood and the risk for germ-cell testicular cancer. Cancer Epidemiol
Biomarkers Prev 2003;12:669–73.
4. Frontini MG, Srinivasan SR, Berenson GS. Longitudinal changes in risk
variables underlying metabolic Syndrome X from childhood to young
adulthood in female subjects with a history of early menarche: the Bo-
galusa Heart Study. Int J Obes Relat Metab Disord 2003;27:1398–404.
5. Freedman DS, Khan LK, Serdula MK, Dietz WH, Srinivasan SR,
Berenson GS. The relation of menarcheal age to obesity in childhood
and adulthood: the Bogalusa heart study. BMC Pediatr 2003;3:3.
6. Ibanez L, Ferrer A, Marcos MV, Hierro FR, de Zegher F. Early puberty:
rapid progression and reduced final height in girls with low birth weight.
Pediatrics 2000;106:E72.
7. Adair LS. Size at birth predicts age at menarche. Pediatrics 2001;107:E59.
8. Cooper C, Kuh D, Egger P, Wadsworth M, Barker D. Childhood growth
and age at menarche. Br J Obstet Gynaecol 1996;103:814–7.
9. dos Santos Silva I, De Stavola BL, Mann V, Kuh D, Hardy R, Wadsworth
ME. Prenatal factors, childhood growth trajectories and age at menarche.
Int J Epidemiol 2002;31:405–12.
10. Persson I, Ahlsson F, Ewald U, et al. Influence of perinatal factors on the
onset of puberty in boys and girls: implications for interpretation of link
with risk of long term diseases. Am J Epidemiol 1999;150:747–55.
11. Sloboda DM, Hart R, Doherty DA, Pennell CE, Hickey M. Age at
menarche: influences of prenatal and postnatal growth. J Clin Endo-
crinol Metab 2007;92:46–50.
12. Tam CS, de Zegher F, Garnett SP, Baur LA, Cowell CT. Opposing in-
fluences of prenatal and postnatal growth on the timing of menarche.
J Clin Endocrinol Metab 2006;91:4369–73.
13. Kroke A, Manz F, Kersting M, et al. The DONALD Study. History,
current status and future perspectives. Eur J Nutr 2004;43:45–54.
14. Preece MA, Baines MJ. A new family of mathematical models de-
scribing the human growth curve. Ann Hum Biol 1978;5:1–24.
15. WHO. Physical status: the use and interpretation of anthropometry.
Report of a WHO Expert Committee. World Health Organ Tech Rep Ser
1995;854:1–452.
16. Kromeyer-Hauschild K, Wabitsch M, Kunze D, et al. Perzentile fu
¨r den
Body-mass-Index fu
¨r das Kindes- und Jugendalter unter Heranziehung
verschiedener deutscher Stichproben. (Percentiles of body mass index in
children and adolescents evaluated from different regional German
studies.) Monatsschr Kinderheilkunde 2001;149:807–18 (in German).
17. Slaughter MH, Lohman TG, Boileau RA, et al. Skinfold equations for
estimation of body fatness in children and youth. Hum Biol 1988;60:
709–23.
18. Hauspie R, Molinari L. Parametric models for postnatal growth. In:
Hauspie R, Cameron N, Molinari L, eds. Methods in human growth
research. Cambridge, United Kingdom: Cambridge University Press,
2004:205–33.
19. Molinari L, Gasser T. The human growth curve. In: Hauspie R, Cameron
N, Molinari L, eds. Methods in human growth research. Cambridge,
United Kingdom: Cambridge University Press, 2004:27–54.
20. Buyken AE, Karaolis-Danckert N, Remer T. Association of prepubertal
body composition in healthy girls and boys with the timing of early and
late pubertal markers. Am J Clin Nutr 2009;89:221–30.
21. Voigt M, Friese K, Schneider KTM, Jorch G, Hesse V. Kurzmitteilung
zu den Perzentilwerten fu
¨r die Ko
¨rpermaße Neugeborener. (A short
communication on the reference values for body weight and height in
newborns.) Geburtsh Frauenheilk 2002;62:274–6 (in German).
22. Dunger DB, Ahmed ML, Ong KK. Early and late weight gain and the
timing of puberty. Mol Cell Endocrinol 2006;254–255:140–5.
23. Monteiro PO, Victora CG. Rapid growth in infancy and childhood and
obesity in later life: a systematic review. Obes Rev 2005;6:143–54.
24. Cole TJ. Secular trends in growth. Proc Nutr Soc 2000;59:317–24.
25. Lazar L, Pollak U, Kalter-Leibovici O, Pertzelan A, Phillip M. Pubertal
course of persistently short children born small for gestational age
(SGA) compared with idiopathic short children born appropriate for
gestational age (AGA). Eur J Endocrinol 2003;149:425–32.
26. Silventoinen K, Haukka J, Dunkel L, Tynelius P, Rasmussen F. Genetics
of pubertal timing and its associations with relative weight in childhood
and adult height: the Swedish Young Male Twins Study. Pediatrics 2008;
121:e885–91.
27. Herman-Giddens ME, Kaplowitz PB, Wasserman R. Navigating the
recent articles on girls’ puberty in Pediatrics: what do we know and
where do we go from here? Pediatrics 2004;113:911–7.
28. de Muinich Keizer SM, Mul D. Trends in pubertal development in
Europe. Hum Reprod Update 2001;7:287–91.
29. Euling SY, Herman-Giddens ME, Lee PA, et al. Examination of US
puberty-timing data from 1940 to 1994 for secular trends: panel find-
ings. Pediatrics 2008;121(suppl 3):S172–91.
30. United Kingdom Testicular Cancer Study Group. Aetiology of testicular
cancer: association with congenital abnormalities, age at puberty, in-
fertility, and exercise. BMJ 1994;308:1393–9.
31. Garner MJ, Turner MC, Ghadirian P, Krewski D. Epidemiology of
testicular cancer: an overview. Int J Cancer 2005;116:331–9.
32. Guven A, Cinaz P, Ayvali E. Are growth factors and leptin involved in
the pathogenesis of premature adrenarche in girls? J Pediatr Endocrinol
Metab 2005;18:785–91.
33. Jacobsen BK, Oda K, Knutsen SF, Fraser GE. Age at menarche, total
mortality and mortality from ischaemic heart disease and stroke: the
Adventist Health Study, 1976-88. Int J Epidemiol 2009;38:245–52.
34. Clavel-Chapelon F, Gerber M. Reproductive factors and breast cancer
risk. Do they differ according to age at diagnosis? Breast Cancer Res
Treat 2002;72:107–15.
35. World Cancer Research Fund/American Institute for Cancer Research.
Food, nutrition, physical activity, and the prevention of cancer: a global
perspective. Washington, DC: AICR, 2007.
EARLY LIFE EXPOSURES AND PUBERTY ONSET 1565
by guest on June 22, 2013ajcn.nutrition.orgDownloaded from
