1
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s
De elopmen o a Gene ic
Risk Sco e o p edic he isk
o o e weigh and obesi y
in Eu opean adolescen s
om he HELENA s udy
Miguel Se al‑Co es1*, Se gio Sab oso‑Lasa2, Pila De Miguel‑E ayo1,3,4,
Ma cela Gonzalez‑G oss4,5,6, E a Ges ei o6, C is ina Molina‑Hidalgo7, S e aan De Henauw8,
F ede ic Go and9, Ch is ina Ma ogianni10, Yannis Manios10, Ma ia Plada11, Ku Widhalm12,
An hony Ka a os13, É a E ha d 14, Aline Mei haeghe15, Diego Salaza ‑To osa16,
Jona an Ruiz17, Luis A. Mo eno1,3,4, Luis Ma iano Es eban18 & Idoia Labayen19
Obesi y is he esul o in e ac ions be ween genes and en i onmen al ac o s. Since monogenic
e iology is only known in some obesi y‑ ela ed genes, a gene ic isk sco e (GRS) could be use ul
o de e mine he gene ic p edisposi ion o obesi y. The e o e, he aim o ou s udy was o build a
GRS able o p edic gene ic p edisposi ion o o e weigh and obesi y in Eu opean adolescen s. A
o al o 1069 adolescen s (51.3% emale), aged 11–19 yea s pa icipa ing in he Heal hy Li es yle in
Eu ope by Nu i ion in Adolescence (HELENA) c oss‑sec ional s udy we e geno yped. The sample
was di ided in non‑o e weigh (non‑OW) and o e weigh /obesi y (OW/OB). F om 611 single
nucleo ide polymo phisms (SNP) a ailable, a i s sc eening o 104 SNPs uni a ia ely associa ed wi h
obesi y (p < 0.20) was es ablished selec ing 21 signi ican SNPs (p < 0.05) in he mul i a ia e model.
Unweigh ed GRS (uGRS) was calcula ed by summing he numbe o isk alleles and weigh ed GRS
(wGRS) by mul iplying he isk alleles o each es ima ed coe icien . The a ea unde cu e (AUC) was
calcula ed in uGRS (0.723) and wGRS (0.734) using en old in e nal c oss‑ alida ion. Bo h uGRS and
wGRS we e signi ican ly associa ed wi h body mass index (BMI) (p < .001). Bo h GRSs could po en ially
be conside ed as use ul gene ic ools o e alua e indi idual’s p edisposi ion o o e weigh /obesi y in
Eu opean adolescen s.
OPEN
1G ow h, Exe cise, NU i ion and De elopmen (GENUD) Resea ch G oup, Ins i u o Ag oalimen a io de A agón
(IA2), Ins i u o de In es igación Sani a ia A agón (IIS A agón), Uni e sidad de Za agoza, Za agoza, Spain. 2Spanish
Na ional Cance Resea ch Cen e (CNIO), Mad id, Spain. 3Ins i u o de In es igación Sani a ia A agón (IIS A agón),
Za agoza, Spain. 4Cen o de In es igación Biomédica en Red de Fisiopa ología de la Obesidad y la Nu ición
(CIBERObn), Ins i u o de Salud Ca los III, Mad id, Spain. 5Ins i u e o Nu i ional and Food Sciences, Nu i ional
Physiology, Uni e si y o Bonn, Bonn, Ge many. 6ImFine Resea ch G oup, Depa men o Heal h and Human
Pe o mance, Facul ad de Ciencias de la Ac i idad Física y del Depo e-INEF, Uni e sidad Poli écnica de Mad id,
Mad id, Spain. 7EFFECTS 262 Depa men o Medical Physiology, School o Medicine, Uni e si y o G anada,
18071 G anada, Spain. 8Depa men o Public Heal h, Facul y o Medicine and Heal h Sciences, Ghen Uni e si y,
Ghen , Belgium. 9Facul y o Medicine, Uni e si y Lille, Lille, F ance. 10Depa men o Nu i ion and Die e ics, School
o Heal h Science and Educa ion, Ha okopio Uni e si y, A hens, G eece. 11Uni e si y o C e e School o Medicine,
C e e, G eece. 12Di ision o Gas oen e ology and Hepa ology, Depa men o In e nal Medicine III, Medical
Uni e si y o Vienna, Aus ia and Aus ian Academic Ins i u e o Clinical Nu i ion, Vienna, Aus ia. 13Facul y
o Medicine, Uni e si y o C e e, C e e, G eece. 14Depa men o Pedia ics, Medical School, Uni e si y o Pécs,
Pecs, Hunga y. 15UMR1167, RID-AGE, Risk Fac o s and Molecula De e minan s o Aging-Rela ed Diseases,
Cen e Hosp. Uni Lille, Ins i u Pas eu de Lille, Uni e si é de Lille, Lille, F ance. 16Depa men o Ecology and
E olu iona y Biology, Uni e si y o A izona, Tucson, AZ, USA. 17Depa men o de Ac i idad Física y Depo e,
Facul y o Spo Sciences, Uni e sidad de G anada, G anada, Spain. 18Escuela Poli écnica de La Almunia,
Uni e sidad de Za agoza, Za agoza, Spain. 19Depa men o Heal h Sciences, Public Uni e si y o Na a a,
Pamplona, Spain. *email: mse al@uniza .es
2
Vol:.(1234567890)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s/
Childhood obesi y is a majo public heal h p oblem1. Pedia ic obesi y inc eases he isk o physical and psy-
chological heal h p oblems al eady in childhood, and la e in adul hood2. Mo e so, adiposi y ela ed diso de s
p edominan ly diagnosed in adul s such as ype 2 diabe es melli us (T2DM) and ca dio ascula diseases migh
o igina e in ea ly li e, and po en ially educe li e expec ancy. O e he las wo decades, la ge-scale s udies ha e
been un eiling new common a ian s in locus o ce ain genes ela ed wi h childhood and adul obesi y3–5. A
leas 97 loci ha e been associa ed wi h obesi y6. Cu en ly, he FTO gene s ill emains he locus explaining he
la ges associa ion wi h obesi y in adul s, child en and adolescen s7,8. In his ega d, p e ious s udies ha e shown
ha each copy o he FTO s9939609 polymo phism A allele is associa ed wi h 2.8% highe body a in Eu opean
adolescen s9,10. Some s udies ha e ound some associa ions be ween single nucleo ide polymo phisms (SNPs)
wi h obesi y isk ac o s, being po en ially use ul as ea ly li e isk indica o s in child en and adul s11. Howe e ,
indi idual SNPs can explain li le o disease a iance12. Se e al s udies ha e demons a ed he po en ial alue o
o he gene ic app oaches ha combine a numbe o SNPs o de elop a gene ic isk sco e (GRS) by summing he
numbe o isk alleles: unweigh ed GRS (uGRS) o by mul iplying he numbe o isk alleles o each es ima ed
coe icien : weigh ed GRS (wGRS)13–15. The c ea ion and alida ion o obesi y-speci ic GRS se s a landma k in
pe sonalised gene ic isk p edic ion o obesi y and obesi y- ela ed diseases16. Di e en obesi y- ela ed GRS ha e
been cons uc ed in adul s15–18 and child en19,20 wi h signi ican obesi y-gene associa ions, being implemen ed
on a a ie y o e hnic popula ion backg ounds. Wi hin Eu opean popula ions, Seyednas ollah e al.21 compu ed
wo weigh ed GRS (wGRS) o 97 and 19 SNPs p e iously ela ed o he isk o obesi y in wo coho s including
2262 Finnish child en and adolescen s (3–18yea s). Fu he , Viljakainen e al.22 de eloped a wGRS o p edic
he isk o o e weigh and obesi y in a coho o 1142 Finnish p eadolescen s (11.3 ± 0.2yea s) conside ing body
mass index (BMI) and 30 BMI- ela ed SNPs om p e ious genome-wide associa ion s udies (GWAS). As only
ew s udies es ing obesi y isk in Eu opean adolescen s wi h GRSs ha e been conduc ed, he aim o he p esen
s udy was o de elop a GRS o o e weigh and obesi y in adolescen s pa icipa ing in he Heal hy Li es yle in
Eu ope by Nu i ion in Adolescence (HELENA) c oss-sec ional s udy.
Me hods
S udy design and popula ion. The da a we e ex ac ed om he HELENA mul icen ic and c oss-sec-
ional s udy con aining a o al sample o 4356 adolescen s (51.6% emales), aged 11–19yea s old, om 10 Eu o-
pean ci ies loca ed in sepa a ed geog aphical poin s in Eu ope in 2006–2007. Thei size o he ci ies was la ge
enough o ensu e pa icipan s di e si y23. The main objec i e o he HELENA s udy was o ob ain compa able
da a o a la ge sample o Eu opean adolescen s on nu i ion and heal h- ela ed pa ame e s by a s anda dised
p ocedu e24. Mo e so, he s udy was pe o med ollowing he e hical guidelines o he Decla a ion o Helsinki
1964 ( e ision o 2013), he Good Clinical P ac ice, and he legisla ion abou clinical esea ch in humans in each
o he pa icipa ing coun ies and was app o ed by he E hics Commi ee o each ci y pa icipa ing in he s udy25.
The p o ocol was app o ed by he E hical Commi ee (Comi é de É ica de la In es igación de la Comunidad
Au ónoma de A agón: CEICA). W i en in o med consen and assen o pa icipa e in he s udy we e ob ained
om adolescen s and hei pa en s be o e being en olled. One hi d o he subjec s (N = 1172) om he o al
sample we e andomly selec ed o blood sampling24. A e including speci ic inclusion c i e ia om genomic
pa ame e s (SNPs) and an h opome y (BMI), a o al o 1069 adolescen s (51.3% emales) we e inally consid-
e ed o he analysis in he p esen s udy. The low cha o he selec ed sample is displayed in Supplemen a y
Fig.1.
Physical examina ion. All measu emen s we e pe o med by ained esea che s ollowing s anda d p o-
ocols. Weigh and heigh we e measu ed ollowing s anda d p ocedu es26. BMI was calcula ed om heigh and
weigh (kg/m2)27 and ca ego ised in o non-o e weigh (non-OW) and o e weigh , including obesi y (OW/OB),
acco ding o he age- and sex-speci ic BMI in e na ional cu -o s p oposed by he Wo ld Obesi y Fede a ion28.
FMI was calcula ed was calcula ed di iding a mass (FM) by heigh squa ed (in me e s).
Blood collec ion and geno yping. The blood samples we e collec ed in o e nigh as ing s a e. A s and-
a dised me hodology o blood collec ion, anspo and analysis was pe o med by a ce i ied labo a o y29.
Blood o DNA ex ac ion was collec ed in EDTA K3 ubes and s o ed a he Ins i u e o Nu i ional and Food
Sciences (IEL) o he Uni e si y o Bonn, and sen o he Labo a oi e d’Analyse Genomique Cen e de Ressou ces
Biologiques (LAG-CRB) e BB- 0033-00071 Ins i u Pas eu de Lille, F-59000 Lille, F ance. DNA was ex ac ed
om whi e blood cells wi h he Pu egene ki (QIAGEN, Cou aboeu , F ance) and s o ed a − 20°C. The geno-
yping was done by an Illumina sys em (Illumina, Inc, San Diego, Cali o nia) using he Golden- Ga e echnology
(Sampling p ocedu e scheme, GoldenGa e; So wa e, Inc, San F ancisco, Cali o nia). In e ms o gene selec-
ion wi hin he HELENA s udy, a candida e gene app oach was used. Fi s , a ce ain numbe o beha iou and
me abolic pa hways ela ed o he adolescence´s heal h we e iden i ied. These included included ood in ake,
ea ing beha iou , ood choices and p e e ences, ene gy and adipose issue me abolism, glucose, insulin, lipid and
lipop o ein me abolism among o he s. Finally, SNPs playing a key ole in genes coding o he abo emen ioned
pa hways we e selec ed. The HapMap da abase was used o selec ag and independen SNPs. SNPs we e selec ed
wi h a mino allele equency (MAF) abo e 0.1 and ag SNPs wi h 2 abo e 0.8. I ag SNPs desc ibed o a single
gene exceeded in numbe (mo e han ~ 20), only SNPs signi ican ly associa ed wi h app op ia e pheno ypes in
p e ious s udies we e selec ed, i a ailable. Las ly, SNPs om he NCBI da abase we e used when a limi ed num-
be o SNPs we e a ailable in he HapMap da abase.
S a is ical analysis. Desc ip i e cha ac e is ics by sex a e shown as median and in e qua ile ange (IQR)
o con inuous a iables and as absolu e and ela i e equency o ca ego ical ones. The s a is ical es s used o
3
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s/
compa e di e ences by sex we e Pea son´s chi-squa e o ca ego ical a iables and Mann–Whi ney-Wilcoxon
es o con inuous a iables. Pea son’s chi-squa e s a is ic es was used o analyse he Ha dy–Weinbe g equilib-
ium. Shapi o–Wilk non-pa ame ic es o check no mali y o a iables was pe o med.
All s a is ical analyses we e pe o med using RS udio Ve sion 1.2.5001 (RS udio Team (2015). RS udio: In e-
g a ed De elopmen o R. RS udio, Inc., Bos on, MA URLh p://www. s ud io.com/) and he signi icance le el
was se a p < 0.05.
De elopmen o he gene ic isk sco e. Candida e gene app oach was he p ocedu e based o selec he
genes in he HELENA s udy. Fi s , ele an adolescen ´s beha iou s and me abolic pa hways ela ed o heal h
we e iden i ied. Second, key p o eins ha , acco ding o he li e a u e, play a ole in hese pa hways we e also
iden i ied. Thi d, a selec ion o SNPs coding o hese p o eins was pe o med. Fou h, in o de o selec and
ag SNPs independen ly, he HapMap da abase (2007 elease) was used. SNPs wi h a mino allele equency
(MAF) abo e 0.1 and ag SNPs wi h 2 abo e 0.8 we e selec ed. I oo many ag SNPs desc ibed o a single gene
(mo e han ~ 20) we e iden i ied, only SNPs signi ican ly associa ed wi h app op ia e pheno ypes in p e ious
publica ions we e selec ed, i a ailable. Finally, SNPs om he NCBI da abase we e included when a oo limi ed
numbe o SNPs we e a ailable in he HapMap da abase. Based on he abo e, a o al o 611 SNPs ela ed o obe-
si y and obesi y- ela ed pheno ypes a ailable in he HELENA da ase 30 we e used o build a GRS conside ing
BMI as obesi y- ela ed a iable in o de o p edic a majo p edisposi ion o o e weigh /obesi y in Eu opean
adolescen s24. Each SNP was ecoded as 0, 1, o 2 depending on he numbe o isk alleles de ined in p e ious
li e a u e, espec i ely. A u he selec ion o SNPs was pe o med using gene alised linea model (GLM) o
es ablish an ini ial cu o poin (p < 0.20) o e ine he sea ch o 104 SNPs. Then, a s ep by s ep algo i hm was
applied o selec he signi ican SNPs unde he p < 0.05 h eshold in a mul i a ia e model o sho lis a inal
numbe o 21 SNPs signi ican ly associa ed wi h BMI. The co espondence be ween ac ual and p edic ed p ob-
abili ies o his model was analysed by a calib a ion cu e. The unweigh ed GRS (uGRS) was calcula ed by sum-
ming he numbe o isk alleles om he 21 SNP a ian s wi h a escaling, conside ing he SNPs ha appea as
p o ec o ac o s. The wGRS was he esul o mul iplying he numbe o isk alleles a each locus (0, 1, 2) o
each es ima ed coe icien o he mul i a ia e model. Pa icipan s wi h missing da a we e dismissed in he GRS
analysis (N = 3287). Recei e ope a ing cha ac e is ics (ROC) cu e analysis31 was applied o es he diagnos ic
accu acy o he GRS o classi y po en ial pa icipan s o obesi y associa ed dis u bances32. The a ea unde cu e
(AUC) was calcula ed in uGRS and wGRS conside ing weigh s a us as bina y a iable (i.e., non-OW s. OW/
OB). Selec ion o uGRS o e wGRS o p oceed wi h he design o he inal model was pe o med by he highe
alue o he AUC compa ed using he Delong es . The model was in e nally alida ed pe o ming en old c oss
alida ion analysis. Fo his analysis, he whole da ase was di ided in 10 g oups, using 9 o hem o build he
p edic i e model and he one o emains o alida e his model. This p ocedu e was epea ed aking in o accoun
all possible ways o selec he 9 subg oups, ensu ing di e en o ms o alida e GRS wi h da a no used in he
building model p ocess. Mo eo e , we e alua ed he dis ibu ion o uGRS and wGRS alues o NON-OW and
OW/OB in a boxplo o g aphically analyse he pe o mance o he GRS. In o de o p o ide he bes cu -o o
he use o he GRS as a dicho omic a iable, he maximisa ion o he Youden index33 was explo ed (see Table3).
Las ly, o es he GRS eliabili y wi h a gene al adiposi y es ima e o he han BMI, simple linea eg ession mod-
els (LRM) we e pe o med o e alua e he associa ion be ween a mass index (FMI) and bo h wGRS and uGRS.
In o med consen . In o med consen was signed by pa en s o all pa icipan s. The da ase s used and/o
analysed du ing he cu en s udy a e a ailable om he co esponding au ho on easonable eques .
Resul s
Desc ip ion o he s udy sample: demog aphics. The sample was composed o 520 boys and 549 gi ls.
Main cha ac e is ics o s udy pa icipan s a e shown in Supplemen a y Table1. The median age o pa icipan s
ba ely di e ed be ween males (14.6yea s, IQR: 13.6—15. 7) and emales (14.6yea s, IQR: 13.5–15.7) (p = 0.722).
The p e alence o OW/OB was highe in males (22.2%), han in emales (18.8%) (p = 0.009).
Associa ions be ween SNPs and o e weigh /obesi y: Building and alida ion o GRS. Ini ially,
104 SNPs po en ially associa ed wi h BMI we e selec ed (Supplemen a y Table2 and we e en e ed in he mul-
i a ia e model o build he GRS. F om hem, we ound a inal numbe o 21 SNPs signi ican ly associa ed wi h
OW/OB in he HELENA s udy Table1. Table2 shows he uni a ia e and mul i a ia e model´s odds a io (OR) o
each o he selec ed SNPs o he GRS build up. A o es plo is displayed in Fig.1 o p esen he OR´s di ec ion
(p o ec i e/ isk) o each SNP. Wi hin ou GRS, AMPD1 s2010899, PPARG s4135275, NR3C1 s4912905, LPA s
9,355,296, IL-6 s1524107, CNTFR s2183013, IGF1 s1019731, THRA s1568400 and FASN s4246444 had a
p o ec i e ole in he p edic ion o OW/OB whe eas NR3C1 s7701443, NR3C1 s13182800, CD36 s3211867,
CNTF s2515362, DRD2 s1800497, FTO s9939609, CETP s4783961, NOS2A s8068149, THRA s7502966,
ANGPTL4 s1044250, LXRβ s17373080 and PTPN1 s2143511 inc eased he isk o OW/OB. Supplemen a y
Fig.2 shows he calib a ion cu es analysing he co espondence be ween p obabili ies o o e weigh and he
eal ou come. I can be obse ed ha he e is a good ag eemen be ween p edic ed and ac ual p obabili ies, hus,
he panel o SNPs shows a good adjus men in o de o p edic OW/OB.
Using he p edic ABEL R package34, om he mul i a ia e logis ic eg ession model buil o p edic OW/
OB, an uGRS and a wGRS we e de i ed. The p edic i e abili y o he GRS by means o he ROC cu e, AUCs
and Youden index o uGRS and wGRS models a e displayed in Fig.2. The esul s o he GRSs´ AUC (uGRS:
0.7198,wGRS: 0.7336) showed a mode a e abili y o disc imina e OW/OB s a us. AUC´s compa isons displayed
s a is ically signi ican di e ences be ween uGRS and wGRS o p edic OW/OB (p = 0.043). The disc imina ion
4
Vol:.(1234567890)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s/
abili y o wGRS and uGRS was in e nally alida ed by c oss- alida ion using 10 olds. Bo h GRS p o ide obus
p edic ions as showed by he AUC esul s (0.723 and 0.734 o uGRS and wGRS espec i ely). The dis ibu ion
o uGRS and wGRS alues o he g oups o NON-OW and OW/OB by boxplo s is displayed in Fig.3. Bo h
GRS disc imina e be ween g oups, bu he e is no a cu -o poin ha can clea ly sepa a e NON-OW and OW/
OB g oups. The Youden index was 23.5 o uGRS (speci ici y 69.4%, sensi i i y 63.6%), and − 0.126 o wGRS
(speci ici y 61.1%, sensi i i y 74.6%). A mo e gene al analysis o sensi i i y, speci ici y, posi i e and nega i e
p edic i e alue, and accu acy is shown in Table3.
Associa ions be ween GRS and adiposi y. LRM showed a signi ican associa ion be ween GRS and
FMI (p ≤ 2e−16) in bo h weigh ed (β = 0.877) and unweigh ed (β = 0.312) a ian s o he o al sample.
Discussion
In he p esen s udy, wo GRSs, uGRS and wGRS, including 21 SNPs associa ed wi h BMI, we e success ully
de eloped o assess he isk o o e weigh and obesi y in Eu opean adolescen s. Hence, o he bes o ou knowl-
edge, his a e he i s GRSs wi h hese cha ac e is ics in a di e sely dis ibu ed sample o Eu opean adolescen s.
The e a e ew p e ious s udies ocusing on BMI-speci ic GRSs wi h o e weigh and obesi y in Eu opean
pedia ic popula ions and none exclusi ely in Eu opean adolescen s, which ein o ces he po en ial o ou GRS
analysis. In a coho o 1142 Finnish p eadolescen s, Viljakainen e al.22 cons uc ed a wGRS o p edic he isk
o o e weigh (1.39- old inc eased odds) and obesi y (1.41- old inc eased odds) using 30 BMI- ela ed SNPs, s a -
ing ha hei GRS was poo in p edic ing sho - e m longi udinal changes in BMI. In wo Finnish child en and
adolescen coho s, Seyednas ollah e al.21 de eloped wo wGRS o 97 and 19 SNPs p e iously ela ed o he isk
o obesi y, and ob ained a ligh ly be e p edic ion accu acy wi h he 19-SNP GRS han in ou s udy (AUC = 0.769
s. 0.734). Howe e , none o he SNPs used in he wo GRSs in Eu opean adolescen s abo e men ioned concu
wi h he SNPs u ilised o de elop ou GRS.
The majo i y o he GRSs de eloped in pedia ic and adolescen popula ions ha e been pe o med in non-
Eu opean subjec s, based on SNPs associa ed wi h obesi y isk om p e ious GWAS. Compa a i ely o he
Eu opean GRSs, he non-Eu opean s udies p o ided a ewe numbe o selec ed SNPs ela ed o obesi y isk o
de elop hei GRS. In a c oss-sec ional s udy o B azilian child en and adolescen s (mean age 11.9 ± 2.8yea s)19,
he BMI-speci ic wGRS composed o 3 SNPs was associa ed wi h a 2.65- old inc eased isk o o e weigh and
obesi y. In a Chinese coho o child en aged 7.3 o 11.1yea s, Fang J e al.35 de eloped an uGRS o 11 BMI- ela ed
SNPs which explained 0.11kg/m2 BMI inc ease in hose child en ca ying BMI suscep ibili y alleles. L D e al.36
Figu e1. Fo es plo o single nucleo ide polymo phisms (SNPs) nega i ely (OR < 1) and posi i ely (OR > 1)
associa ed wi h isk o OW/OB. Legend: SNPs o de ed by ch omosome numbe . P o ec i e SNPs agains isk o
o e weigh /obesi y a e shown in he uppe pa o he o es plo ; SNPs wi h p edisposi ion isk o o e weigh /
obesi y a e shown in he bo om pa . Mul i a ia e model Odds Ra io (O.R.) and 95% con idence in e als (C.I.)
displayed.
5
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s/
posi i ely associa ed he cumula i e e ec o 5 BMI- ela ed SNPs GRS o obesi y isk by mo e han se en old
inc eased odds in indi iduals ca ying 5–7 isk alleles (age 11.6 ± 2.5, N = 2977). Finally, SNPs p e iously ela ed
o o he ca diome abolic isk ac o s (hype ension) in a Chinese adolescen popula ion (aged 12.2 ± 3.0yea s)
we e used o de elop a 3 SNP GRS posi i ely associa ed o obesi y isk37. The inc eased isk o obesi y associ-
a ed o he men ioned GRSs in non-Eu opean subjec s seem o ha e a consis ency o some SNPs showed in he
p esen s udy, despi e acknowledging ha he subjec s o igin does no allow o make compa isons in his ega d.
The elabo a ion o a BMI-GRS comp ised p o ec o and isk in he same model. Some o hose SNPs ha e
been signi ican ly associa ed wi h he isk o obesi y in p e ious s udies. The A allele o FTO s9939609 polymo -
phism has been consis en ly associa ed wi h highe BMI and wais ci cum e ence in se e al s udies in adul s11,
adolescen s9 and child en38. In coho s o child en and adul s wi h Eu opean ances y, F ayling e al.11 and Wille
e al.39 ound he s onges associa ions o he A isk allele o FTO s9939609 wi h BMI. These indings con i m
he ole o FTO s9939609 in ou GRS as isk ac o o OW/OB. In a s udy by Boko e al.40, CD36 s3211867
inc eased he isk o obesi y by almos wo olds in a coho o Hunga ian obese (N = 307) and no mal weigh
(N = 339) adolescen s. Al hough he s udy had wo independen samples wi h limi ed sample size, he indings
a e consis en wi h he esul s o ou s udy. Mo eo e , in a pooled-s udy by Solaas e al.41, au ho s obse ed
signi ican associa ion be ween LXRβ s17373080 and he isk o T2DM and OW/OB by 1.59- old inc eased
odds. Equally, in ou s udy, LXRβ s17373080 SNP was associa ed wi h a 1.38 old highe isk o OW/OB. O
no e, he las wo s udies included pa icipan s om he HELENA coho . On he o he hand, he p esen s udy
showed ha THRA s1568400 SNP was nega i ely associa ed wi h he isk o OW/OB whe eas he same SNP
was associa ed wi h highe BMI in a coho o Spanish adul s42.
O he SNPs included in he GRS de eloped in he p esen s udy ha e also been associa ed wi h obesi y- ela ed
ca diome abolic isk ac o s in e hnically di e se adul s, bu no wi h o e weigh o obesi y isk. Thus, in a Eu o-
pean popula ion, DRD2 s180049743 modi ied he ela ionship be ween bi h weigh and adul hood educa ional
a ainmen in Finnish subjec s and FASN s424644444 a enua ed he e ec on low densi y lipop o eins (LDL)
peak pa icle diame e when consuming a high amoun o a in a Canadian coho . Wi hin non-Eu opean back-
g ound, in Chinese popula ion, NR3C1 s770144345 was signi ican ly associa ed wi h a highe isk o me abolic
synd ome and CC alleles o IL-6 s152410746 had a highe isk o de eloping neph opa hy in T2DM subjec s. In
addi ion, PPARG s413527547 was posi i ely associa ed wi h glyca ed hemoglobin and as ing plasma glucose in
Taiwanese men al heal h pa ien s. In p egnan Tu kish women, LPA s935529648 was posi i ely ela ed o ascula
Table 1. Main cha ac e is ics o he 21 single nucleo ide polymo phisms (SNPs) included in he gene ic
isk sco e. Associa ion o SNPs in ela ion o body mass index (BMI) displayed in p alues (p). AMPD1
(Adenosine Monophospha e Deaminase 1); ANGPTL4 (Angiopoie in like p o ein 4); BMI (Boy Mass Index);
CD36 (clus e o di e en ia ion 36); CETP (Choles e yl es e ans e p o ein); CNTF (Cilia y Neu o ophic
Fac o ); CNTFR (Cilia y Neu o ophic Fac o Recep o ); DRD2 (Dopamine Recep o D2); FASN (Fa y acid
syn hase); FTO (Fa mass and obesi y-associa ed gene); HWE (Ha dy–Weinbe g Equilib ium); IGF1 (Insulin
Like G ow h Fac o 1); IL-6 (In e leukin-6); LPA (Lipop o ein A); LXRB (Li e X ecep o be a); MAF (Mino
allele equency); NOS2A (Ni ic oxide syn hase 2); NR3C1 (Nuclea ecep o sub amily 3, g oup C, membe
1); PPARG (Pe oxisome P oli e a o Ac i a ed Recep o Gamma); PTPN1 (P o ein Ty osine Phospha ase Non-
Recep o Type 1); THRA (Thy oid Ho mone Recep o Alpha).
s code Nea es gene Alleles (majo /mino ) MAF pGeno yping success a e HWE
s2010899 AMPD1 A/C 0.44 0.012 99.8 0.349
s4135275 PPARG A/G 0.19 0.024 99.9 0.655
s4912905 NR3C1 C/G 0.23 0.004 100.0 0.665
s7701443 NR3C1 A/G 0.40 1.35 × 10–4 100.0 0.127
s13182800 NR3C1 C/A 0.24 5.96 × 10–4 99.9 0.183
s9355296 LPA G/A 0.13 0.013 100.0 0.216
s1524107 IL-6 G/A 0.07 0.006 100.0 0.879
s3211867 CD36 C/A 0.07 0.033 100.0 0.728
s2183013 CNTFR C/G 0.17 0.001 100.0 0.934
s2515362 CNTF A/G 0.44 0.024 99.9 0.184
s1800497 DRD2 G/A 0.18 0.049 99.8 0.763
s1019731 IGF1 C/A 0.11 0.005 100.0 0.744
s9939609 FTO T/A 0.40 5.81 × 10–4 100.0 0.322
s4783961 CETP A/G 0.50 0.014 100.0 0.558
s8068149 NOS2A G/A 0.46 0.010 100.0 0.136
s7502966 THRA A/G 0.44 0.025 99.7 0.264
s1568400 THRA A/G 0.26 0.023 100.0 0.349
s4246444 FASN C/A 0.27 0.008 94.7 0.461
s1044250 ANGPTL4 G/A 0.29 0.005 99.6 0.051
s17373080 LXRβ G/C 0.32 0.005 99.7 0.523
s2143511 PTPN1 A/G 0.43 0.004 99.9 0.337

6
Vol:.(1234567890)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s/
Table 2. Associa ion o he 21 single nucleo ide polymo phisms (SNPs) o he gene ic isk sco e wi h OW/OB.
SNPs o de ed by ch omosome numbe and p o ec o / isk na u e o he SNP. *Uni a ia e (indi idual model
o SNP-O e weigh /Obesi y associa ion) and mul i a ia e model (mul iple model SNP-O e weigh /Obesi y
associa ion) displayed wi h Odds Ra io (O.R.) and 95% con idence in e al (C.I.).
s code Ch omosome
Uni a ia e* Mul i a ia e*
O.R. (95% C.I.) pO.R. (95% C.I.) p
s2010899 1 0.822 (0.665–1.014) 0.068 0.749 (0.596–0.939) 0.012
s4135275 3 0.738 (0.556–0.967) 0.031 0.709 (0.522–0.951) 0.012
s4912905 5 0.810 (0.630–1.034) 0.096 0.639 (0.467–0.867) 0.024
s9355296 6 0.636 (0.448–0.886) 0.009 0.629 (0.430–0.901) < 0.001
s1524107 7 0.602 (0.363–0.950) 0.038 0.481 (0.278–0.795) 0.014
s2183013 9 0.674 (0.495–0.903) 0.010 0.595 (0.425–0.818) 0.033
s1019731 12 0.681 (0.468–0.967) 0.038 0.571 (0.379–0.837) 0.049
s1568400 17 0.794 (0.623–1.006) 0.059 0.742 (0.571–0.957) 0.023
s4246444 17 0.756 (0.592–0.957) 0.022 0.711 (0.548–0.914) 0.009
s7701443 5 1.170 (0.946–1.447) 0.146 1.919 (1.432–2.579) 0.005
s13182800 5 1.374 (1.094–1.721) 0.006 1.771 (1.340–2.344) < 0.001
s3211867 7 1.487 (1.009–2.159) 0.040 1.576 (1.029–2.384) 0.006
s2515362 11 1.189 (0.973–1.452) 0.090 1.283 (1.032–1.594) 0.002
s1800497 11 1.204 (0.924–1.559) 0.162 1.329 (0.998–1.761) 0.024
s9939609 16 1.496 (1.219–1.838) < 0.001 1.575 (1.263–1.968) 0.005
s4783961 16 1.272 (1.034–1.568) 0.023 1.325 (1.058–1.663) 0.014
s8068149 17 1.190 (0.974–1.456) 0.088 1.331 (1.071–1.656) 0.010
s7502966 17 1.165 (0.952–1.427) 0.138 1.287 (1.031–1.607) 0.025
s1044250 19 1.211 (0.976–1.500) 0.079 1.390 (1.100–1.754) 0.006
s17373080 19 1.313 (1.061–1.624) 0.012 1.388 (1.102–1.748) 0.005
s2143511 20 1.294 (1.056–1.588) 0.013 1.386 (1.108–1.736) 0.004
Figu e2. Recei e ope a ing cha ac e is ics (ROC) cu es o he wo gene ic isk sco es, unweigh ed (uGRS)
and weigh ed (wGRS). A eas unde cu es (AUC) a e indica ed. The s aigh line ep esen s he ROC expec ed
by chance only.
7
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s/
in lamma ion as u u e ca dio ascula e en indica o . Finally, in a la ge adul coho o 5 di e en ances y
g oups, CETP s478396149 was in ol ed in sleep-associa ed ad e se high densi y lipop o eins (HDL) p o iles.
In con as , as a as we know, se e al SNPs included ou GRSs (i.e., AMPD1 s2010899, NR3C1 s4912905,
CNTFR s2183013, IGF1 s1019731 as p o ec o ac o s and NR3C1 s13182800, CNTF s2515362, NOS2A
s8068149, THRA s7502966, ANGPTL4 s1044250 and PTPN1 s2143511 as isk ac o s) a e new p edic i e
ac o s, as hey had no p e iously been associa ed wi h obesi y o obesi y ela ed diseases no had been signi i-
can ly ele an in p e ious s udies.
Addi ionally, he p esen GRS was posi i ely es ed o e alua e i s abili y o p edic he isk o o e weigh and
obesi y in o he adiposi y es ima es (FMI). P e ious s udies ha e also iden i ied po en ial in e ac ions be ween
an obesi y-GRS and die on FMI in English child en (9y s)50. Monne eau e al.51 cons uc ed 15 SNPs-wGRS
ela ed o child BMI in child en (6y s) om Ne he lands signi ican ly associa ed o o al a mass. Mo e so,
ano he s udy52 showed a BMI-based GRS signi ican ly associa ed o highe body a mass in Finnish child en
and adolescen s.
Al hough using he ex e nal weigh om me a-analyses is he gold s anda d o build a GRS, when he ex e nal
weigh s a e no a ailable, he uGRS is commonly used35,53. In he p esen app oach, in e nal weigh s om he
gene ic e ec s o he same s udy we e used. The wGRS ou pe o med he uGRS in e ms o s a is ical powe (0.734
s. 0.723). Con en ionally, i is accep ed ha he AUC in a ROC analysis should be > 0.8 o be o clinical alue
o sc eening54. When cons uc ing he GRS models, AUC ell sho o his h eshold combining gene ic ac o s
alone. As SNPs hemsel es ha e li le p edic o capaci y, we should conside he esul s ob ained o cons uc
he uGRS and wGRS as s a is ically accep able, so ou wGRS could be eplica ed in o he coho s wi h simila
cha ac e is ics. Thus, ou indings add a signi ican con ibu ion o obesi y-speci ic GRS ha may imp o e he
p edic i e alues o obesi y bioma ke s in adolescen s.
Figu e3. Boxplo o he dis ibu ion o unweigh ed gene ic isk sco e (uGRS) and weigh ed gene ic isk sco e
(wGRS). Legend: Values o he g oups o Non-o e weigh (NON-OW) and O e weigh /Obesi y (OW/OB).
8
Vol:.(1234567890)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s/
O he au ho s55 sugges ha adi ional p edic o s, such as amily his o y and childhood obesi y ha e s onge
p edic i e powe han models based on he es ablished gene ic a ian s. None heless, he limi ed p edic i e abili y
o gene ic a ian s does no unde alue he ole o gene disco e y o obesi y as, based on he li e a u e, gene ic
analyses ha e al eady p o ided wi h p omising insigh s in ol ing BMI egula ion6. As such, he p esen GRSs, o
u u e GRS comp ising addi ional SNPs om o he genes which do no ha e (a p io i) p e ious epo ed associa-
ions wi h obesi y, could yield p omising esul s o minimise he isk o ca dio ascula e en s ela ed o obesi y.
Howe e , he p esen s udy has some limi a ions. The esul s should be alida ed in la ge pedia ic s udy
popula ions, also using obesi y incidence, in o de o es he eliabili y o his obesi y-speci ic GRS in o he popu-
la ions wi h simila e hnici y. Despi e ha some s udies on child en epo ing he p edic ion o adul hood obesi y
e icien ly 20 o 30yea s la e 21, di e en gene ic ac o s migh a ec he sho - e m changes in BMI, especially
du ing pe iods o apid g ow h56. Due o he c oss-sec ional na u e o he s udy, no cause-e ec ela ionship can
be de e mined. Addi ionally, al hough he model o de elop he GRSs was in e nally alida ed pe o ming en old
c oss alida ion analysis, we unde s and ha he op imal si ua ion would ha e been an ex e nal alida ion in an
independen coho . Mo eo e , only selec ed isk loci a e a ailable in he HELENA s udy. Since he es ablished
common a ian s om GWAS explain a small p opo ion o he BMI a ia ion6, i is likely ha o he loci om
a e a ian s, s ill o be disco e ed, will eme ge when la ge sample sizes a e included in GWAS. Fu he mo e,
he e is no da a a ailable ega ding he ela edness o he e hnic o igins among he s udied pa icipan s; he allele
equencies and hei e ec size migh be di e en om non-Eu opean popula ions and he ou come should no
be ep oduced o o he e hnici ies. Since he HELENA s udy selec ed genes based on candida e genes ins ead
o GWAS, he o e lapping e ec obse ed be ween SNPs o Eu opean and non-Eu opean adolescen s could be
possible. Mo e common SNPs in non-Eu opean GRSs we e ound han in Eu opean GRSs. This inding could be
due o he highe numbe o GRSs de eloped in o he e hnici ies in compa ison o he numbe o GRS pe o med
in Eu opean popula ion. Also, we used he same da a in he SNPs selec ion p ocess and in he building model,
hus li le bias can be p oduced. The e o e, he esul s showed in he p esen s udy should be conside ed ca e ully.
Fu he s udies wi h la ge sample size could p o ide key in o ma ion o his po en ial gene ic p edisposi ion
o obesi y. On he o he hand, he p esen s udy has also some s eng hs. The mul icen ic design o HELENA
s udy in ol ed he pa icipa ion o adolescen s om 10 Eu opean ci ies. This allowed he esea che s o use a
la ge da abase wi h ele an and di e se in o ma ion om di e en popula ions ac oss Eu ope. Addi ionally, only
ew GRSs o p edic he o e weigh and obesi y isk ha e been de eloped pa icula ly in Eu opean adolescen s,
an unde s udied popula ion om he ea ly ea men and p e en ion pe spec i e21,22. Simila ly o Viljakainen
e al.22, he p oposed gene ic sco e o p edisposi ion o obesi y de ined in his s udy migh e icien ly con ibu e
o disce n popula ion a isk o o e weigh and obesi y and no jus obesi y alone.
In conclusion, ou indings sugges ha he GRSs de eloped in he p esen s udy (uGRS and wGRS) could
be conside ed as a use ul gene ic ool o e alua e indi idual’s p edisposi ion o OW/OB, allowing o ad ance in
he p e en ion and managemen o he disease om ea ly s ages in li e. Fu u e GRS de elopmen wi h la ge
samples will be able o de ec new a ian s p e iously no ela ed o obesi y ha could in luence he gene ic isk
o obesi y, o he han he common ones.
Table 3. Speci ici y, sensi i i y, nega i e and posi i e p edic i e alue and accu acy analysis p esen ed in
pe cen ages (%). Abb e ia ions: uGRS (Unweigh ed Gene ic Risk Sco e); wGRS (Weigh ed Gene ic Risk
Sco e); Acc (accu acy); NPV (nega i e p edic i e alue); PPV (posi i e p edic i e alue); Sens (sensi i i y);
Spec (speci ici y).
Cu -o Spec Sens NPV PPV Acc Cu -o Spec Sens NPV PPV Acc
uGRS wGRS
30 100 0.42 77.99 100 78.01 4.14 100 0.42 77.99 100 78.01
29 99.88 3.81 78.56 90.00 78.67 3.51 99.88 5.50 78.86 92.85 79.04
28 99.15 6.35 78.89 68.18 78.67 3.30 99.15 8.89 79.34 75.00 79.23
27 96.39 13.55 79.74 51.61 78.11 2.98 96.63 15.25 80.09 56.25 78.67
26 92.07 26.69 81.59 48.83 77.64 2.61 92.07 27.54 81.76 49.61 77.82
25 83.31 44.91 84.22 43.26 74.83 2.26 83.31 48.72 85.15 45.27 75.67
24 69.38 63.55 87.04 37.03 68.10 1.88 69.38 65.25 87.57 37.65 68.47
23 54.74 73.72 88.03 31.57 58.93 1.50 54.74 79.23 90.29 33.15 60.14
22 39.13 89.40 92.87 29.38 50.23 1.10 39.13 89.40 92.87 29.38 50.23
21 26.17 93.22 93.16 26.34 40.97 0.77 26.17 93.22 93.16 26.34 40.97
20 14.04 97.88 95.90 24.39 32.55 0.35 14.04 97.03 94.35 24.23 32.36
19 7.92 98.72 95.65 23.30 27.97 0.03 7.92 98.72 95.65 23.30 27.97
18 2.88 99.57 96.00 22.50 24.22 -0.54 2.88 99.57 96.00 22.50 24.22
17 1.32 100 100 22.30 23.10 -0.82 1.32 100 100 22.30 23.10
16 0.48 100 100 22.15 22.45 -1.20 0.48 100 100 22.15 22.45
15 0.00 100 100 22.09 22.17 -1.67 0.12 100 100 22.09 22.17
9
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:3067 | h ps://doi.o g/10.1038/s41598-021-82712-4
www.na u e.com/scien i ic epo s/
Recei ed: 21 July 2020; Accep ed: 22 Janua y 2021
Re e ences
1. WHO | Childhood o e weigh and obesi y. WHO. h ps ://www.o g/en i y/die p hysic alac i i y /child hood/en/index .h ml (2017).
2. Gu nani, M., Bi ken, C. & Hamil on, J. Childhood obesi y: causes, consequences, and managemen . Pedia Clin No h Am 62,
821–840. h ps ://doi.o g/10.1016/j.pcl.2015.04.001 (2015).
3. Spelio es, E. K. e al. Associa ion analyses o 249,796 indi iduals e eal 18 new loci associa ed wi h body mass index. Na Gene
42, 937–948. h ps ://doi.o g/10.1038/ng.686 (2010).
4. B ad ield, J. P. e al. A genome-wide associa ion me a-analysis iden i ies new childhood obesi y loci. Na .Gene 44, 526–531. h ps
://doi.o g/10.1038/ng.2247 (2012).
5. Mey e, D. e al. Genome-wide associa ion s udy o ea ly-onse and mo bid adul obesi y iden i ies h ee new isk loci in Eu opean
popula ions. Na Gene 41, 157–159. h ps ://doi.o g/10.1038/ng.301 (2009).
6. Locke, A. E. e al. Gene ic s udies o body mass index yield new insigh s o obesi y biology. Na u e 518, 197–206. h ps ://doi.
o g/10.1038/na u e1417 7 (2015).
7. Loos, R. J. & Yeo, G. S. The bigge pic u e o FTO: he i s GWAS-iden i ied obesi y gene. Na . Re . Endoc inol. 10, 51–61. h ps
://doi.o g/10.1038/n end o.2013.227 (2014).
8. Kilpeläinen, T. O. e al. Physical ac i i y a enua es he in luence o FTO a ian s on obesi y isk: a me a-analysis o 218,166 adul s
and 19,268 child en. PLoS Med. h ps ://doi.o g/10.1371/jou n al.pmed.10011 16 (2011).
9. Labayen, I. e al. Die a y a in ake modi ies he in luence o he FTO s9939609 polymo phism on adiposi y in adolescen s: he
HELENA c oss-sec ional s udy. Nu . Me ab. Ca dio asc. Dis. 26, 937–943. h ps ://doi.o g/10.1016/j.numec d.2016.07.010 (2016).
10. Ruiz, J. R. e al. A enua ion o he e ec o he FTO s9939609 Polymo phism on o al and cen al body a by physical ac i i y in
adolescen s: he HELENA s udy. A ch. Pedia ics Adolesc. Med. h ps ://doi.o g/10.1001/a chp edia ics.2010.29 (2010).
11. F ayling, T. M. e al. A common a ian in he FTO gene is associa ed wi h body mass index and p edisposes o childhood and
adul obesi y. Science 316, 889–894. h ps ://doi.o g/10.1126/scien ce.11416 34 (2007).
12. Wang, S., He, S., Yuan, F. & Zhu, X. Tagging SNP-se selec ion wi h maximum in o ma ion based on linkage disequilib ium
s uc u e in genome-wide associa ion s udies. Bioin o ma ics 33(14), 2078–2081. h ps ://doi.o g/10.1093/bioin o ma ics/b x15 1
(2017).
13. Ka hi esan, S. e al. Polymo phisms associa ed wi h choles e ol and isk o ca dio ascula e en s. N Engl J Med 358, 1240–1249.
h ps ://doi.o g/10.1056/NEJMo a0706 728 (2008).
14. Janssens, A. C. e al. P edic i e es ing o complex diseases using mul iple genes: ac o ic ion?. Gene Med 8, 395–400. h ps ://
doi.o g/10.1097/01.gim.00002 29689 .18263 . 4 (2006).
15. Mo ison, A. C. e al. P edic ion o co ona y hea disease isk using a gene ic isk sco e: he A he oscle osis Risk in Communi ies
S udy. Am J Epidemiol 166, 28–35. h ps ://doi.o g/10.1093/aje/kwm06 0 (2007).
16. Co ella, D. e al. Polymo phism o he T ansc ip ion Fac o 7-Like 2 Gene (TCF7L2) in e ac s wi h obesi y on Type-2 diabe es in
he PREDIMED s udy emphasizing he he e ogenei y o gene ic a ian s in ype-2 diabe es isk p edic ion: ime o obesi y-speci ic
gene ic isk sco es. Nu ien s h ps ://doi.o g/10.3390/nu812 0793 (2016).
17. Belsky, D. W. e al. De elopmen and e alua ion o a gene ic isk sco e o obesi y. Biodemog . Soc. Biol. 59, 85–100. h ps ://doi.
o g/10.1080/19485 565.2013.77462 8 (2013).
18. Goumidi, L. e al. E ec s o es ablished BMI-associa ed loci on obesi y- ela ed ai s in a F ench ep esen a i e popula ion sample.
BMC Gene . h ps ://doi.o g/10.1186/1471-2156-15-62 (2014).
19. Todendi, P. F. e al. Gene ic isk sco e based on a mass and obesi y-associa ed, ansmemb ane p o ein 18 and ib onec in ype III
domain con aining 5 polymo phisms is associa ed wi h an h opome ic cha ac e is ics in Sou h B azilian child en and adolescen s.
B . J. Nu . 121, 93–99. h ps ://doi.o g/10.1017/s0007 11451 80027 38 (2019).
20. Zhao, H., Wilkinson, A., Shen, J., Wu, X. & Chow, W. H. Gene ic polymo phisms in genes ela ed o isk- aking beha iou s p e-
dic ing body mass index ajec o y among Mexican Ame ican adolescen s. Pedia . Obes. 12, 356–362. h ps ://doi.o g/10.1111/
ijpo.12151 (2017).
21. Seyednas ollah, F. e al. P edic ion o adul hood obesi y using gene ic and childhood clinical isk ac o s in he ca dio ascula isk
in young Finns s udy. Ci c. Ca dio asc. Gene . h ps ://doi.o g/10.1161/ci cg ene i cs.116.00155 4 (2017).
22. Viljakainen, H. e al. Gene ic isk sco e p edic s isk o o e weigh and obesi y in Finnish p eadolescen s. Clin. Obes. 9, e12342.
h ps ://doi.o g/10.1111/cob.12342 (2019).
23. Mo eno, L. A. e al. Nu i ion and li es yle in eu opean adolescen s: he HELENA (Heal hy Li es yle in Eu ope by Nu i ion in
Adolescence) s udy. Ad . Nu . 5, 615s–623s. h ps ://doi.o g/10.3945/an.113.00567 8 (2014).
24. Mo eno, L. A. e al. Design and implemen a ion o he Heal hy Li es yle in Eu ope by Nu i ion in Adolescence C oss-Sec ional
S udy. In . J. Obes. (Lond.) 32(Suppl 5), S4-11. h ps ://doi.o g/10.1038/ijo.2008.177 (2008).
25. Beghin, L. e al. Quali y assu ance o e hical issues and egula o y aspec s ela ing o good clinical p ac ices in he HELENA C oss-
Sec ional S udy. In . J. Obes. (Lond.) 32(Suppl 5), S12-18. h ps ://doi.o g/10.1038/ijo.2008.179 (2008).
26. Nagy, E. e al. Ha moniza ion p ocess and eliabili y assessmen o an h opome ic measu emen s in a mul icen e s udy in ado-
lescen s. In . J. Obes. (Lond.) 32(Suppl 5), S58-65. h ps ://doi.o g/10.1038/ijo.2008.184 (2008).
27. Cole, T. J., Bellizzi, M. C., Flegal, K. M. & Die z, W. H. Es ablishing a s anda d de ini ion o child o e weigh and obesi y wo ldwide:
in e na ional su ey. BMJ 320, 1240–1243. h ps ://doi.o g/10.1136/bmj.320.7244.1240 (2000).
28. Cole, T. J. & Lobs ein, T. Ex ended in e na ional (IOTF) body mass index cu -o s o hinness, o e weigh and obesi y. Pedia .
Obes. 7, 284–294. h ps ://doi.o g/10.1111/j.2047-6310.2012.00064 .x (2012).
29. Gonzalez-G oss, M. e al. Sampling and p ocessing o esh blood samples wi hin a Eu opean mul icen e nu i ional s udy:
e alua ion o bioma ke s abili y du ing anspo and s o age. In . J. Obes. (Lond.) 32(Suppl 5), S66–S75. h ps ://doi.o g/10.1038/
ijo.2008.185 (2008).
30. Goumidi, L. e al. Heal hy li es yle by nu i ion in adolescence (HELENA). A new EU unded p ojec . The apie h ps ://doi.
o g/10.2515/ he a pie:20070 50 (2007).
31. Hanley, J. A. & McNeil, B. J. The meaning and use o he a ea unde a ecei e ope a ing cha ac e is ic (ROC) cu e. Radiology
143, 29–36. h ps ://doi.o g/10.1148/ adio logy.143.1.70637 47 (1982).
32. Ca ayol, J., To es, F., Konig, I. R., Hage , J. & Ziegle , A. E alua ing diagnos ic accu acy o gene ic p o iles in a ec ed o sp ing
amilies. S a . Med. 29, 2359–2368. h ps ://doi.o g/10.1002/sim.4006 (2010).
33. Liu, X. Classi ica ion accu acy and cu poin selec ion. S a . Med. 31, 2676–2686. h ps ://doi.o g/10.1002/sim.4509 (2012).
34. Kundu, S., Aulchenko, Y. S., an Duijn, C. M. & Janssens, A. C. P edic ABEL: an R package o he assessmen o isk p edic ion
models. Eu . J. Epidemiol. 26, 261–264. h ps ://doi.o g/10.1007/s1065 4-011-9567-4 (2011).
35. Fang, J. e al. Polygenic isk, adhe ence o a heal hy li es yle, and childhood obesi y. Pedia . Obes. 14, e12489. h ps ://doi.
o g/10.1111/ijpo.12489 (2019).
36. L , D. e al. Gene ic a ia ions in SEC16B, MC4R, MAP2K5 and KCTD15 we e associa ed wi h childhood obesi y and in e ac ed
wi h die a y beha io s in Chinese school-age popula ion. Gene 560, 149–155. h ps ://doi.o g/10.1016/j.gene.2015.01.054 (2015).