Polygenic risk, rapid childhood growth, and the development of obesity: evidence from a 4-decade longitudinal study.

Abstract

To test how genomic loci identified in genome-wide association studies influence the development of obesity.
A 38-year prospective longitudinal study of a representative birth cohort.
The Dunedin Multidisciplinary Health and Development Study, Dunedin, New Zealand.
One thousand thirty-seven male and female study members.
We assessed genetic risk with a multilocus genetic risk score. The genetic risk score was composed of single-nucleotide polymorphisms identified in genome-wide association studies of obesity-related phenotypes. We assessed family history from parent body mass index data collected when study members were 11 years of age.
Body mass index growth curves, developmental phenotypes of obesity, and adult obesity outcomes were defined from anthropometric assessments at birth and at 12 subsequent in-person interviews through 38 years of age.
Individuals with higher genetic risk scores were more likely to be chronically obese in adulthood. Genetic risk first manifested as rapid growth during early childhood. Genetic risk was unrelated to birth weight. After birth, children at higher genetic risk gained weight more rapidly and reached adiposity rebound earlier and at a higher body mass index. In turn, these developmental phenotypes predicted adult obesity, mediating about half the genetic effect on adult obesity risk. Genetic associations with growth and obesity risk were independent of family history, indicating that the genetic risk score could provide novel information to clinicians.
Genetic variation linked with obesity risk operates, in part, through accelerating growth in the early childhood years after birth. Etiological research and prevention strategies should target early childhood to address the obesity epidemic.

Full text

Page 1

Confidential. Do not distribute. Pre-embargo material.
ARTICLE
Polygenic Risk, Rapid Childhood Growth,
and the Development of Obesity
Evidence From a 4-Decade Longitudinal Study
Daniel W. Belsky, PhD; Terrie E. Moffitt, PhD; Renate Houts, PhD; Gary G. Bennett, PhD; Andrea K. Biddle, PhD;
James A. Blumenthal, PhD; James P. Evans, MD, PhD; HonaLee Harrington, BA; Karen Sugden, PhD;
Benjamin Williams, BS; Richie Poulton, PhD; Avshalom Caspi, PhD
Objective: To test how genomic loci identified in ge-
nome-wide association studies influence the develop-
ment of obesity.
Design: A 38-year prospective longitudinal study of a
representative birth cohort.
Setting:TheDunedinMultidisciplinaryHealthandDe-
velopment Study, Dunedin, New Zealand.
Participants: One thousand thirty-seven male and fe-
male study members.
Main Exposures: We assessed genetic risk with a mul-
tilocusgeneticriskscore.Thegeneticriskscorewascom-
posed of single-nucleotide polymorphisms identified in
genome-wideassociationstudiesofobesity-relatedphe-
notypes. We assessed family history from parent body
mass index data collected when study members were 11
years of age.
Main Outcome Measures: Body mass index growth
curves, developmental phenotypes of obesity, and adult
obesityoutcomesweredefinedfromanthropometricas-
sessmentsatbirthandat12subsequentin-personinter-
views through 38 years of age.
Results:Individualswithhighergeneticriskscoreswere
more likely to be chronically obese in adulthood. Ge-
netic risk first manifested as rapid growth during early
childhood. Genetic risk was unrelated to birth weight.
Afterbirth,childrenathighergeneticriskgainedweight
more rapidly and reached adiposity rebound earlier and
atahigherbodymassindex.Inturn,thesedevelopmen-
tal phenotypes predicted adult obesity, mediating about
half the genetic effect on adult obesity risk. Genetic as-
sociations with growth and obesity risk were indepen-
dentoffamilyhistory,indicatingthatthegeneticriskscore
could provide novel information to clinicians.
Conclusions:Geneticvariationlinkedwithobesityrisk
operates,inpart,throughacceleratinggrowthintheearly
childhoodyearsafterbirth.Etiologicalresearchandpre-
vention strategies should target early childhood to ad-
dress the obesity epidemic.
Arch Pediatr Adolesc Med. 2012;166(6):515-521
O
tiple single-nucleotide polymorphisms
(SNPs)associatedwithahigheradultbody
mass index (BMI; calculated as weight in
kilograms divided by height in meters
squared).1The next step is to understand
how these SNPs influence the develop-
ment of obesity. Individual differences in
obesity risk emerge during gestation and
arefurtherestablishedduringinfancyand
childhoodthroughacceleratedgrowthtra-
jectories.2,3Therefore,examinationofde-
velopmentalphenotypesinrelationtoge-
neticriskrepresentsapromisingapproach
BESITY IS KNOWN TO BE
heritable, and genome-
wide association studies
(GWASs) have begun to
uncover the molecular
rootsofthisheritabilitybyidentifyingmul-
to understand the pathogenesis of obe-
sity.4-6In this study, we asked how SNPs
with replicated GWAS evidence for asso-
ciation with adult BMI relate to growth
acrossthefirst4decadesoflifeandtoadult
obesityinabirthcohortfolloweduppro-
spectively from birth through 38 years of
age.
The SNPs identified in GWASs con-
tribute small increments to obesity risk.7
AggregatingGWAS-identifiedSNPstopro-
duceagenome-wideindex(ageneticrisk
score [GRS]) yields a quantitative mea-
sure of inherited predisposition toward a
trait, such as BMI.8This approach has
See also pages 522
and 576
Author Affiliations are listed at
the end of this article.
ARCH PEDIATR ADOLESC MED/VOL 166 (NO. 6), JUNE 2012 WWW.ARCHPEDIATRICS.COM
515
©2012 American Medical Association. All rights reserved.

Page 2

Confidential. Do not distribute. Pre-embargo material.
shown promise in the study of complex diseases, such
as diabetes mellitus and heart disease.9,10In this study,
weusedamultilocusgenomeriskscoretotesthowage-
netic predisposition to higher adult BMI might also re-
latetodevelopmentalphenotypesofgrowthduringpro-
posedcriticalperiodsinthedevelopmentofobesity.The
following 3 developmental phenotypes are of interest:
growth during gestation, postnatal growth, and the adi-
posity rebound. All correlate with adult BMI and are
thought to program risk for adult obesity.11-13There-
fore,wetestedthehypothesisthatpolygenicriskforadult
obesity is mediated by these developmental phenotypes
of rapid early growth (Figure 1). Understanding when
indevelopmentgeneticriskforobesitymanifestscanhelp
to refine research and intervention targets.
Ifgeneticriskismediatedthroughearlygrowth,knowl-
edge of how measured genetic risk compares with pa-
rental BMI in predicting children’s growth and obesity
risk is important. We thus tested whether obesity risk
information contained in the GRS was independent of
obesityriskinformationcontainedintheBMIsofthechil-
dren’s parents. That is, does the GRS contain novel in-
formation about children’s risk for obesity beyond their
family history?
METHODS
PARTICIPANTS
ParticipantsaremembersoftheDunedinMultidisciplinaryHealth
and Development Study, a longitudinal investigation of health
and behavior in a complete birth cohort. Study members (1037
members;91%ofeligiblebirths;52%male)wereallindividuals
bornbetweenApril1972andMarch1973inDunedin,NewZea-
land,whowereeligibleforthelongitudinalstudybasedonresi-
dence in the province and who participated in the first fol-
low-up assessment at 3 years of age. The cohort represents the
full range of socioeconomic status in the general population of
New Zealand’s South Island and is primarily white. Assess-
ments were performed at birth and at ages 3, 5, 7, 9, 11, 13, 15,
18, 21, 26, 32, and, most recently, 38 years, with greater than
95%retention.Ateachassessment,studymembersarebrought
totheDunedinresearchunitforafulldayofinterviewsandex-
aminations. The Otago Ethics Committee approved each phase
of the study. The study protocol was approved by the institu-
tionalethicalreviewboardsoftheparticipatinguniversities.In-
formed consent was obtained from all study members.
MAIN EXPOSURES
Obesity GRS
Wederiveda32-SNPGRSfrompublishedGWASsofBMI,obe-
sity, weight, and waist circumference in populations of Euro-
pean descent. The construction of the GRS is described in the
eMethods(http://www.archpediatrics.com).Wevalidatedour
GRS as a measure of obesity risk in data from the Atheroscle-
rosisRiskinCommunities(ARIC)sample.16ARICcohortmem-
bers of European descent who had higher GRSs were larger as
measured by BMI, weight, and waist circumference (r?0.10
[P?1?10−20])andweremorelikelytobeobese(relativerisk,
1.73 [95% CI, 1.51-1.97] for individuals in the highest vs the
lowest quintile of the GRS distribution).
Wegenotypedthe32GRSSNPsintheDunedinStudycohort
with a commercially available array (BeadPlex Array; Illumina,
Inc)usingDNAextractedfromwholeblood(93%ofthesample)
orbuccalswabs(7%ofthesample).Ofthe32GRSSNPs,29were
calledsuccessfullyinmorethan95%ofthecohort,andwecon-
structed the final score from these SNPs (eTable 1). Compari-
sonofthe29-SNPGRSwiththeoriginal32-SNPGRSintheARIC
samplerevealednodifferencesinscoredistributionoreffectsizes.
DunedinStudymemberscarried15to36riskalleles(mean[SD],
26.04 [3.32]). After weighting, GRS values ranged from 13.71
to 35.04 (mean [SD], 24.71 [3.59]) (eFigure 1). The GRS was
standardized to have a mean of 0 and an SD of 1 for analyses.
Family History of Obesity
Parental BMI was available for 97.82% of the cohort. Parental
BMIs were computed from self-reports of height and weight
when children were 11 years of age. To measure familial pre-
dispositiontoobesity,parentalBMIswerestandardizedwithin
sex,andthestandardizedscoreswereaveragedtocreateasingle
family history score.
OUTCOME MEASURES
Body Mass Index
Individuals’ height and weight were measured at ages 3, 5, 7,
9, 11, 13, 15, 18, 21, 26, 32, and 38 years. Height was mea-
sured to the nearest millimeter using a portable stadiometer
(Harpenden; Holtain, Ltd). Weight was measured to the near-
est 0.1 kg at ages 3, 5, 7, 9, 11, 13, 15, 18, 21, 26, 32, and 36
yearsusingcalibratedscales.Individualswereweighedinlight
clothing. Obesity was defined at 15 years of age using US Cen-
ters for Disease Control and Prevention cutoff points
(BMI?24.64forboys;BMI?25.46forgirls),whichshowpre-
dictivevalidityforobesityandcoronaryheartdiseaseinyoung
adulthoodthatissimilartotheInternationalObesityTaskForce
cutoff points.17Obesity was defined at ages 18 to 38 years as a
BMI of 30 or greater. Individuals who met obesity criteria for
at least 50% of 6 measurements from ages 15 to 38 years were
classified as chronically obese.18
Growth during
gestation
(birth weight)
Postnatal growth
(weight gain,
birth through 3 y of age)
Adiposity rebound
(age and BMI at the nadir
of the childhood BMI
growth curve)
Developmental Phenotypes of Rapid Early Growth
Polygenic
risk
Adult
obesity
Time
Figure 1. Developmental phenotypes of rapid early growth hypothesized to
mediate polygenic risk for obesity. The genetic epidemiology of obesity
indicates that a large number of common polymorphisms each contribute
small, additive increments to risk for obesity.14,15The combined influence of
these polymorphisms can be summarized in a polygenic risk profile.8The
developmental epidemiology of obesity highlights the following 3
developmental phenotypes of rapid early growth that predispose children to
become obese in later life: (1) growth during gestation, (2) postnatal growth,
and (3) adiposity rebound.11,12We tested the hypothesis that these
developmental phenotypes would mediate polygenic risk for adult obesity.
BMI indicates body mass index.
ARCH PEDIATR ADOLESC MED/VOL 166 (NO. 6), JUNE 2012 WWW.ARCHPEDIATRICS.COM
516
©2012 American Medical Association. All rights reserved.

Page 3

Confidential. Do not distribute. Pre-embargo material.
Additional Measures of Adiposity
At ages 7 and 9 years, tricep and subscapular skinfold thick-
nessesweremeasuredbytrainedanthropometrists.Atages26,
32,and38years,waistgirthwasmeasuredbyaveraging2mea-
surements of the perimeter at the level of the noticeable waist
narrowing.Atages32and38years,fatmasswasmeasuredusing
a body composition analyzer (BC 418; Tanita) to assess bio-
electrical impedence.19
Developmental Phenotypes
of Early Growth
Rate of early-childhood weight gain was assessed as the differ-
encebetweenweightatbirth(fromhospitalrecords)andweight
at 3 years of age. Adiposity rebound was calculated as the na-
dir of each individual’s childhood BMI curve fitted across ages
3through13years.Weusedmultilevellongitudinalmodeling
to fit individual growth curves.20Models included linear and
quadratic slope terms and were adjusted for sex. Children in
our sample experienced adiposity rebound at about 6 years of
age (mean [SD] age, 6.11 [1.10] years) at a BMI of approxi-
mately 16 (mean [SD], 15.57 [1.00]).
STATISTICAL ANALYSIS
We analyzed life-course growth using a multilevel longitudinal
growthmodel20fittedtoBMImeasurementsatages3,5,7,9,11,
13, 15, 18, 21, 26, 32, and 38 years. We set the intercept at 13
yearsofage.Wemodeledseparatelinearandquadraticslopesfor
growthduringchildhood(ages3-13years)andadulthood(ages
13-38years).TheinterceptcapturedthesamplemeanBMIat13
yearsofage(?=19.97).Slopecoefficientscapturedannualchange/
acceleration in BMI. Linear slope terms captured change in BMI
across childhood (?=1.19) and adulthood (?=0.51). Quadratic
slope terms captured acceleration of change, that is, the concav-
ity of the trajectory in childhood (?=0.08) and the convexity of
thetrajectoryinadulthood(?=−0.01).Allmodeltermsweresta-
tistically significant (P?.001).
We tested genetic influence on growth by modeling the in-
tercept and linear slope terms of the life-course growth curve
as functions of the GRS and covariates. The GRS coefficients
measured the effect of a 1-SD increase in genetic risk on BMI
at 13 years of age (intercept) and on the linear change per year
inBMIfromages3through13years(childhoodslope)andages
13 through 38 years (adulthood slope).
Wetestedgeneticassociationswithcross-sectionalmeasure-
ments of BMI and with other quantitative traits using linear re-
gressionmodels.TheGRScoefficientswerestandardizedtoeffect-
sizecorrelations(Pearsonr)foreaseofinterpretation.Wetested
genetic associations with obesity risk using Poisson regression
models. The GRS coefficients were exponentiated to compute
relative risks. We tested mediation of genetic risk for obesity
through developmental phenotypes of early growth using the
structural equation described by MacKinnon and Dwyer.21Me-
diation analyses decomposed GRS-obesity associations into di-
rect (unmediated) and indirect (mediated through a develop-
mentalphenotype)components.Statisticaltestsofmediationwere
conducted using methods described by Preacher and col-
leagues.22-24
All models were adjusted for sex and constituted the 98%
(n=856) of Dunedin Study members of European descent
with available BMI, family history, and genotype data. We
used SAS statistical software (version 9.2)25for growth mod-
eling and mediation analyses and STATA (version 11.0)26for
other analyses.
RESULTS
CHILDREN WITH HIGHER GRSs WERE LARGER
AND GREW FASTER DURING CHILDHOOD
AND DURING ADULTHOOD
ChildrenwithhigherGRSshadhigherBMIsateveryage
assessed,fromages3through38years(Table 1).Inthe
life-coursegrowthmodel,higherGRSspredictedhigher
Table 1. Descriptive Statistics and Correlations With GRS
and Family History Score for Anthropometric Assessments
Among 856 Individuals of European Descent
Assessmenta
BMI
3
5
7
9
11
13
15
18
21
26
32
38
Alternative measures of
adiposity
Subscapular skinfold
thickness, mm
7
9
Tricep skinfold
thickness, mm
7
9
Waist circumference,
mm
26
32
38
Fat mass, kg
32
38
Gestational and childhood
growth
Birth weight, kg
Weight gain from birth
to age 3 y, kg
Adiposity rebound
Age, y
BMI
Mean (SD)
Measurement
Correlation, r
With
GRSb
With
Family
History
Scoreb
16.33 (1.28)
15.88 (1.18)
15.82 (1.29)
16.33 (1.62)
17.49 (2.09)
19.59 (2.47)
20.37 (2.60)
22.73 (3.02)
23.74 (3.41)
24.89 (4.27)
26.03 (4.80)
26.92 (5.13)
0.08c
0.13e
0.17e
0.18e
0.16e
0.18e
0.17e
0.15e
0.18e
0.16e
0.13e
0.14e
0.11d
0.19e
0.24e
0.21e
0.22e
0.23e
0.30e
0.31e
0.31e
0.29e
0.31e
0.34e
5.87 (1.92)
6.67 (2.80)
0.07d
0.11e
0.12d
0.10d
8.46 (2.46)
11.06 (4.12)
0.07c
0.09d
0.10d
0.10d
798.68 (96.40)
841.81 (110.95)
858.39 (122.84)
0.14e
0.11e
0.12e
0.24e
0.25e
0.26e
21.41 (10.54)
23.59 (10.96)
0.10d
0.11d
0.23e
0.27e
3.38 (0.52)
11.31 (1.52)
0.00
0.09c
0.07c
0.02
6.11 (1.10)
15.57 (1.00)
−0.13e
0.17e
−0.21e
0.23e
Abbreviations: BMI, body mass index (calculated as weight in kilograms
divided by height in meters squared); GRS, genetic risk score.
aUnless otherwise indicated, all measurements are given by year of age at
which the measurement was obtained.
bAll correlations were adjusted for sex. Tests of statistical significance
were conducted using heteroskedasticity robust standard errors.
cP ? .05.
dP ? .01.
eP ? .001.
ARCH PEDIATR ADOLESC MED/VOL 166 (NO. 6), JUNE 2012WWW.ARCHPEDIATRICS.COM
517
©2012 American Medical Association. All rights reserved.

Page 4

Confidential. Do not distribute. Pre-embargo material.
meanlevelsofBMI(intercept?=0.38[P?.001]),faster
growth in childhood (?=0.03 [P?.001]), and faster
growthinadulthood(?=0.02[P=.02]). Figure 2shows
life-coursegrowthcurvesforchildrenwithhigh,low,and
average GRSs.
To rule out the possibility that variation at the FTO lo-
cus (GenBank NC_007316.4) accounted for our ob-
served GRS-growth associations, we repeated the analy-
sis,adjustingslopeandinterceptestimatesfortheFTOSNP
rs9939609. This SNP is the best-replicated GWAS result
forBMI,27hasbeenshowntoinfluencegrowth,6,28andcar-
ried the largest weight of any SNP in our GRS. Associa-
tions of GRS and growth were unchanged by adjustment
forrs9939609.Independentoftheirrs9939609genotype,
childrenwithhigherGRSswerelargeracross4decadesof
follow-up(intercept?=0.40[P?.001])andgrewfasterdur-
ingchildhoodandduringadulthood(childhoodlinearslope
?=0.03 [P=.003]; adult linear slope ?=0.02 [P=.01]).
To rule out the possibility that GRS-growth associa-
tions reflected associations with height or muscle mass
and not with adiposity, we tested associations between
the GRS and childhood skinfold thicknesses and adult
waist-girth and fat-mass measurements. These measure-
ments are less susceptible to inflation as a result of body
size and are considered to be more direct measures of
body fat.19The GRS correlations with these alternative
measures of adiposity were statistically significant and
were similar to GRS correlations with BMI (Table 1).
CHILDREN WITH HIGHER GRSs
WERE AT GREATER RISK FOR OBESITY
ACROSS 2 DECADES OF ADULT FOLLOW-UP
As teenagers (ages 15-18 years), 5.5% of Dunedin Study
children had BMIs in the obese range; in their third de-
cade of life (ages 21-26 years), 11.2% met criteria for
obesity; and in their fourth decade of life (ages 32-38
years), 22.3% met criteria for obesity, consistent with
the nationwide prevalence among New Zealanders of
Europeandescent(http://socialreport.msd.govt.nz/health
/obesity.html).Weclassified8.4%ofthesampleaschroni-
callyobese. Figure 3showsobesityprevalencesforchil-
dren at low (below average) and high (above average)
genetic risk. Children at high genetic risk were 1.61 to
2.41timesmorelikelytobeobeseintheirsecond,third,
andfourthdecadesoflifeandwere1.90timesmorelikely
to be chronically obese across more than 3 assessments
compared with children at low genetic risk.
POLYGENIC RISK FOR ADULT OBESITY
IS MEDIATED BY DEVELOPMENTAL
PHENOTYPES OF RAPID CHILDHOOD GROWTH
To determine whether genetic risk for obesity was me-
diated by rapid early growth, we investigated relation-
shipsamongthechildren’sGRSs,theirgrowthduringges-
tationandchildhood,andtheirobesityoutcomesacross
ages 15 to 38 years.
The first developmental period theorized to entrain
adultobesityriskisgestation.However,theGRSwasnot
associated with fetal growth as indexed by birth weight
(r=0.00[P?.90])(Table1).Nevertheless,by3yearsof
age,childrenathighergeneticriskhadhigherBMIsrela-
tive to their peers (r=0.08 [P=.04]), raising the ques-
tionwhethergrowthduringaseconddevelopmentalpe-
riod, from birth to 3 years of age, mediated the genetic
risk for obesity. Children at higher genetic risk gained
moreweightfrombirthto3yearsofage(r=0.09[P=.01])
(Table 1). Consistent with previous research,29,30chil-
drenwithmorerapidweightgainduringtheseyearswere
more likely to become obese (Table 2). Decomposi-
tion of GRS-obesity associations into direct and indirect
effects indicated that weight gain from birth to 3 years
of age mediated statistically significant portions of ge-
30.00
20.00
25.00
15.00
38 1318 23 28 3338
Age, y
Body Mass Index
Figure 2. Life-course growth curves for children with high, low, and average
genetic risk scores (GRSs). Individuals with higher-obesity GRSs were larger
and grew more rapidly as children and adults. The solid line represents the
population mean trajectory (average genetic risk). Dashed lines are for
subgroups within 1 SD of the GRS (high and low genetic risk). Trajectories
were derived from the life-course growth model (intercept fitted at 13 years
of age; linear and quadratic slopes fitted during ages 3-13 years and 13-38
years), including intercept and linear slope effects for the GRS. Analyses
included 856 individuals of European descent. Body mass index is calculated
as weight in kilograms divided by height in meters squared.
40
20
30
0
10
Second Third Fourth Chronic
(Ages 15-38 y)
Decade of Life
Obese, % of Individuals
RR = 2.41
(1.31-4.41)
RR = 1.76
(1.20-2.63)
RR = 1.61
(1.25-2.10)
RR = 1.90
(1.21-3.07)
Low genetic riskHigh genetic risk
Figure 3. Obesity prevalence among low and high genetic risk cohort
members in their second, third, and fourth decades of life and chronically
across ages 15 to 38 years. Individuals with higher genetic risk scores
(GRSs) were more likely to be obese across 2 decades of adult follow-up.
Error bars and numbers in parentheses reflect 95% CIs. The GRS was
dichotomized at the sample mean to create low and high genetic risk
categories. Relative risks (RRs) (95% CIs) are reported from Poisson
regression models adjusted for sex that included the 856 individuals of
European descent in the analysis sample.
ARCH PEDIATR ADOLESC MED/VOL 166 (NO. 6), JUNE 2012WWW.ARCHPEDIATRICS.COM
518
©2012 American Medical Association. All rights reserved.

Page 5

Confidential. Do not distribute. Pre-embargo material.
neticriskforobesityintheteenageyearsandforchronic
obesity, but not for obesity, in the third or fourth de-
cade of life individually (Table 2).
Adiposity rebound, when children begin to gain
body fat after losing it during early childhood, is a
third period in development theorized to entrain adult
obesity. For children at higher genetic risk, adiposity
rebound occurred earlier in development and at
higher BMI (r=−0.13 for age and r=0.17 for BMI
[P?.001 for both]) (Table 1). Consistent with previ-
ous research,11,31children with earlier adiposity
rebound and higher BMI at adiposity rebound were
more likely to become obese (Table 2). Decomposition
of GRS-obesity associations into direct and indirect
effects revealed that adiposity rebound mediated large
and statistically significant portions of genetic risk for
obesity in the second, third, and fourth decades of life
and for chronic obesity (Table 2).
THE GRS CONTAINED INFORMATION
ABOUT CHILDREN’S GROWTH
AND THEIR RISK FOR OBESITY IN ADULTHOOD
THAT WAS NOT AVAILABLE
IN THEIR FAMILY HISTORIES
Higher genetic risk predicted faster growth and in-
creased risk for obesity in children with normal-weight
and overweight parents (Figure 4). That is, the GRS
contributed independent and additive information to
the prediction of children’s growth and their risk for
obesity in adulthood beyond the family history infor-
mation (eTable 2).
COMMENT
Weconductedadevelopmentalgeneticinvestigationinto
theetiologyofobesityinaprospectivebirthcohortstudy
with4decadesoffollow-up.Wemeasuredpolygenicrisk
forobesityusingamultilocusGRSderivedfromGWASs
ofobesity-relatedphenotypes.Ouranalysesrevealedthat
polygenic risk for obesity was partly mediated by rapid
growthintheearlychildhoodyearsafterbirth.Thisfind-
ing supported our hypothesis that developmental phe-
notypes were critical in linking a genetic predisposition
to adult obesity. Furthermore, risk for obesity mea-
sured by the genetic risk score was independent of risk
information available in parental BMI.
These findings have implications for clinical practice
andfordevelopmentalandepidemiologicresearch.First,
the results suggest promise for using genetic informa-
tion in obesity risk assessments. Parental BMI has been
proposed as a screening measure to target obesity pre-
vention in children on the basis of effect-size correla-
tions only slightly larger than those we report for our
GRS.32New developments in genome science, includ-
ing next-generation sequencing, may uncover new vari-
ants that further improve the performance of SNP-
basedriskassessments.33-35Moreover,theGRScontained
informationaboutchildren’sfutureobesityriskthatcould
not be derived from measurements of parents, suggest-
ingthatpositivefamilyhistorymaynotalwaysbeanap-
propriateprerequisiteforgenetictesting.Second,ourfind-
ingsillustratehowpolygenicinfluencesondevelopment
can be investigated using the GRS. Prospective cohort
Table 2. Mediation of Polygenic Risk for Adult Obesity by Developmental Phenotypes of Rapid Early Growtha
Obesity by Decade of Life, No. (%) Ever Obese
Second
(n = 47) [5.5]
Third
(n = 96) [11.2]
Fourth
(n = 191) [22.3]
Chronicb
(n = 72) [8.4]
Bivariate
Modelsc
Multivariate
Modeld
Polygenic Risk for Obesity Is Partly Mediated by Weight Gain Between Birth and Age 3 y
1.30 (1.00-1.69) 1.37 (1.13-1.68) 1.34 (1.10-1.63)
Bivariate
Modelsc
Multivariate
Modeld
Bivariate
Modelsc
Multivariate
Modeld
Bivariate
Modelsc
Multivariate
Modeld
GRS, RR
(95% CI)
Weight gain, RR
(95% CI)
Mediation ratioe
P valuef
1.39 (1.08-1.81) 1.23 (1.08-1.39)1.21 (1.07-1.37)1.37 (1.09-1.74) 1.32 (1.05-1.67)
1.78 (1.38-2.30)1.72 (1.33-2.22)1.32 (1.10-1.59) 1.28 (1.06-1.54) 1.15 (1.02-1.31)1.13 (1.00-1.28)1.44 (1.15-1.80)1.39 (1.12-1.73)
0.16
.02
0.07
.06
0.06
.12
0.10
.04
Polygenic Risk for Obesity Is Partly Mediated by Age and BMI at Adiposity Rebound
1.18 (0.93-1.49)1.37 (1.13-1.68) 1.20 (1.00-1.45)GRS, RR
(95% CI)
Age, RR
(95% CI)
BMI, RR
(95% CI)
Mediation ratioe
P valuef
1.39 (1.08-1.81) 1.23 (1.08-1.39)1.14 (1.01-1.29)1.37 (1.09-1.74) 1.20 (0.96-1.50)
0.57 (0.48-0.68)0.57 (0.48-0.68)0.66 (0.58-0.75)0.66 (0.58-0.75) 0.77 (0.70-0.85)0.78 (0.70-0.85)0.66 (0.57-0.76) 0.66 (0.57-0.76)
2.13 (1.70-2.66)2.09 (1.66-2.64)1.61 (1.36-1.89)1.56 (1.32-1.85) 1.35 (1.20-1.51)1.33 (1.19-1.49)1.72 (1.44-2.05)1.68 (1.39-2.02)
0.58
?.001
0.44
?.001
0.41
?.001
0.45
?.001
Abbreviations: BMI, body mass index; GRS, genetic risk score; RR, relative risk.
aAll analyses were adjusted for sex and included the 856 individuals of European descent in the analysis sample. Weight gain from birth through age 3 years and
adiposity rebound measures were standardized to have means of 0 and standard deviations of 1 for analyses.
bObese at 50% or more of assessments between ages 15 and 38 years.
cIndicates bivariate effects of the GRS and the developmental phenotypes from Poisson regression models.
dIndicates the independent effects of the GRS and the developmental phenotypes from multivariate Poisson regression models.
eDescribes how much of the effect of genetic risk is mediated by the developmental phenotype. Mediation ratios were calculated from the indirect and direct effects
estimated from structural equations (eTable 3).
fCalculated from the Sobel test of mediation.
ARCH PEDIATR ADOLESC MED/VOL 166 (NO. 6), JUNE 2012 WWW.ARCHPEDIATRICS.COM
519
©2012 American Medical Association. All rights reserved.

Page 6

Confidential. Do not distribute. Pre-embargo material.
studiescontainingrepeatedmeasuresarenecessarytoelu-
cidate developmental processes leading to complex dis-
eases.36However, to date, small single-locus effect sizes
have made it challenging to incorporate genetic infor-
mationintoongoingcohortstudies.Toaddressthechal-
lenge of small effects, we used a multilocus profile. The
resultingGRSenablesmeasurementofalarger,genome-
wide effect size and reduces the number of hypothesis
tests to 1, making follow-up of GWAS findings tractable
in cohort studies that are needed to study development.
Third, the longitudinal results illustrate that investiga-
tions of obesity as an outcome to developmental pro-
cesses can inform public health initiatives and research
priorities by identifying specific phases in development
when genetic risk becomes manifest and thus might be
amenabletointervention.Childhoodgrowthingeneral—
and,inparticular,growthduringtheperiodbetweenbirth
and the adiposity rebound—should be a focus for fu-
tureresearchtounderstandgeneticcontributionstothe
development of obesity.
We acknowledge 3 limitations. First, we derived our
GRSfromGWASsofEuropeansandconductedourstudy
inindividualsofEuropeandescent;theseresultsmaynot
generalizetootherpopulations.37Second,ourfamilyhis-
tories included only parents. More complete family his-
tories might have greater overlap with the GRS. Third,
we were unable to characterize growth trajectories dur-
ingtheearlieststagesoflife;regularfollow-upoftheco-
hort did not begin until 3 years of age. However, results
fromouranalysesofbirthweightandofweightgainfrom
birth though 3 years of age were consistent with previ-
ous genetic investigations of this interval that did in-
clude repeated measurements.5,6,38,39Moreover, we were
able to capture growth from 3 years of age and onward
with a high degree of resolution; our study included 12
measurements taken during the subsequent 35 years. In
addition to repeated measures of height and weight, our
study included more direct measures of adiposity, in-
cluding childhood measurements of skinfold thick-
nesses and adult measurements of waist circumference
and fat mass, all of which were associated with our GRS
in parallel to BMI. Thus, the results present compelling
evidencethatSNPsidentifiedinGWASsofadultBMIand
otherobesity-relatedphenotypespredisposetomorerapid
growth in childhood, leading to increased risk for obe-
sity in adulthood, and provide information not forth-
coming from a simple analysis of family history.
Accepted for Publication: February 2, 2012.
Author Affiliations: Department of Health Policy and
Management, Gillings School of Public Health (Drs Bel-
sky and Biddle), and Department of Genetics, School of
Medicine(DrEvans),UniversityofNorthCarolina,Cha-
pelHill;DepartmentofPsychologyandNeuroscience(Dr
Belsky, Moffitt, Houts, Bennett, Sugden, and Caspi; Ms
Harrington; and Mr Williams) and Institute for Ge-
nomeSciencesandPolicy(DrBelsky,Moffitt,Houts,Sug-
den,andCaspi;MsHarrington;andMrWilliams),Duke
University, and Department of Psychiatry and Behav-
ioralSciences(DrBelsky,Moffitt,Houts,Blumenthal,Sug-
den,andCaspi;MsHarrington;andMrWilliams),Duke
UniversityMedicalCenter,Durham,NorthCarolina;So-
cial, Genetic, and Developmental Psychiatry Centre, In-
stituteofPsychiatry,King’sCollegeLondon,London,En-
gland(DrsMoffitt,Sugden,andCaspiandMrWilliams);
and Dunedin Multidisciplinary Health and Develop-
mentResearchUnit,UniversityofOtago,Dunedin,New
Zealand (Dr Poulton).
Correspondence: Daniel W. Belsky, PhD, Institute for
Genome Sciences and Policy, Duke University, Grey
House, Ste 201, Duke University Box 104410, Durham,
NC 27708 (dbelsky@duke.edu).
Author Contributions: Study concept and design: Bel-
sky, Moffitt, and Caspi. Acquisition of data: Moffitt, Sug-
den, Williams, Poulton, and Caspi. Analysis and inter-
pretationofdata:Belsky,Moffitt,Houts,Bennett,Biddle,
Blumenthal,Evans,Harrington,Sugden,Williams,Poul-
ton, and Caspi. Drafting of the manuscript: Belsky, Mof-
fitt, and Caspi. Critical revision of the manuscript for im-
portantintellectualcontent:Belsky,Moffitt,Houts,Bennett,
Biddle, Blumenthal, Evans, Harrington, Sugden, Wil-
liams, Poulton, and Caspi. Statistical analysis: Belsky,
Houts,andCaspi.Obtainedfunding:Belsky,Moffitt,Poul-
ton,andCaspi.Administrative,technical,andmaterialsup-
30.00
20.00
25.00
15.00
Body Mass Index
A
3813 18 23 283338
Age, y
40
20
30
0
10
Obese, % of Individuals
Low familial risk, low genetic risk
Low familial risk, high genetic risk
High familial risk, low genetic risk
High familial risk, high genetic risk
B
SecondThirdFourthChronic
(Ages 15-38 y)
Decade of Life
High familial risk, high genetic risk
Low familial risk, low genetic risk
Low familial risk, high genetic risk
High familial risk, low genetic risk
Figure 4. Influence of genetic risk and family history on growth and obesity
risk. The genetic risk score (GRS) contained information about children’s
growth and their obesity risk that was not available in their family histories.
Genetic risk and family history made independent and additive contributions
to life-course growth predictions and to adult obesity risk in 856 individuals
of European descent. A, Life-course growth curves show that genetic risk
and family history made additive contributions to growth predictions. B, Bar
graph shows that genetic risk and family history made additive contributions
to children’s risk of becoming obese. Error bars reflect 95% CIs. Statistical
analyses illustrating the independence of the GRS and family history in
predicting growth and obesity risk are presented in eTable 2. Body mass
index is calculated as weight in kilograms divided by height in meters
squared.
ARCH PEDIATR ADOLESC MED/VOL 166 (NO. 6), JUNE 2012WWW.ARCHPEDIATRICS.COM
520
©2012 American Medical Association. All rights reserved.

Page 7

Confidential. Do not distribute. Pre-embargo material.
port:Moffitt,Harrington,Sugden,Williams,Poulton,and
Caspi. Study supervision: Moffitt, Poulton, and Caspi.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grant
G0601483 from the UK Medical Research Council,
grant AG032282 from the National Institute on Aging,
andgrantMH077874fromtheNationalInstituteofMen-
tal Health. Additional support was provided by the Ja-
cobsFoundationandfellowship1R36HS020524-01from
theAgencyforHealthcareResearchandQuality(DrBel-
sky). The Dunedin Multidisciplinary Health and Devel-
opment Research Unit was supported by the New Zea-
land Health Research Council.
Online-OnlyMaterial:TheeMethods,3eTables,eFigure,
andeReferencesareavailableathttp://www.archpediatrics
.com.
AdditionalContributions:WethanktheDunedinStudy
members, their families, unit research staff, and study
founder Phil Silva, PhD.
REFERENCES
1. McCarthy MI. Genomics, type 2 diabetes, and obesity. N Engl J Med. 2010;363
(24):2339-2350.
2. Oken E, Gillman MW. Fetal origins of obesity. Obes Res. 2003;11(4):496-506.
3. FrongilloEA,LamplM.Earlyidentificationofchildrenatriskofdevelopingobesity.
Arch Pediatr Adolesc Med. 2011;165(11):1043-1044.
4. North KE, Graff M, Adair LS, et al. Genetic epidemiology of BMI and body mass
change from adolescence to young adulthood. Obesity (Silver Spring). 2010;
18(7):1474-1476.
5. Elks CE, Loos RJ, Sharp SJ, et al. Genetic markers of adult obesity risk are as-
sociated with greater early infancy weight gain and growth. PLoS Med. 2010;
7(5):e1000284. doi:10.1371/journal.pmed.1000284.
6. SovioU,Mook-KanamoriDO,WarringtonNM,etal;EarlyGrowthGeneticsCon-
sortium. Association between common variation at the FTO locus and changes
in body mass index from infancy to late childhood: the complex nature of ge-
netic association through growth and development. PLoS Genet. 2011;7(2):
e1001307. doi:10.1371/journal.pgen.1001307.
7. McCarthy MI, Abecasis GR, Cardon LR, et al. Genome-wide association studies
forcomplextraits:consensus,uncertaintyandchallenges.NatRevGenet.2008;
9(5):356-369.
8. Plomin R, Haworth CMA, Davis OSP. Common disorders are quantitative traits.
Nat Rev Genet. 2009;10(12):872-878.
9. KathiresanS,MelanderO,AnevskiD,etal.Polymorphismsassociatedwithcholes-
terolandriskofcardiovascularevents.NEnglJMed.2008;358(12):1240-1249.
10. Meigs JB, Shrader P, Sullivan LM, et al. Genotype score in addition to common
risk factors for prediction of type 2 diabetes [published correction appears in
N Engl J Med. 2009;360(6):648]. N Engl J Med. 2008;359(21):2208-2219.
11. Rolland-CacheraMF,DeheegerM,MaillotM,BellisleF.Earlyadiposityrebound:
causes and consequences for obesity in children and adults. Int J Obes (Lond).
2006;30(suppl 4):S11-S17.
12. DietzWH.Overweightinchildhoodandadolescence.NEnglJMed.2004;350(9):
855-857.
13. Taylor RW, Grant AM, Goulding A, Williams SM. Early adiposity rebound: re-
view of papers linking this to subsequent obesity in children and adults. Curr
Opin Clin Nutr Metab Care. 2005;8(6):607-612.
14. Yang WJ, Kelly T, He J. Genetic epidemiology of obesity. Epidemiol Rev. 2007;
29:49-61.
15. Speliotes EK, Willer CJ, Berndt SI, et al; MAGIC; Procardis Consortium. Asso-
ciation analyses of 249,796 individuals reveal 18 new loci associated with body
mass index. Nat Genet. 2010;42(11):937-948.
16. GENEVA ARIC Project: Quality Control Report for the ARIC GWAS database. Ac-
cession No. phs000090.v1.p1 ed. 2009. http://www.ncbi.nlm.nih.gov/projects
/gap/cgi-bin/study.cgi?study_id=phs000090.v1.p1. Accessed May 18, 2011.
17. Janssen I, Katzmarzyk PT, Srinivasan SR, et al. Utility of childhood BMI in the
prediction of adulthood disease: comparison of national and international
references. Obes Res. 2005;13(6):1106-1115.
18. Abdullah A, Wolfe R, Stoelwinder JU, et al. The number of years lived with obe-
sityandtheriskofall-causeandcause-specificmortality.IntJEpidemiol.2011;
40(4):985-996.
19. LobsteinT,BaurL,UauyR.Obesityinchildrenandyoungpeople:acrisisinpub-
lic health. Obes Rev. 2004;5(suppl 1):4-104.
20. SingerJD,WillettJB.AppliedLongitudinalDataAnalysis.NewYork,NY:Oxford
University Press; 2003.
21. MackinnonDP,DwyerJH.Estimatingmediatedeffectsinpreventionstudies.Eval
Rev. 1993;17(2):144-158.
22. Preacher KJ, Hayes AF. SPSS and SAS procedures for estimating indirect ef-
fects in simple mediation models. Behav Res Methods Instrum Comput. 2004;
36(4):717-731.
23. PreacherKJ,HayesAF.Asymptoticandresamplingstrategiesforassessingand
comparingindirecteffectsinmultiplemediatormodels.BehavResMethods.2008;
40(3):879-891.
24. Preacher KJ, Kelley K. Effect size measures for mediation models: quantitative
strategies for communicating indirect effects. Psychol Methods. 2011;16(2):
93-115.
25. SAS [computer program]. Version 9.2. Durham, NC: SAS Institute Inc; 2002-
2008.
26. Stata/MP for Windows [computer program]. Version 11.1. College Station, TX:
Stata Corp LP; 2010.
27. HindorffLA,JunkinsHA,MehtaJP,ManolioTA.ACatalogofPublishedGenome-
WideAssociationStudies.http://www.genome.gov/gwastudies/.AccessedApril
30, 2010.
28. Hardy R, Wills AK, Wong A, et al. Life course variations in the associations be-
tween FTO and MC4R gene variants and body size. Hum Mol Genet. 2010;
19(3):545-552.
29. Eriksson J, Forse ´n T, Tuomilehto J, Osmond C, Barker D. Size at birth, child-
hood growth and obesity in adult life. Int J Obes Relat Metab Disord. 2001;
25(5):735-740.
30. Li H, Stein AD, Barnhart HX, Ramakrishnan U, Martorell R. Associations be-
tween prenatal and postnatal growth and adult body size and composition. Am
J Clin Nutr. 2003;77(6):1498-1505.
31. Williams SM, Goulding A. Patterns of growth associated with the timing of adi-
posity rebound. Obesity (Silver Spring). 2009;17(2):335-341.
32. Whitaker KL, Jarvis MJ, Beeken RJ, Boniface D, Wardle J. Comparing maternal
andpaternalintergenerationaltransmissionofobesityriskinalargepopulation-
based sample. Am J Clin Nutr. 2010;91(6):1560-1567.
33. EichlerEE,FlintJ,GibsonG,etal.Missingheritabilityandstrategiesforfindingthe
underlying causes of complex disease. Nat Rev Genet. 2010;11(6):446-450.
34. GibsonG.HintsofhiddenheritabilityinGWAS.NatGenet.2010;42(7):558-560.
35. Qiu J, Hayden EC. Genomics sizes up. Nature. 2008;451(7176):234. doi:10
.1038/451234a.
36. Manolio TA, Bailey-Wilson JE, Collins FS. Genes, environment and the value of
prospective cohort studies. Nat Rev Genet. 2006;7(10):812-820.
37. MatiseTC,AmbiteJL,BuyskeS,etal;PAGEStudy.ThenextPAGEinunderstand-
ing complex traits: design for the analysis of Population Architecture Using Ge-
netics and Epidemiology (PAGE) Study. Am J Epidemiol. 2011;174(7):849-859.
38. Andersson EA, Pilgaard K, Pisinger C, et al. Do gene variants influencing adult
adiposity affect birth weight? a population-based study of 24 loci in 4,744 Dan-
ishindividuals.PLoSOne.2010;5(12):e14190.doi:10.1371/journal.pone.0014190.
39. Kilpela ¨inen TO, den Hoed M, Ong KK, et al; Early Growth Genetics Consortium.
Obesity-susceptibility loci have a limited influence on birth weight: a meta-
analysis of up to 28,219 individuals. Am J Clin Nutr. 2011;93(4):851-860.
ARCH PEDIATR ADOLESC MED/VOL 166 (NO. 6), JUNE 2012 WWW.ARCHPEDIATRICS.COM
521
©2012 American Medical Association. All rights reserved.