Eur J Pediatr (2013) 172:12071214

DOI 10.1007/s00431-013-2017-4

ORIGINAL ARTICLE

Effects of birth size, post-natal growth and current size

on insulin resistance in 9-year-old children: a prospective
cohort study
Melissa J. Whitrow & Michael J. Davies & Lynne C. Giles &
Bianca L. De Stavola & Julie A. Owens & Oana Maftei &
Vivienne M. Moore

Received: 3 December 2012 / Revised: 18 April 2013 / Accepted: 19 April 2013 / Published online: 8 May 2013
# Springer-Verlag Berlin Heidelberg 2013

Abstract The influence of pre-natal conditions on later type

2 diabetes risk factors such as insulin resistance (IR) may be
mediated by post-natal growth trajectory. We aimed to investigate the association of body size at birth and 9 years
with IR at 9 years. Using data from a prospective Australian
cohort study, we examined the influence of body size from
birth to 9 years [z-score for weight or body mass index
(BMI)] on IR at 9 years (estimated by homeostasis model
assessment). At age 9 years, 151 children provided a fasting
blood sample. z-BMI at age 9 was positively associated with
IR. Birth z-BMI was inversely associated with IR only after
adjustment for z-BMI at age 9 years. This may be interpreted as an effect of accelerated growth between birth and
9 years on IR. There was a statistically significant interaction between birth and 9-year z-BMI. Results from regression models including z-BMI at all available time points
M. J. Whitrow (*) : M. J. Davies : J. A. Owens : O. Maftei
Discipline of Obstetrics & Gynaecology, School of Paediatrics &
Reproductive Health, The University of Adelaide, Mail Drop 327,
Adelaide, SA 5005, Australia
e-mail: melissa.whitrow@adelaide.edu.au
M. J. Whitrow : L. C. Giles : O. Maftei : V. M. Moore
Discipline of Public Health, School of Population Health,
The University of Adelaide, Adelaide, SA, Australia
M. J. Whitrow : M. J. Davies : L. C. Giles : J. A. Owens :
V. M. Moore
Research Centre for the Early Origins of Health and Disease,
Robinson Institute, The University of Adelaide,
Adelaide, SA, Australia
B. L. De Stavola
Department of Medical Statistics, Faculty of Epidemiology and
Population Health, London School of Hygiene and Tropical
Medicine, London, UK

(birth, 6 and 12 months, and 2, 3.5 and 9 years) indicate a

possible inverse association between body size at 3.5 years
and HOMA-IR at 9 years. Results were similar when the
analyses were repeated with z-weight substituted for z-BMI.
These results add to the body of evidence concerning the
importance of growth in early life for later IR, and highlight
a possible interaction between pre- and post-natal growth.
The potential influence of growth at around 3.5 years for
HOMA-IR at 9 years warrants further investigation.
Keywords Birth weight . Body mass index . Body weight .
Childhood obesity . Growth . Insulin resistance
Abbreviations
BMI
Body mass index
HOMA-IR Homeostasis model assessment of insulin
resistance
IR
Insulin resistance
z-BMI
Standard deviation score for body mass index
z-Weight
Standard deviation score for weight

Introduction
Perturbations in development in early life may contribute to
disorders of insulin action and glucose homeostasis, such as
insulin resistance (IR) and type 2 diabetes [31]. The association of birth size, as an indicator of pre-natal conditions,
and these disorders has been examined in many studies.
Both low and high birth weight have been implicated, but
often in separate studies of heterogeneous populations [14,
25, 31].

1208

There is some evidence that the influence of pre-natal conditions on later glucoseinsulin homeostasis is mediated by
subsequent growth trajectory [19]. However, fewer studies have
simultaneously characterized pre-natal and post-natal growth
and type 2 diabetes risk factors such as IR. In studies with data
at two ages (birth weight and current, pre-adolescence, weight),
current weight is the strongest predictor of IR, with birth weight
tending to be statistically significant only after adjustment for
current size in some studies [1, 21, 30], and remaining nonsignificant in others [33]. The interpretation of this is contentious [21, 22, 29], but it has been suggested it points to a role of
the post-natal environment affecting growth and associated
factors between the two stages [5, 22].
Longitudinal studies with measures of body size at multiple time points have identified birth to 3 years [26], birth to
7 years [7] and 18 months to late childhood [11] as potential
critical periods of growth for insulin indicators of metabolic
health (insulin sensitivity, IR and fasting insulin, respectively). A chronic high or rising (regardless of birth size)
growth trajectory between birth and 14 years has also been
associated with higher IR and fasting insulin [15]. In contrast, Gardner and colleagues [12] found current weight
(9 years) rather than weight velocity most influential, and
Larnkjaer and colleagues [18] found no effect of growth in
infancy on IR at 17 years. The heterogeneity in the sample
used (pre- or term birth), timing and type of measurement
(including weight, weight for height and body composition),
insulin measure and statistical methods all contribute to
these inconsistencies.
Despite calls for studies to use the life course plot methods
developed by Cole [5] to jointly investigate the influence of
pre- and post-natal growth on IR [21, 25], to date no studies
have used this method in a sample of children born at term.
These methods use a set of regression equations that enables
reporting of (1) the effect of accelerated growth (a change in a
childs relative position in the distribution of size) between
two time points and (2) the specific effect of body size at
multiple points in childhood on IR. Thus, the life course plots
generated with such methods help to elucidate the magnitude
and direction of any observed associations.
The aim of this study was to investigate the associations,
and any interactions, between size at birth and growth across
childhood with a surrogate estimate of IR at 9 years. We
applied the life course plot methods proposed by Cole [5] to
data from a contemporary cohort of children born at term
who have been prospectively followed from before birth.

Methods
The childs primary care giver provided written informed
consent to participate in the Generation 1 study using the
consent form of the universitys Human Research Ethics

Eur J Pediatr (2013) 172:12071214

Committee. This consent was renewed before commencing

each wave of data collection. In addition, children provided
verbal informed consent for study participation in the most
recent wave of data collection (at 9 years); prior to this, the
children were deemed too young to provide informed consent.
Ethics approval from The University of Adelaide Human
Research Ethics Committee was obtained for the study and
consent procedure.
The Generation 1 cohort is a prospective longitudinal
study of women and their children. Women were recruited
in the first 16 weeks of pregnancy in 1998 to 2000 from four
antenatal clinics in Adelaide, South Australia. Of the 605
women recruited, 557 (92 %) completed the pregnancy
phase of the study and had a live singleton baby.
Participating women were similar to all women having
children in South Australia in 19982000 for a range of
social and economic indicators [23].
Thus far, data have been collected at 16 and 32 weeks
gestation, when the children were born, at ages 3, 6, 9 and
12 months, and 2, 3.5, 5 and 9 years. Data pertaining to the
children and their wider family circumstances have been
collected at each wave via personal interviews in all but
the 3-month and 5-year waves, when data were collected by
phone interview and postal collection, respectively.
Special arrangements were made for midwives to weigh
and measure the babies at birth, for study purposes. These
measurements were compared with those recorded in routine records. At 6 through 12 months, and 2 and 3.5 years
weight (Heine S-831,61 scales at <3 years, Tanita TBF 531
scales at 3.5 years) and height/length (using a board and tape
measure) were measured. At 9 years, weight (Tanita TBF
521 body fat scales) and height (using portable stadiometers) was measured. All measures were taken by trained
research nurses using protocols based on World Health
Organisation guidelines [32]. BMI at 9 years was derived
as weight (in kilogram)/height (in square meter). Standard
deviation scores for weight and BMI at birth (z-weight0 and
z-BMI0) and 9 years (z-weight9y and z-BMI9y) were derived
based on the 1990 standards for British children [6].
Gestational age was incorporated in the calculation of the
z-weight0 and z-BMI0.
At the follow-up appointment at 9 years, 443 children (and
mothers) completed questionnaires and had anthropometric
measurements taken. Following the main study appointment,
164 children consented to a fasting (12 h) blood sample.
Plasma glucose was measured by a Hitachi 912 automated
sample system using the Glucose HK assay kit and quality
controls: Precinorm U and Precipath U (Roche Diagnostics,
NSW, Australia), and insulin by a commercial radioimmunoassay kit specific for human insulin (Linco, Abacus ALS).
One child was excluded from further analyses due to results
indicative of a non-fasting blood sample. IR was estimated
using the homeostasis model assessment (HOMA-IR=fasting

1208

There is some evidence that the influence of pre-natal conditions on later glucoseinsulin homeostasis is mediated by
subsequent growth trajectory [19]. However, fewer studies have
simultaneously characterized pre-natal and post-natal growth
and type 2 diabetes risk factors such as IR. In studies with data
at two ages (birth weight and current, pre-adolescence, weight),
current weight is the strongest predictor of IR, with birth weight
tending to be statistically significant only after adjustment for
current size in some studies [1, 21, 30], and remaining nonsignificant in others [33]. The interpretation of this is contentious [21, 22, 29], but it has been suggested it points to a role of
the post-natal environment affecting growth and associated
factors between the two stages [5, 22].
Longitudinal studies with measures of body size at multiple time points have identified birth to 3 years [26], birth to
7 years [7] and 18 months to late childhood [11] as potential
critical periods of growth for insulin indicators of metabolic
health (insulin sensitivity, IR and fasting insulin, respectively). A chronic high or rising (regardless of birth size)
growth trajectory between birth and 14 years has also been
associated with higher IR and fasting insulin [15]. In contrast, Gardner and colleagues [12] found current weight
(9 years) rather than weight velocity most influential, and
Larnkjaer and colleagues [18] found no effect of growth in
infancy on IR at 17 years. The heterogeneity in the sample
used (pre- or term birth), timing and type of measurement
(including weight, weight for height and body composition),
insulin measure and statistical methods all contribute to
these inconsistencies.
Despite calls for studies to use the life course plot methods
developed by Cole [5] to jointly investigate the influence of
pre- and post-natal growth on IR [21, 25], to date no studies
have used this method in a sample of children born at term.
These methods use a set of regression equations that enables
reporting of (1) the effect of accelerated growth (a change in a
childs relative position in the distribution of size) between
two time points and (2) the specific effect of body size at
multiple points in childhood on IR. Thus, the life course plots
generated with such methods help to elucidate the magnitude
and direction of any observed associations.
The aim of this study was to investigate the associations,
and any interactions, between size at birth and growth across
childhood with a surrogate estimate of IR at 9 years. We
applied the life course plot methods proposed by Cole [5] to
data from a contemporary cohort of children born at term
who have been prospectively followed from before birth.

Methods
The childs primary care giver provided written informed
consent to participate in the Generation 1 study using the
consent form of the universitys Human Research Ethics

Eur J Pediatr (2013) 172:12071214

Committee. This consent was renewed before commencing

each wave of data collection. In addition, children provided
verbal informed consent for study participation in the most
recent wave of data collection (at 9 years); prior to this, the
children were deemed too young to provide informed consent.
Ethics approval from The University of Adelaide Human
Research Ethics Committee was obtained for the study and
consent procedure.
The Generation 1 cohort is a prospective longitudinal
study of women and their children. Women were recruited
in the first 16 weeks of pregnancy in 1998 to 2000 from four
antenatal clinics in Adelaide, South Australia. Of the 605
women recruited, 557 (92 %) completed the pregnancy
phase of the study and had a live singleton baby.
Participating women were similar to all women having
children in South Australia in 19982000 for a range of
social and economic indicators [23].
Thus far, data have been collected at 16 and 32 weeks
gestation, when the children were born, at ages 3, 6, 9 and
12 months, and 2, 3.5, 5 and 9 years. Data pertaining to the
children and their wider family circumstances have been
collected at each wave via personal interviews in all but
the 3-month and 5-year waves, when data were collected by
phone interview and postal collection, respectively.
Special arrangements were made for midwives to weigh
and measure the babies at birth, for study purposes. These
measurements were compared with those recorded in routine records. At 6 through 12 months, and 2 and 3.5 years
weight (Heine S-831,61 scales at <3 years, Tanita TBF 531
scales at 3.5 years) and height/length (using a board and tape
measure) were measured. At 9 years, weight (Tanita TBF
521 body fat scales) and height (using portable stadiometers) was measured. All measures were taken by trained
research nurses using protocols based on World Health
Organisation guidelines [32]. BMI at 9 years was derived
as weight (in kilogram)/height (in square meter). Standard
deviation scores for weight and BMI at birth (z-weight0 and
z-BMI0) and 9 years (z-weight9y and z-BMI9y) were derived
based on the 1990 standards for British children [6].
Gestational age was incorporated in the calculation of the
z-weight0 and z-BMI0.
At the follow-up appointment at 9 years, 443 children (and
mothers) completed questionnaires and had anthropometric
measurements taken. Following the main study appointment,
164 children consented to a fasting (12 h) blood sample.
Plasma glucose was measured by a Hitachi 912 automated
sample system using the Glucose HK assay kit and quality
controls: Precinorm U and Precipath U (Roche Diagnostics,
NSW, Australia), and insulin by a commercial radioimmunoassay kit specific for human insulin (Linco, Abacus ALS).
One child was excluded from further analyses due to results
indicative of a non-fasting blood sample. IR was estimated
using the homeostasis model assessment (HOMA-IR=fasting

Eur J Pediatr (2013) 172:12071214

insulin (U/mL)fasting glucose (mmol/L)/22.5), as recommended for epidemiological studies of children [17].
Variables were considered potential confounders if there
was reasonable evidence in the literature for an independent
association between them and the exposure (birth weight) and
outcome (child HOMA-IR). There are a range of pregnancy
factors that could contribute to birth weight, but are not
independent contributors to IR. In the absence of any established confounders, maternal education at the time of pregnancy (as a proxy for socioeconomic status) was the only
candidate [16]. Pubertal stage is on the pathway between
growth and IR, and hence was not included as a confounder.
The analysis set for the present study comprised the
151 children born after 37 weeks of gestation with valid
measures of height, weight and insulin resistance at
9 years. The distribution of HOMA-IR was positively
skewed, which was addressed in the analyses. The age
in days at which the 9-year blood sample was taken
was included in all models. All models that included
birth size were adjusted for gestational age in days,
even though all births were term.
To overcome previously identified problems with interpreting back transformation of log-transformed data [24], generalized linear models (GLMs) with log-link function and a
Gaussian distribution were used to estimate the effects of standardized birth weight and birth BMI on HOMA-IR at 9 years
[4]. In the GLMs that were initially fit, HOMA-IR at age 9 years
was regressed on standardized birth size (i.e. z-weight0 or zBMI0). Maternal education was also included in these initial
models as a potential confounder. We then added a quadratic
term for standardized birth size to allow for the possibility of a
non-linear association between HOMA-IR and birth size.
A second series of GLMs that regressed HOMA-IR at
age 9 years on body size at birth and at age 9 years were
then fit so as to derive regression coefficients for subsequent
display in life course plots [5]. Such plots can be used to
demonstrate the effect of growththat is, change, over time
in a childs relative position in the distribution of body
sizeon HOMA-IR. We considered four distinct parameterizations of the regression model, with the four models
distinguished by the interaction terms included.
Specifically, the regression models included interactions
between (1) birth size and size at age 9 years, (2) birth size
and change in size, (3) change in size and size at age 9 years
and (4) average of birth size and size at age 9 years, and
change in size, respectively.
In a final set of fitted models, HOMA-IR at age 9 years was
regressed on all available measurements of body size (as well as
maternal education). Separate models were fit for z-weight and
z-BMI. The number of observations available for these analyses
was slightly smaller (n=143) due to missing interim measurements of body size between the ages of 6 months and 3.5 years.
The coefficients from these models were also displayed in life

1209

course plots. This set of analyses enables the identification of

any periods of growth that are more influential than others on
HOMA-IR.
Results are presented as predicted percent change in HOMAIR and either 95 % confidence intervals (95 % CI) or transformed standard errors (SE) derived using the delta method.

Results
At 9 years, 151 children provided a fasting blood sample, had
gestational age 37 weeks and valid measures of height,
weight, glucose and insulin. At 9 years, the median fasting
glucose was 4.9 mmol/L [interquartile range (IQR), 4.75.2],
the median insulin was 11.3 U/mL (IQR, 9.013.0) and the
median HOMA-IR score was 2.3 (IQR, 1.93.0). A total of
6.6 % of children had mild fasting hyperglycaemia
(>5.6 mmol/L) and normal fasting plasma insulin levels.
The mean maternal age at the birth of these study children was 30.45.3 years, and for 44 mothers (29.1 %) this
was their first child. Just over half (n=78, 51.7 %) of the
women had completed high school (but not further education) prior to the birth of the study child. Mothers of children
not providing a fasting blood sample at 9 years were no
younger or less educated than those who took part (p>0.5),
and the sample included the full spectrum of maternal education levels. Table 1 shows the body size characteristics of
children included in the analyses.
The initially fitted regression models showed that the two
measures of body size at age 9 years were positively associated with HOMA-IR (adjusted for current age) (Table 2).
Neither z-weight 0 nor z-BMI 0 was associated with
HOMA-IR (adjusted for gestational age). Including quadratic
terms for birth size in the statistical models provided no evidence
for a non-linear relationship between birth size and HOMA-IR.
Figure 1 presents a life course plot of the percent change
in HOMA-IR associated with z-BMI0 and z-BMI9y when
both variables were included in the same model (solid line).
After controlling for z-BMI9y, z-BMI0 was inversely associated (p = 0.008) with HOMA-IR at 9 years. z-BMI 9y
remained positively associated with HOMA-IR (p<0.001)
and the magnitude of this association was slightly larger
after adjustment for z-BMI0.
Results followed a similar pattern when HOMA-IR was
instead regressed on z-weight0 and z-weight9y (p=0.036 for
z-weight0, p<0.001 for z-weight9y). The addition of body
size at age 9 years (z-weight9y or z-BMI9y) significantly
improved the fit of each model compared to models with
birth size alone. Coefficients for z-BMI and z-weight were
unchanged after adjustment for maternal education.
These models suggest an effect of accelerated growth
between birth and 9 years, which was supported by modeling the change in z-score (9 years minus birth). A 1 standard

1210
Table 1 Characteristics of children included and excluded from
the analyses

Age- and sex-specific standard

deviation scores based on 1990
UK reference population

Eur J Pediatr (2013) 172:12071214

Characteristic

Excluded from analyses

Mean SD

Included in analyses
Mean SD

Birth characteristics

(n=368)

(n=151)

Gestational age (weeks)

Weight (g)
Length (cm)
BMI (kg/m2)
z-weighta
z-BMIa
Characteristics at 9 years
Age (years)
Weight (kg)
Height (cm)
BMI (kg/m2)
z-weighta
z-BMIa

39.71.2
3,478.0487.0
50.42.1
13.71.3
0.131.0
0.471.0b
(n=260)
9.70.4
35.18.8
138.46.6
18.23.4
0.481.1
0.491.2

39.71.2
3,490.8465.4
50.52.2
13.71.3
0.181.0
0.481.0
(n=151)
9.60.2
33.57.2
137.66.3
17.62.6
0.331.1
0.331.1

deviation increase in z-BMI between birth and 9 years was

associated with an increase in HOMA-IR of 11.04 % (95 %
CI 7.30, 14.92). Similarly, a 1 standard deviation
change in z-weight between birth and 9 years was
associated with an increase in HOMA-IR of 11.92 %
(95 % CI 7.67, 16.35).

There was a significant interaction between z-BMI0 and

z-BMI9y (p=0.043). To display this result, analyses were
repeated dichotomizing z-BMI0 (< or median birth BMI,
displayed in Fig. 1). Both the inverse association of z-BMI0
and the positive association of z-BMI9y with HOMA-IR
were greater in children in the low birth BMI group. The

Table 2 Association of birth

size or body size at 9 years with
insulin resistance at 9 years
Measure of birth sizea,b
z-weight0
z-weight0 squared
z-BMI0
z-BMI0 squared
Measure of current (9 years) body sizeb,c
z-weight9y

All birth size models adjusted

for gestational age (days) and
age at time of blood sample
(days)

Age- and sex-specific z-scores

based on 1990 UK reference
population

All size at 9-year models adjusted for age at time of blood sample
(days)

% change in HOMA-IR associated with a 1 unit

change in the predictor variable (95% CI)

p value

1.19 (6.66, 4.59)

2.41 (1.45, 6.42)
3.89 (9.00, 1.50)
3.37 (0.42, 7.30)

0.679
0.224
0.154
0.082

15.59 (5.80, 20.83)

<0.001

z-BMI9y
13.81 (8.91, 18.92)
Measure of birth size and body size at age 9 years
z-weight0
5.28 (9.97, 0.36)
z-weight9y
16.73 (11.61, 22.08)
z-BMI0
6.16 (10.46, 1.66)
z-BMI9y
14.70 (9.90, 19.70)
Interactions between birth size and current body size
1. z-weight0 z-weight9y
1.86 (6.90, 3.46)
2. z-weight0 (z-weight9y z-weight0)
2.99 (5.10, 11.77)
3. z-weight9y (z-weight9y z-weight0)
3.68 (11.61, 4.95)
4. (z-weight0 z-weight9y)/2(z-weight9y
0.40 (5.21, 4.65)
z-weight0)
4.46 (8.59, 0.15)
1. z-BMI0 z-BMI9y
2. z-BMI0 (z-BMI9y z-BMI0)
1.95 (9.52, 6.25)
3. z-BMI9y (z-BMI9y z-BMI0)
2.09 (10.12, 6.66)
4. (z-BMI0 +z-BMI9y)/2(z-BMI9y z-BMI0) 1.90 (6.62, 3.06)

<0.001
0.036
<0.001
0.008
<0.001
0.486
0.480
0.392
0.873
0.043
0.631
0.628
0.446

Eur J Pediatr (2013) 172:12071214

1211

Fig. 1 Life course

plotpercent change in
HOMA-IR (95 % CI) at
9 years associated with a 1 SD
change in BMI at birth or a 1
SD change in BMI at 9 years
(age- and sex-specific standard
deviation scores based on 1990
UK reference population)
plotted against the age at
measurement (n = 151). (All
models included the relevant
measure of birth size and are
adjusted for current size,
gestational age (days) and age
at time of blood sample (days)

steeper slope for children with below median birth BMI

indicates that the effect of accelerated growth across 9 years
was greater for these children, relative to those above median birth BMI.
When body sizes at all available time points (birth, 6 and
12 months, and 2, 3.5 and 9 years) were included simultaneously in the model, only birth and 9 years were significant
(for both z-weight and z-BMI; life course plot for z-BMI
shown in Fig. 2). The coefficients for body size at age
9 years were attenuated, and the coefficients for birth size
were slightly larger, with the addition of the interim body
sizes to the model. Whilst the interim measures of body size
were not significant, the dip in the plot suggests a possible
inverse association between body size at 3.5 years and
HOMA-IR at 9 years. After adjustment for maternal education, the coefficient for z-weight0 was slightly attenuated (from
6.01, 95 % CI 11.59, 0.09 to 5.66, 95 % CI 11.14, 0.17;
p=0.057). The other coefficients for z-weight and z-BMI were
unchanged.

Discussion
This is the first study to use the methods proposed by Cole
[5] to investigate the influence of size at birth and growth
across childhood on a surrogate measure of IR in late
childhood using a representative sample of children born
at term. These life course plot methods allow the simultaneous assessment of the effects of size at different time
points, and the effects of growth between time points, on
IR. We found an inverse association between birth size and
HOMA-IR at 9 years only after adjustment for body size at
9 years. This observed effect persisted after adjusting for

maternal education and was similar when using z-weight or

z-BMI as the measure of body size. Children with lower
birth BMI and high BMI at age 9 had higher HOMA-IR, as
indicated by the significant interaction, and the accompanying life course plot displays this. Our results may also
suggest growth around 3.5 years could be important for
HOMA-IR at 9 years. Taken together, our results suggest
an influential role for growth (and the environment and
factors affecting this) between birth and late childhood
on the development of IR [5, 29]. For practitioners, the
implication is that accelerated growth, especially in
children who were small at term birth, appears to have
adverse consequences for metabolic health. Weight gain
that outstrips height gain in these children should not be
encouraged.
The interpretation and underlying mechanisms of an association between birth size and health outcomes after adjustment for current size is contentious and has particularly
been debated with regard to adult blood pressure [20, 29].
Our series of models suggest post-natal growth between
birth and 9 years is important for IR development, and that
this is magnified in children with lower birth weight. Whilst
we have not described body size beyond weight and BMI,
evidence elsewhere suggests that periods of rapid weight
gain in any stage of life are characterized by a disproportionate gain in fat compared to lean body tissue [9]. The
magnified effect in children of lower birth weight may arise
from altered functional capacity in key glucoregulatory tissues and in other body systems that can affect these, by
exposure to pre-natal substrate deprivation at critical periods
of development. These outcomes then interact with adequate
or excess nutrition availability post-natally, to drive growth
and obesity [3, 27].

1212

Eur J Pediatr (2013) 172:12071214

Fig. 2 Life course

plotpercent change in
HOMA-IR (95 % CI) at
9 years associated with a 1 SD
change in BMI at various time
points between birth and
9 years (age- and sex-specific
standard deviation scores based
on 1990 UK reference
population) plotted against the
age at measurement (n=143).
All models adjusted for
gestational age (days) and age
at time of blood sample (days),
measure of body size at all other
ages and maternal education

In non-human species such as sheep and rodents, restriction of placental growth, a common cause of intra-uterine
growth restriction, reduces weight and BMI at birth [8, 13].
In sheep, this is followed by early onset catch-up growth and
central obesity and insulin resistance, with the latter likely to
reflect both obesity-mediated impairment as well as intrinsic
abnormalities in skeletal muscle or liver [8, 13]. This pattern
of growth and fat accumulation appears, in part, attributable
to impaired appetite control and possibly leptin resistance
[8], but may involve additional mechanisms [2, 9]. There are
similar patterns in rodents where exposure to excess nutrition post-natally interacts with pre-natal nutritional restriction to induce later obesity and sometimes insulin resistance
[3, 27].
The only previous study of pre- and post-natal growth
and IR in children employing the same analytical methods
as this study was in pre-term infants [5, 11]. That study also
found an inverse association between birth weight and IR
only after adjustment for current size. In contrast to our
results, weight gain between 18 months and 912 years
was identified as being of greater importance than weight
gain from birth to 18 months. The likely explanation for the
contrasting findings is the different pattern of post-natal
growth experienced by pre-term compared to term infants.
When body size at each time point was included in the
model, there was a suggestion that growth around 3.5 years
may be particularly relevant for IR. Outcomes from other
studies with similar measures of body size at multiple time
points, but using different analytical methods, have been
conflicting [7, 18, 26], ranging from no effect of post-natal
growth [18], to birth to 3 years [26], birth to 7 years [7] or only
current weight [12] being significant. All of these studies,
including ours, are restricted by the number and timing of

the available measures to conclusively detect critical periods

of growth. The one study investigating growth trajectories and
IR [15] found both chronic high and rising trajectories
were important (the latter in agreement with the significant
interaction between birth size and current size reported here),
but did not identify potential critical periods. BMI increases
steeply during infancy (peaking between 6 and 12 months)
then falls during toddlerhood and the preschool years, reaching its lowest point at around 56 years (the adiposity rebound) [28]. Our interpretation of the potential role of weight
at around 3.5 years for HOMA-IR is that early adiposity
rebound has been consistently associated with increased risk
of diabetes and obesity in adulthood [10, 34], and hence,
lower weight gain at this age relative to the UK reference
group is associated with lower IR. Whilst our study was
strengthened by the repeated, rigorous measurements in infancy and toddlerhood, we were unable to pinpoint the nadir
in BMI and hence timing of the adiposity rebound. This
requires a series of measurements at very short intervals
between 4 and 6 years, which is often not feasible. Hence,
we cannot draw conclusions about the precise timing of critical periods between or beyond the measurement time points
used in our and other studies.
Our study is also limited by the number of children consenting to a blood sample at 9 years, potentially introducing
bias and reducing statistical power. However, the sample of
children consenting for a blood sample were no different to
children not consenting in terms of maternal age and education, and the sample included representation of the entire
spectrum of these variables. Our significant results for current
and birth size are in line with other studies [35]; however, we
are unable to rule out insufficient power to detect interactions.
HOMA-IR is one method of measuring IR in children and,

Eur J Pediatr (2013) 172:12071214

whilst not as informative as a glucose tolerance test, it is an

acceptable alternative in a large cohort childhood study [17].
Identifying the determinants of the emergence of IR in
childhood is a significant public health issue: children exhibiting signs of poor glucoseinsulin homeostasis are more
likely to proceed to type 2 diabetes and cardiovascular disease
in their adult life. These results suggest research and public
health initiatives focus on the underlying drivers of the pace of
growth across childhood, particularly in children born at term
with relatively low weight-for-length. Further research would
benefit from measures of height and weight encompassing the
time of the adiposity rebound to re-examine the possible effect
we found of growth at around 3.5 years.
Acknowledgments We are grateful to the women and children who
have participated in the Generation 1 study and also to other family
members who have supported the study. We thank Prof. Jeffrey Robinson (of the Discipline of Obstetrics & Gynaecology, The University
of Adelaide) for his role in cohort establishment. We also thank
Kendall Smith for study co-ordination, Suzanne Edwards for statistical
analysis (both of the Discipline of Public Health, The University of
Adelaide) and Dr Miles DeBlasio (of the Discipline of Obstetrics &
Gynaecology, The University of Adelaide) for analyses of blood samples and the staff members involved in interviews, data base construction and data entry.
Funding This study was supported by grants from the Faculty of Health
Sciences at the University of Adelaide, the Dairy Research and
Development Corporation, the Channel 7 Childrens Research
Foundation, the Australian Research Council (Future Fellowship
FT100101018 to MD) and the Australian National Health and Medical
Research Council (Grants 465455, 465437, Career Development Award in
Population Health 349548 to MD and Australian Based Public Health
Training Fellowship 627033 to LG). The funders had no role in study
design, data collection and analysis, decision to publish, or preparation of
the manuscript.

Conflict of interest The authors declare no conflicts of interest and

no financial relationships with the grant funding agencies that sponsored the research.

References
1. Bavdekar A, Yajnik CS, Fall CH, Bapat S, Pandit AN, Deshpande
V, Bhave S, Kellingray SD, Joglekar C (1999) Insulin resistance
syndrome in 8-year-old Indian children: small at birth, big at
8 years, or both? Diabetes 48:24222429
2. Beltrand J, Levy-Marchal C (2008) Pathophysiology of insulin
resistance in subjects born small for gestational age. Best Pract
Res Clin Endocrinol Metab 22:503515
3. Bol VV, Delattre AI, Reusens B, Raes M, Remacle C (2009)
Forced catch-up growth after fetal protein restriction alters the
adipose tissue gene expression program leading to obesity in
adult mice. Am J Physiol Regul Integr Comp Physiol 297:
R291R299
4. Breslow NE (1996) Generalized linear models: checking
assumptions and strengthening conclusions. Statistica Applicata
8:2341

1213
5. Cole TJ (2007) Ch 6: the life course plot in life course
analysis. In: Pickles A, Maughan B, Wadsworth M (eds)
Epidemiological methods in life course research. Oxford University
Press, Oxford
6. Cole TJ, Freeman JV, Preece MA (1995) Body mass reference
curves for the UK, 1990. Arch Dis Child 73:2529
7. Crowther NJ, Cameron N, Trusler J, Gray IP (1998) Association
between poor glucose tolerance and rapid post natal weight gain in
seven-year-old children. Diabetologia 41:11631167
8. de Blasio MJ, Blache D, Gatford KL, Robinson JS, Owens JA
(2010) Placental restriction increases adipose leptin gene expression and plasma leptin and alters their relationship to feeding
activity in the young lamb. Pediatr Res 67:603608
9. Dulloo AG, Jacquet J, Seydoux J, Montani JP (2006) The thrifty
'catch-up fat' phenotype: its impact on insulin sensitivity during
growth trajectories to obesity and metabolic syndrome. Int J Obes
30:S23S35
10. Eriksson JG, Forsn T, Tuomilehto J, Osmond C, Barker DJP
(2003) Early adiposity rebound in childhood and risk of type 2
diabetes in adult life. Diabetologia 46:190194
11. Fewtrell MS, Doherty C, Cole TJ, Stafford M, Hales CN, Lucas A
(2000) Effects of size at birth, gestational age and early growth in
preterm infants on glucose and insulin concentrations at 912
years. Diabetologia 43:714717
12. Gardner DSL, Hosking J, Metcalf BS, Jeffery AN, Voss LD,
Wilkin TJ (2009) Contribution of early weight gain to childhood
overweight and metabolic health: a longitudinal study (EarlyBird
36). Pediatrics 123:e67e73
13. Gatford KL, Simmons RA, De Blasio MJ, Robinson JS, Owens JA
(2010) Placental programming of postnatal diabetes and impaired
insulin action after IUGR. Placenta 31:S60S65
14. Harder T, Rodekamp E, Schellong K, Dudenhausen JW,
Plagemann A (2007) Birth weight and subsequent risk of type 2
diabetes: a meta-analysis. Am J Epidemiol 165:849857
15. Huang R-C, de Klerk NH, Smith A, Kendall GE, Landau LI, Mori
TA, Newnham JP, Stanley FJ, Oddy WH, Hands B, Beilin LJ
(2011) Lifecourse childhood adiposity trajectories associated with
adolescent insulin resistance. Diabetes Care 34:10191025
16. Joseph KS, Kramer MS (1996) Review of the evidence on fetal
and early childhood antecedents of adult chronic disease.
Epidemiol Rev 18:158174
17. Keskin M, Kurtoglu S, Kendirci M, Atabek ME, Yazici C (2005)
Homeostasis model assessment is more reliable than the fasting
glucose/insulin ratio and quantitative insulin sensitivity check index for assessing insulin resistance among obese children and
adolescents. Pediatrics 115:e500e503
18. Larnkjr A, Schack-Nielsen L, Mlgaard C, Ingstrup HK, Holst
JJ, Michaelsen KF (2010) Effect of growth in infancy on body
composition, insulin resistance, and concentration of appetite hormones in adolescence. Am J Clin Nutr 91:16751683
19. Lawlor DA, Davey Smith G, Ebrahim S (2003) Life course influences on insulin resistance. Diabetes Care 26:97103
20. Lawlor DA, Leon DA, Rasmussen F (2007) Growth trajectory
matters: interpreting the associations among birth weight, concurrent body size, and systolic blood pressure in a cohort study of
378,707 Swedish men. Am J Epidemiol 165:14051412
21. Lawlor DA, Riddoch CJ, Page AS, Anderssen SA, Froberg K,
Harro M, Stansbie D, Davey Smith G (2005) The association of
birthweight and contemporary size with insulin resistance among
children from Estonia and Denmark: findings from the European
Youth Heart Study. Diabet Med 22:921930
22. Lucas A, Fewtrell MS, Cole TJ (1999) Fetal origins of adult
diseasethe hypothesis revisited. BMJ 319:245249
23. Moore VM, Davies MJ, Willson KJ, Worsley A, Robinson JS
(2004) Dietary composition of pregnant women is related to size
of the baby at birth. J Nutr 134:18201826

1214
24. Moran JL, Solomon PJ, Peisach AR, Martin J (2007) New models
for old questions: generalized linear models for cost prediction. J
Eval Clin Pract 13:381389
25. Newsome CA, Shiell AW, Fall CHD, Phillips DIW, Shier R, Law
CM (2003) Is birth weight related to later glucose and insulin
metabolism? A systematic review. Diabet Med 20:339348
26. Ong KK, Petry CJ, Emmett PM, Sandhu MS, Kiess W, Hales
CN, Ness AR, Dunger DB (2004) Insulin sensitivity and secretion in normal children related to size at birth, postnatal growth,
and plasma insulin-like growth factor-I levels. Diabetologia
47:10641070
27. Shahkhalili Y, Moulin J, Zbinden I, Aprikian O, Mace K (2010)
Comparison of two models of intrauterine growth restriction for
early catch-up growth and later development of glucose intolerance and obesity in rats. Am J Physiol Regul Integr Comp Physiol
298:R141R146
28. Silverwood RJ, De Stavola BL, Cole TJ, Leon DA (2009) BMI
peak in infancy as a predictor for later BMI in the Uppsala Family
Study. Int J Obes 33:929937
29. Tu Y-K, West R, Ellison GTH, Gilthorpe MS (2005) Why evidence for the fetal origins of adult disease might be a statistical
artifact: the reversal paradox for the relation between birth
weight and blood pressure in later life. Am J Epidemiol 161:2732
30. Whincup PH, Cook DG, Adshead F, Taylor SJC, Walker M,
Papacosta O, Alberti KGMM (1997) Childhood size is more

Eur J Pediatr (2013) 172:12071214

31.

32.

33.

34.
35.

strongly related than size at birth to glucose and insulin levels in

1011-year-old children. Diabetologia 40:319326
Whincup PH, Kaye SJ, Owen CG, Huxley R, Cook DG, Anazawa
S, Barrett-Connor E, Bhargava SK, Birgisdottir BE, Carlsson S, de
Rooij SR, Dyck RF, Eriksson JG, Falkner B, Fall C, Forsn T, Grill
V, Gudnason V, Hulman S, Hyppnen E, Jeffreys M, Lawlor DA,
Leon DA, Minami J, Mishra G, Osmond C, Power C, RichEdwards JW, Roseboom TJ, Sachdev HS, Syddall H, Thorsdottir
I, Vanhala M, Wadsworth M, Yarbrough DE (2008) Birth weight
and risk of type 2 diabetes: a systematic review. JAMA 300:2886
289732
WHO HWCC (1995) Physical Status: the uses and interpretation
of anthropometry (report of a WHO Expert Committee). Geneva,
Switzerland
Wilkin TJ, Metcalf BS, Murphy MJ, Kirkby J, Jeffery AN, Voss
LD (2002) The relative contributions of birth weight, weight
change, and current weight to insulin resistance in contemporary
5-year-olds. Diabetes 51:34683472
Williams SM (2005) Weight and height growth rate and the timing
of adiposity rebound. Obesity 13:11231130
Yajnik CS, Deshpande SS, Jackson AA, Refsum H, Rao S, Fisher DJ,
Bhat DS, Naik SS, Coyaji KJ, Joglekar CV, Joshi N, Lubree HG,
Deshpande VU, Rege SS, Fall CHD (2008) Vitamin B12 and folate
concentrations during pregnancy and insulin resistance in the offspring: the Pune Maternal Nutrition Study. Diabetologia 51:2938