RESEARCH Open Access
DNA me hyla ion and body mass index
om bi h o adolescence: me a-analyses o
epigenome-wide associa ion s udies
Flo ianne O. L. Vehmeije
1,2,3†
, Leanne K. Küpe s
4,5,6†
, Gemma C. Sha p
4
, Lucas A. Salas
7,8,9,10
, Saman ha Len
11
,
De eje D. Jima
12,13
, Gwen Tindula
14
, Sa ah Reese
15
, Cancan Qi
16,17
, Olena G uzie a
18,19
, Ch is ian Page
20,21
,
Faisal I. Rezwan
22,23
, Philip E. Mel on
24,25
, Ellen Noh
26,27
, Geò gia Esca amís
10,28,29
, Pe e Rzehak
30
, Anni Heiskala
31
,
Tong Gong
32
, Samuli T. Tuominen
33
, Lu Gao
34
, Jason P. Ross
35
, Anne P. S a ling
36,37
, John W. Holloway
23,38
,
Paul Youse i
4
, Gunn Ma i Aas ang
39
, Law ence J. Beilin
40
, Anna Be gs öm
18,19
, Elisabe h Binde
41,42
, Leda Cha zi
43
,
E a Co peleijn
6
, Da ina Czama a
41
, B enda Eskenazi
44
, Susan Ewa
45
, Na alia Fe e
46
, Vei G o e
30
,
Da iusz G usz eld
47
, Si i E. Håbe g
20
, Ca h ine Hoyo
13,48
, Ka en Huen
14
, Robe Ka lsson
32
, Inge Kull
49,50
,
Jean-Paul Langhend ies
51
, Johanna Lepeule
52
, Ma ia C. Magnus
4,5,20
, Rachel L. Magui e
48,53
, Pe e L. Molloy
35
,
Clai e Monne eau
1,3
, T e o A. Mo i
40
, Emily Oken
54
, Ka i Räikkönen
33
, She yl Ri as-Shiman
54
,
Ca los Ruiz-A enas
8,9,10
, Syl ain Sebe
31
, Vilhelmina Ullema
32
, El i a Ve duci
55
, Judi h M. Vonk
6,17
,
Cheng-jian Xu
16,17,56,57
, I ana V. Yang
36,58,59
, Hongmei Zhang
60
, Weiming Zhang
61
, Wil ied Ka maus
60
,
Dana Dabelea
36,37,62
, Be e ly S. Muhlhausle
63
, Ca ie V. B e on
34
, Ja i Lah i
33,64
, Ca a ina Almq is
32,65
,
Ma jo-Rii a Ja elin
31,66,67,68
, Be hold Kole zko
30
, Ma ine V ijheid
8,9,10
, Tho kild I. A. Sø ensen
4,69
, Rae-Chi Huang
70
,
Syed Hasan A shad
38,71
, Wenche Nys ad
72
, E ik Melén
49,50
, Ge a d H. Koppelman
16,17
, S ephanie J. London
15
,
Nina Holland
14
, Ma iona Bus aman e
8,9,10
, Susan K. Mu phy
53
, Ma ie-F ance Hi e
54,73,74
, And ea Bacca elli
75†
,
Ca oline L. Rel on
4†
, Ha old Sniede
6†
, Vincen W. V. Jaddoe
1,2,3†
and Janine F. Felix
1,2*†
Abs ac
Backg ound: DNA me hyla ion has been shown o be associa ed wi h adiposi y in adul hood. Howe e , whe he simila
DNA me hyla ion pa e ns a e associa ed wi h childhood and adolescen body mass index (BMI) is la gely unknown. Mo e
insigh in o his ela ionship a younge ages may ha e implica ions o u u e p e en ion o obesi y and i s ela ed ai s.
(Con inued on nex page)
© The Au ho (s). 2020 Open Access This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License,
which pe mi s use, sha ing, adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long as you gi e
app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons licence, and indica e i
changes we e made. The images o o he hi d pa y ma e ial in his a icle a e included in he a icle's C ea i e Commons
licence, unless indica ed o he wise in a c edi line o he ma e ial. I ma e ial is no included in he a icle's C ea i e Commons
licence and you in ended use is no pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will need o ob ain
pe mission di ec ly om he copy igh holde . To iew a copy o his licence, isi h p://c ea i ecommons.o g/licenses/by/4.0/.
The C ea i e Commons Public Domain Dedica ion wai e (h p://c ea i ecommons.o g/publicdomain/ze o/1.0/) applies o he
da a made a ailable in his a icle, unless o he wise s a ed in a c edi line o he da a.
* Co espondence: [email p o ec ed]
†
Flo ianne O.L. Vehmeije and Leanne K. Küpe s a e au ho s who con ibu ed
equally o he wo k.
†
And ea Bacca elli, Ca oline L. Rel on, Ha old Sniede , Vincen W.V. Jaddoe,
and Janine F. Felix a e au ho s join ly supe ised he wo k.
1
The Gene a ion R S udy G oup, E asmus MC, Uni e si y Medical Cen e
Ro e dam, Room Na-2918, E asmus MC, PO Box 2040, 3000 CA Ro e dam,
he Ne he lands
2
Depa men o Pedia ics, E asmus MC, Uni e si y Medical Cen e Ro e dam,
Ro e dam, he Ne he lands
Full lis o au ho in o ma ion is a ailable a he end o he a icle
Vehmeije e al. Genome Medicine (2020) 12:105
h ps://doi.o g/10.1186/s13073-020-00810-w
(Con inued om p e ious page)
Me hods: We examined whe he DNA me hyla ion in co d blood and whole blood in childhood and adolescence was
associa ed wi h BMI in he age ange om 2 o 18 yea s using bo h c oss-sec ional and longi udinal models. We pe o med
me a-analyses o epigenome-wide associa ion s udies including up o 4133 child en om 23 s udies. We examined he
o e lap o indings epo ed in p e ious s udies in child en and adul s wi h hose in ou analyses and calcula ed en ichmen .
Resul s: DNA me hyla ion a h ee CpGs (cg05937453, cg25212453, and cg10040131), each in a di e en age ange, was
associa ed wi h BMI a Bon e oni signi icance, P<1.06×10
−7
, wi h a 0.96 s anda d de ia ion sco e (SDS) (s anda d e o
(SE) 0.17), 0.32 SDS (SE 0.06), and 0.32 BMI SDS (SE 0.06) highe BMI pe 10% inc ease in me hyla ion, espec i ely. DNA
me hyla ion a nine addi ional CpGs in he c oss-sec ional childhood model was associa ed wi h BMI a alse disco e y
a e signi icance. The s eng h o he associa ions o DNA me hyla ion a he 187 CpGs p e iously iden i ied o be
associa ed wi h adul BMI, inc eased wi h ad ancing age ac oss childhood and adolescence in ou analyses. In addi ion,
co ela ion coe icien s be ween e ec es ima es o hose CpGs in adul s and in child en and adolescen s also inc eased.
Among he op indings o each age ange, we obse ed inc easing en ichmen o he CpGs ha we e p e iously
iden i ied in adul s (bi h P
en ichmen
= 1; childhood P
en ichmen
=2.00×10
−4
;adolescenceP
en ichmen
=2.10×10
−7
).
Conclusions: The e we e only minimal associa ions o DNA me hyla ion wi h childhood and adolescen BMI. Wi h he
ad ancing age o he pa icipan s ac oss childhood and adolescence, we obse ed inc easing o e lap wi h al e ed DNA
me hyla ion loci epo ed in associa ion wi h adul BMI. These indings may be compa ible wi h he hypo hesis ha DNA
me hyla ion di e ences a e mos ly a consequence a he han a cause o obesi y.
Keywo ds: Body mass index, Childhood obesi y, DNA me hyla ion, Epigene ics
Backg ound
An accumula ing body o e idence sugges s ha exposu es
in ea ly li e a e associa ed wi h childhood BMI [1]. I is hy-
po hesized ha changes in DNA me hyla ion may unde lie
he associa ions o ea ly-li e exposu es wi h childhood adi-
posi y [2–4]. Thus a , mos o he e idence ega ding
DNA me hyla ion and adiposi y s ems om adul s udies
[5–9]. The la ges epigenome-wide associa ion s udy
(EWAS) in adul s iden i ied c oss-sec ional associa ions
be ween DNA me hyla ion a 187 loci and BMI in o e 10,
000 pa icipan s [5]. P e ious s udies o he associa ions
be ween epigenome-wide DNA me hyla ion and childhood
and adolescen adiposi y we e small and inconclusi e [10–
16]. Candida e gene s udies in childhood iden i ied associa-
ions o DNA me hyla ion in co d and childhood blood
wi h measu es o adiposi y [17–24]. Epigenome-wide asso-
cia ion s udies in child en and adolescen s, wi h sample
sizes anging om 40 o 700 indi iduals, iden i ied a limi ed
numbe o cy osine-phospha e-guanine si es (CpGs) associ-
a ed wi h BMI [11–13,15,25]. Al hough indings o some
s udies sugges ha di e ences in DNA me hyla ion may
p ecede he de elopmen o adiposi y, ecen s udies in
adul s, using me hods such as Mendelian andomiza ion,
posi ha al e a ions in DNA me hyla ion a e p edomin-
an ly he consequence o adiposi y, a he han he cause
[4,5,9,26,27]. The di ec ion o any causal pa hway has
no been obus ly app aised in child en. Ob aining mo e
knowledge on he associa ion be ween DNA me hyla ion
and adiposi y al eady in childhood may ha e implica ions
o u u e p e en ion o obesi y and i s ela ed ai s.
We pe o med a me a-analysis o epigenome-wide asso-
cia ion s udies o BMI in up o 4133 pa icipan s om 23
s udies. We assessed associa ions o DNA me hyla ion in
co d blood, in childhood and adolescence wi h BMI in
child en aged 2–18 yea s. We also compa ed he e ec es-
ima es and examined whe he he e was en ichmen in
ou da a o CpGs p e iously iden i ied o hei associ-
a ion wi h adolescen and adul adiposi y.
Me hods
Pa icipan s
We me a-analyzed epigenome-wide associa ion s udies
o co d o whole blood me hyla ion wi h childhood o
adolescen body mass index (BMI). We used da a om
up o 4133 pa icipan s om 23 s udies collabo a ing in
he P egnancy And Childhood Epigene ics (PACE) Con-
so ium, Li eCycle P ojec , and Nu iP og am P ojec
(Addi ional ile 1: Table S1A-D and Addi ional ile 2:
Supplemen a y Me hods) [28,29]: ALSPAC, BAMSE,
CHAMACOS, CHOP S udy, CHS, DOMInO T ial,
GECKO D en he coho , Gene a ion R S udy, GOYA
s udy, Heal hy S a S udy, HELIX, INMA, IOW F1,
IOW F2, MoBa1, MoBa2, NEST, NFBC 1986, PIAMA
s udy, PREDO s udy, P ojec Vi a, Raine, and STOPPA
( ull names in Supplemen a y Me hods). Coho pa ici-
pan s we e mainly o Eu opean ances y, bu he e we e
also coho s wi h (pa ly) non-Eu opean e hnici ies (A -
ican, Hispanic, and Abo iginals). Mos coho s a e p o-
spec i e bi h coho s. We excluded mul iple bi hs,
siblings (maximum one child pe amily), physician-
diagnosed synd omic obesi y cases, and any ype o ma-
e nal diabe es (including ges a ional diabe es). In o med
consen was ob ained o all pa icipan s, and all s udies
Vehmeije e al. Genome Medicine (2020) 12:105 Page 2 o 15
ecei ed app o al om hei local e hics commi ees
(see Addi ional ile 2: Supplemen a y Me hods).
DNA me hyla ion
DNA me hyla ion was measu ed in co d blood and
whole blood samples, in child en and adolescen s using
he Illumina In inium® HumanMe hyla ion450 BeadChip
assay (Illumina, San Diego, CA, USA) [30]. Each coho
independen ly conduc ed hei p e e ed quali y con ol
and no maliza ion me hod, see Addi ional ile 2: Supple-
men a y Me hods o de ails. Un ans o med no malized
be a alues o indi idual CpG si es we e used as expos-
u e a iables. I mul iple measu emen s o DNA me hy-
la ion and BMI we e a ailable wi hin an age ange, we
used he oldes age wi hin ha ange o which BMI and
DNA me hyla ion we e a ailable a he same ime poin .
Ou lying me hyla ion be a alues we e excluded using
he ollowing me hod: alues < (25 h pe cen ile −3*in e -
qua ile ange (3IQR)) and alues > (75 h pe cen ile +
3IQR) we e emo ed [31]. DNA me hyla ion is
exp essed as he p opo ion o alleles a which he DNA
was me hyla ed a a speci ic si e and hence akes alues
om ze o o one.
Childhood BMI
Heigh and weigh we e measu ed in each s udy using
es ablished p o ocols as desc ibed in de ail in he Add-
i ional ile 2: Supplemen a y Me hods. The p ima y ou -
come was BMI, calcula ed as weigh /heigh
2
in kg/m
2
,ona
con inuous scale measu ed in h ee age anges: 2–5 yea s
(ea ly childhood), 5–10 yea s (la e childhood), and 14–18
yea s (adolescence). I mul iple BMI and DNA me hyla ion
measu emen s we e a ailable, we used he measu emen s
a heoldes agewi hin heage ange o whichBMIand
DNA me hyla ion we e a ailable a he same ime poin .
BMI alues we e hen ans o med in o sex- and age-
adjus ed s anda d de ia ion sco es (SDS) using LMSG ow h
[32–34]. The In e na ional Obesi y Task Fo ce (IOTF)
s anda d was used o de ine cu o s o BMI o unde -
weigh , no mal weigh , o e weigh , and obesi y in child en,
c ea ed wi h he B i ish 1990 g ow h e e ence and in o -
ma ion o pa icipan s on BMI, sex, and age [35,36]. In sec-
onda y analyses, we used a bina y ou come a iable wi h
no mal-weigh child en as con ols and o e weigh o
obese child en as cases. Unde weigh child en we e ex-
cluded om hese seconda y analyses. I a s udy had ≤10
pa icipan s in one o he (case o con ol) g oups, his
s udy was excluded om he seconda y analyses.
Co a ia es
Co a ia es included in all models we e ma e nal co a i-
a es: ma e nal age, ma e nal educa ional le el (coho de -
ini ion), ma e nal smoking s a us du ing p egnancy (any
smoking e sus no smoking), ma e nal p e-p egnancy o
ea ly p egnancy BMI and pa i y (mul ipa ous e sus nul-
lipa ous), and ges a ional age a bi h. Fo de ails on
coho -speci ic collec ion me hods, see Addi ional ile 2:
Supplemen a y Me hods. We es ima ed whi e blood cell
p opo ions (B cells, CD8+ T cells, CD4+ T cells, g anulo-
cy es, NK cells, and monocy es) using he e e ence-based
Houseman me hod wi h he Reinius e e ence in he min i
package in R [37–40]. A sensi i i y analysis using he co d
blood-speci ic Bakulski e e ence was pe o med in he
Gene a ion R and ALSPAC s udies [41]. Ba ch e ec s
we e adjus ed o using coho -speci ic me hods, see Add-
i ional ile 2: Supplemen a y Me hods. Addi ional co a i-
a es added in he c oss-sec ional childhood analyses we e
bi h weigh and b eas eeding. The adolescen analyses
we e addi ionally adjus ed o adolescen age, sex, own
smoking s a us, and pube y s a us. Pube y s a us was
ca ego ized in o ea ly pube y (i bo h b eas and pubic
hai Tanne s ages (o compa able classi ica ion) we e 1,
2, o 3 and i gi ls we e p e-mena cheal o boys did no
ha e oice change ye ) and la e pube y (i ei he b eas o
pubic hai Tanne s ages (o compa able classi ica ion)
we e 4 o 5 o i gi ls we e pos -mena cheal o boys had
had hei oice change) [42–44]. Fu he de ails a e p o-
ided in he s udy-speci ic Addi ional ile 2: Supplemen-
a y Me hods.
S udy-speci ic analyses
Associa ions o DNA me hyla ion wi h childhood o
adolescen BMI we e pe o med in indi idual s udies on
pa icipan s wi h comple e da a on all co a ia es. In
s udies wi h mo e han one e hnic g oup, each g oup
was analyzed sepa a ely. We used obus linea eg es-
sion models o he con inuous ou come o BMI-SDS
and gene alized linea eg ession models o he case/
con ol analyses o o e weigh and obesi y e sus no mal
weigh , acco ding o a p e-speci ied analysis plan. EWAS
analyses we e conduc ed using DNA me hyla ion a
h ee ime poin s: bi h, childhood and adolescence, and
BMI da a collec ed a h ee ime poin s: ea ly childhood
(2–5 yea s), la e childhood (5–10 yea s), and adolescence
(12-18y) (Table 1). We ca ego ized he childhood pe iod
in o ea ly and la e childhood o o e come any age-
speci ic e ec s and he po en ial in luence o he adipos-
i y ebound on he esul s [45]. Depending on da a a ail-
abili y, coho s pa icipa ed in one o mo e o ou
analyses: (analysis A) longi udinal associa ions o co d
blood DNA me hyla ion wi h ea ly childhood BMI (2–
5 yea s; 3295 child en om 13 s udies), (analysis B) lon-
gi udinal associa ions o co d blood DNA me hyla ion
wi h la e childhood BMI (5–10 yea s; 4133 child en om
12 s udies), (analysis C) c oss-sec ional associa ions o
childhood blood DNA me hyla ion wi h childhood BMI
(2–10 yea s; 3371 child en om 11 s udies), and (ana-
lysis D) c oss-sec ional associa ions o adolescen blood
Vehmeije e al. Genome Medicine (2020) 12:105 Page 3 o 15
DNA me hyla ion wi h adolescen BMI (14–18 yea s;
2842 adolescen s om 7 s udies) (Table 1). Pa icipa ing
s udies pe analysis a e shown in Addi ional ile 1: Table
S1A-D.
Co d blood analyses we e adjus ed o ma e nal age,
educa ional le el, smoking s a us, p e-p egnancy o ea ly
p egnancy BMI, pa i y, ges a ional age, ba ch, and es i-
ma ed cell ype p opo ions. The c oss-sec ional analysis
in childhood was addi ionally adjus ed o child co a i-
a es bi h weigh and b eas eeding; in con as , he
c oss-sec ional analysis in adolescence was adjus ed o
he same co a ia es as analysis C plus adolescen sex,
age, and smoking and pube y s a us.
Me a-analyses
A e pe o ming quali y con ol on all s udies, we com-
bined esul s in a ixed-e ec s in e se a iance-weigh ed
me a-analysis using METAL [46,47]. All ollow-up ana-
lyses we e conduc ed in R [39]. The me a-analyses we e
done independen ly by wo s udy g oups, and he esul s
we e compa ed. A e exclusion o p obes ha we e
measu ed in only one s udy, ha mapped o X and Y
ch omosomes and p obes ha co-hyb idized o al e na e
sequences (c oss- eac i e p obes), we included 429,959
p obes o analysis A, 429,959 p obes o analysis B, 429,
957 p obes o analysis C, and 428,967 p obes o ana-
lysis D [48,49]. In he esul iles o he main me a-
analyses, we lagged p obes ha map o DNA con aining
a single nucleo ide polymo phism (SNP), o epe i i e
sequence elemen s, o o DNA ha bo ing an INDEL
(Addi ional ile 3: Table S2A-D) [48,49]. We co ec ed
o mul iple es ing using bo h he Bon e oni co ec-
ion, which gi es a signi icance h eshold o P< 1.16 ×
10
−7
(0.05/429,959), and he less s ingen alse disco e y
a e (FDR) h eshold using he me hod by Benjamini
and Hochbe g [50]. EWAS esul s we e summa ized as
mean (and s anda d e o ) di e ences in BMI-SDS pe
10% inc ease in me hyla ion o each CpG. We c ea ed
olcano plo s o isualize magni ude and di ec ion o e -
ec ( educed o inc eased me hyla ion) along wi h he
le el o s a is ical signi icance. We calcula ed he I
2
s a -
is ic o explo e he e ogenei y ac oss s udies. The I
2
es i-
ma es he p opo ion o a ia ion in he me a-analysis
esul s o each CpG si e ha is due o be ween-s udy
di e ences a he han andom/sampling a ia ion. He -
e ogenei y was de ined as an I
2
alue o > 50 and shown
g aphically in o es plo s. We pe o med lea e-one-ou
analyses, in which we e an he main me a-analysis e-
pea edly wi h one o he 23 s udies emo ed each ime,
o explo e i any s udy in luenced indi idual indings.
We enhanced he anno a ion p o ided by Illumina using
he UCSC Genome B owse . All o he anno a ions use
he human Feb ua y 2009 (GRCh37/hg19) assembly. We
upda ed he gene names manually in all esul iles using
HUGO gene nomencla u e, and in case hey we e no
ound he e, we used he NCBI gene websi e on No em-
be 5, 2019 [51–53].
To explo e he associa ions o he ex eme uppe
alues o he BMI dis ibu ion, we pe o med case/con-
ol analyses (o e weigh and obesi y e sus no mal
weigh ). Unde weigh child en we e excluded om hese
analyses, leading o sample sizes o N= 491 cases and
2540 con ols (analysis A), N= 707 cases and 3217 con-
ols (analysis B), N= 644 cases and 2567 con ols (ana-
lysis C), and N= 507 cases and 2188 con ols (analysis
D) (Addi ional ile 4: Table S3A-D).
To examine whe he any o he Bon e oni-signi ican
o FDR-signi ican CpGs in ou analyses we e close o
Table 1 O e iew o main analyses, seconda y analyses, and sensi i i y analyses
Analysis Main analyses Seconda y analyses: bina y
model (N), cases = o e weigh
and obesi y, con ols = no mal
weigh
Sensi i i y analyses
DNA me hyla ion
in he blood
BMI SD sco es NEu opeans
only (N)
Wi hou s udies >
30% o e weigh
and obesi y (N)
Co d blood analyses
A Bi h (co d blood) Ea ly childhood
(2–5 yea s)
3295 Cases = 491
Con ols = 2540
2902 2989
B Bi h (co d blood) La e childhood
(5–10 yea s)
4133 Cases = 707
Con ols = 3217
3657 3489
C oss-sec ional analyses
C Childhood (whole blood) Childhood
(2–10 yea s)
3371 Cases = 644
Con ols = 2567
3026 3171
D Adolescence (whole blood) Adolescence
(12–18 yea s)
2842 Cases = 507
Con ols = 2188
NA NA
Analyses A and B we e adjus ed o ma e nal age, educa ional le el, smoking s a us, p e-p egnancy o ea ly p egnancy BMI, pa i y, ges a ional age a bi h, ba ch,
and es ima ed cell ype p opo ions
Analyses C was adjus ed o ma e nal age, educa ional le el, smoking s a us, p e-p egnancy o ea ly p egnancy BMI, pa i y, ges a ional age a bi h, ba ch,
es ima ed cell ype p opo ions, bi h weigh , and b eas eeding
Analyses D was adjus ed o ma e nal age, educa ional le el, smoking s a us, p e-p egnancy o ea ly p egnancy BMI, pa i y, ges a ional age a bi h, ba ch,
es ima ed cell ype p opo ions, bi h weigh , b eas eeding, adolescen sex, age smoking and pube y s a us
Vehmeije e al. Genome Medicine (2020) 12:105 Page 4 o 15
BMI SNPs, we assessed i hese CpGs we e loca ed
wi hin a 4-Mb window (± 2 Mb) su ounding he 15 gen-
e ic loci associa ed wi h childhood body mass index [2,
54]. Fo he FDR-signi ican CpGs ha we e lagged be-
cause hey we e po en ially in luenced by a SNP, we
isually inspec ed densi y plo s in he Gene a ion R
S udy o see whe he hese de ia ed om unimodali y
(Addi ional ile 5: Supplemen a y In o ma ion, Fig. S6).
To explo e DNA me hyla ion pa e ns in he egions
a ound he signi ican CpGs, we assessed he associa-
ions o all CpGs loca ed wi hin a 10-kb window (± 5 kb)
su ounding hese CpGs wi h BMI in he ele an
models (Addi ional ile 6: Table S4).
Sensi i i y analyses
To explo e whe he e hnic he e ogenei y may ha e a -
ec ed ou esul s, we epea ed he me a-analyses includ-
ing s udies wi h pa icipan s o Eu opean ances y only
(N= 2902 (excluding h ee s udies o analysis A), N=
3657 (excluding h ee s udies o analysis B), N=3026
(excluding wo s udies o analysis C)), he la ges e hnic
subg oup (Addi ional ile 7: Table S5A-C). E hnici y was
de ined using sel - epo ed ques ionnai es unless speci-
ied o he wise in he s udy-speci ic Supplemen a y
Me hods (Addi ional ile 2). We pe o med addi ional
analyses excluding s udies wi h a high pe cen age (> 30%
(pe cen age calcula ed a e exclusion o unde weigh
child en)) o child en wi h o e weigh and obesi y o ex-
plo e whe he any associa ions ound may be d i en by
mo e ex eme alues o BMI (included N= 2989 (exclud-
ing wo s udies o analysis A), N= 3489 (excluding ou
s udies o analysis B), N= 3171 (excluding one s udy o
analysis C) (Addi ional ile 8: Table S6A-C). We also
pe o med a hi d, conse a i e, sensi i i y analysis in all
age g oups, excluding coho s o non-Eu opeans, s udies
wi h a high pe cen age (> 30%) o child en wi h o e -
weigh o obesi y, and s udies in which he sample was
selec ed on o en iched o any pa icula exposu e o
ou come (Addi ional ile 9: Table S7A-D).
Compa ison wi h p e ious indings
We explo ed whe he CpG si es associa ed wi h child-
hood, adolescen , o adul BMI in p e ious s udies we e
associa ed wi h BMI in ou da a. Fo p e ious candida e
gene s udies and smalle EWASs (N< 1000), we pe -
o med a look-up o he hi s, using a Bon e oni-adjus ed
P alue cu o pe s udy, so o each s udy, he cu o was
0.05/(NCpGs om ha s udy) (Addi ional ile 10:Table
S8) [7,15,17,18,20,21,23,24,55]. I he speci ic CpGs
om a s udy we e no a ailable in ou da ase , we looked
up all CpGs anno a ed o he ele an genes [17,24]. To
es ablish whe he he CpG si es associa ed wi h BMI in
p e iously epo ed la ge EWASs (N≥1000) we e o e -
ep esen ed among ou CpGs wi h he smalles P alues,
we examined he absolu e o e lap o he op CpGs om
li e a u e wi h he op CpGs in ou analyses [5,6,9,25,
26,56–59]. The la e we e de ined using wo cu o s: a
s ingen cu o o P alue < 1 × 10
−5
and a mo e lenien
one o P alue < 0.05. (Addi ional ile 11: Table S9). We
used a hype geome ic es o calcula e en ichmen wi h
he phype unc ion in he R S a s package in R.
We examined he 187 CpGs iden i ied in he la ges
adul s udy (N= 10,261) o da e in mo e de ail in ou e-
sul s [5]. We es ed whe he he en iched CpGs signi i-
can ly o e lapped be ween ou analyses using chi-squa e
es s. We used Pea son’s co ela ion coe icien s o
examine he co ela ions be ween he e ec es ima es o
hese 187 CpGs in adul s and hose in ou analyses [5].
Using Fishe ’s exac es , we calcula ed whe he he co -
ela ion coe icien s a he a ious ages we e signi ican ly
di e en om each o he .
Func ional analyses
We explo ed he po en ial unc ional in e p e a ion o
he mos signi ican ly associa ed CpGs (P alue < 1 ×
10
−4
) in all models using Gene On ology (GO) and
Kyo o Encyclopedia o Genes and Genomes (KEGG) en-
ichmen analyses. We used he missMe hyl package,
which enabled us o co ec o he numbe o p obes
pe gene on he 450K a ay, based on he May 5, 2020,
e sion o he GO and he Oc obe 23, 2019, e sion o
he KEGG sou ce da abases [60]. To il e ou he la ge,
gene al pa hways, we se he numbe o genes o each
gene se be ween 5 and 2000, espec i ely. We epo
nominal P alues < 0.05 and FDR o en ichmen (Add-
i ional ile 12, Table S10).
Resul s
Pa icipan s
We included 2842 o 4133 pa icipan s om 23 independ-
en coho s om he P egnancy And Childhood Epigene -
ics (PACE) Conso ium [28]. We assessed associa ions o
DNA me hyla ion in co d blood wi h BMI in ea ly child-
hood (2–5yea s)(N= 3295, analysis A), DNA me hyla ion
in co d blood wi h BMI in la e childhood (5–10 yea s)
(N= 4133, analysis B), DNA me hyla ion in childhood
wi h BMI in childhood (2–10 yea s) (N= 3371, analysis
C), and DNA me hyla ion in adolescence wi h BMI in
adolescence (12–18 yea s) (N= 2842, analysis D). De ails
o pa icipan s and s udies used in he di e en analyses
a e p esen ed in Table 1, Addi ional ile 1: Table S1A-D
and Addi ional ile 2: Supplemen a y Me hods.
Me a-analyses
The main, seconda y, and sensi i i y analyses a e ou -
lined in Table 1. Genomic in la ion ac o s (lambdas) o
he main me a-analyses anged be ween 0.97 and 1.27
(Addi ional ile 5: Supplemen a y in o ma ion, Fig. 1a-d).
Vehmeije e al. Genome Medicine (2020) 12:105 Page 5 o 15

Genomic in la ion ac o s (lambdas) o all coho -speci ic
analyses a e shown in Addi ional ile 1: Table S1A-D. The
main esul s a e shown in Table 2and Fig. 1. We did no
iden i y associa ions a genome-wide signi icance o DNA
me hyla ion in co d blood wi h BMI in ea ly childhood
(analysis A, Fig. 1a, and Addi ional ile 3: Table S2A).
DNA me hyla ion a one CpG, cg05937453 (SFRP5), in
co d blood was signi ican ly associa ed wi h la e-
childhood BMI (analysis B, Fig. 1b, and Addi ional ile 3:
Table S2B). Fo each 10% inc ease in DNA me hyla ion a
cg05937453 in co d blood, la e-childhood BMI inc eased
0.96 SD (s anda d e o (SE) 0.17). Co d blood DNA
me hyla ion a his CpG was nominally signi ican ly asso-
cia ed wi h BMI in ea ly-childhood (P alue = 0.004), bu
DNA me hyla ion in childhood and adolescence was no
associa ed wi h BMI in he c oss-sec ional analyses (Add-
i ional ile 13: Table S11).
In he c oss-sec ional analysis (analysis C), childhood
DNA me hyla ion a cg25212453 (SLC43A2) was associa ed
wi h childhood BMI a e Bon e oni co ec ion. A 10% in-
c ease in DNA me hyla ion a cg25212453 was associa ed
wi h a 0.32 SD (SE 0.06) inc ease in childhood BMI (Fig. 1c
and Addi ional ile 3:TableS2C).DNAme hyla iona his
CpG a bi h and in adolescence was no associa ed wi h
BMI (Addi ional ile 13: Table S11). DNA me hyla ion in
childhood a nine addi ional CpGs in o nea o he genes
was associa ed wi h childhood BMI using FDR P alue <
0.05 (Fig. 1c and Addi ional ile 3:TableS2C).DNAme hy-
la ion in adolescence a cg10040131 (SFXN5) was associ-
a ed wi h adolescen BMI a e Bon e oni co ec ion
(analysis D, Fig. 1d and Addi ional ile 3: Table 2d). A 10%
inc ease in DNA me hyla ion a cg10040131 was associa ed
wi h a 0.32 SD (SE 0.06) highe BMI in adolescence. DNA
me hyla ion a his CpG in childhood was nominally
signi ican ly associa ed wi h childhood BMI (P alue =
0.0002). The associa ion o DNA me hyla ion a his
CpG in co d blood and BMI in childhood was no sig-
ni ican (Addi ional ile 13: Table S11).
Associa ions o DNA me hyla ion wi h BMI did no show
a p e e en ial di ec ion o e ec in any o he analyses ( ol-
cano plo s, Addi ional ile 5: Supplemen a y In o ma ion,
Fig. S2A-D). We obse ed e y li le e idence o he e ogen-
ei y be ween s udies among he Bon e oni-signi ican ly
associa ed CpG si es, wi h all I
2
≤50 (Addi ional ile 3:
Table 2a-d and o es plo s, Addi ional ile 5: Supplemen-
a y In o ma ion, Fig. S3A, B and L). We ound e idence o
Fig. 1 Manha an plo s o he me a-analyses o DNA me hyla ion and childhood o adolescen BMI. Manha an plo s showing he me a-analysis
esul s o associa ions o DNA me hyla ion in co d blood wi h ea ly childhood BMI (a) and la e childhood BMI (b), o DNA me hyla ion in whole
blood in childhood wi h childhood BMI (c), and o DNA me hyla ion in whole blood in adolescence wi h adolescen BMI (d). The g ay line shows
he Bon e oni-co ec ed signi icance h eshold o mul iple es ing (P< 1.06 × 10
−7
). The o ange line shows he FDR-co ec ed signi icance
h eshold o mul iple es ing
Vehmeije e al. Genome Medicine (2020) 12:105 Page 6 o 15
be ween-s udy he e ogenei y (I
2
> 50) o 3 o he 9 FDR-
signi ican ly associa ed CpG si es (Addi ional ile 3:Table2c
and o es plo s, Addi ional ile 5: Supplemen a y In o ma-
ion, Fig. S3C-K). The esul s o he wel e Bon e oni o
FDR-signi ican ly associa ed CpGs we e s able a e omi -
ing one s udy a a ime (lea e-one-ou analyses, Add-
i ional ile 5: Supplemen a y In o ma ion, Fig. S4A-L).
When BMI was dicho omized in o no mal and o e -
weigh /obesi y, only one CpG in he c oss-sec ional
model in childhood, cg06991974 (PRDM16-DT), showed
e idence o associa ion. In he c oss-sec ional model
du ing childhood, which included 644 child en wi h
o e weigh /obesi y and 2567 no mal-weigh child en,
DNA me hyla ion a cg06991974 was associa ed wi h an
inc eased isk o o e weigh /obesi y in childhood (odds
a io (OR) 3.10, 95% con idence in e al (CI) 2.08, 4.63)
(Addi ional ile 4: Table S3A-D).
None o he h ee indi idual Bon e oni-signi ican CpGs
in he h ee di e en age anges no he 9 FDR-signi ican
CpGs was wi hin a 4-Mb window su ounding he 15 known
gene ic loci associa ed wi h childhood body mass index [54].
Fou o he 12 FDR signi ican CpGs con ained a
single-nucleo ide polymo phism (SNP) [48,49]. We
ound no indica ion o non-unimodal dis ibu ion o
any o hese CpGs sugges ing ha me hyla ion measu e-
men s a hese si es we e no ma kedly a ec ed by SNPs
(Addi ional ile 5: Supplemen a y In o ma ion, Fig. S6).
Two o he h ee Bon e oni-signi ican CpGs (cg05937453
and cg25212453) had o he nea by CpGs wi hin a 10-kb
window (± 5 kb) measu ed on he 450K a ay (Add-
i ional ile 6: Table S4). Cg05937453 (model B) was su -
ounded by 24 o he CpGs, o which one was nominally
signi ican ly associa ed wi h BMI (P alue < 0.05). Bo h we e
loca ed in he TSS200 egion o SFRP5 wi h e ec es ima es
in he same di ec ion. Cg25212453 (model C) was su -
ounded by 13 o he CpGs, o which h ee we e nominally
signi ican (P alues < 0.05). All we e loca ed in he gene body
o SLC43A2 wi h e ec es ima es in he same di ec ion. Re-
sul s o Bon e oni- and FDR-signi ican CpGs a e shown in
Addi ional ile 6:TableS4.
Sensi i i y analyses
Findings we e consis en wi h he main esul s when e-
s ic ed o up o 3657 pa icipan s o Eu opean e hnici y
(Pea son co ela ion coe icien s o he e ec es ima es
ac oss all CpG si es we e 0.86–0.97 and we e 0.99 ac oss
op CpG si es (P alue < 1 × 10
−4
) o all models (Add-
i ional ile 7: Table S5A-C)). Simila ly, when he s udies
wi h a high pe cen age (> 30%) o child en wi h o e -
weigh o obesi y we e excluded, he esul s we e also
consis en wi h he main analyses (Pea son co ela ion
coe icien s o he e ec es ima es ac oss all CpG si es
we e 0.89–0.98 and we e 0.99 ac oss op CpG si es (P
alue < 1 × 10
−4
) o all models (Addi ional ile 8: Table
S6A-C)). Las ly, when he s udies o non-Eu opeans pa -
icipan s, a high pe cen age o child en wi h o e weigh
o obesi y and s udies in which he sample was selec ed
on o en iched o any pa icula exposu e o ou come
Table 2 CpG si es a which DNA me hyla ion was associa ed wi h child o adolescen BMI
CpG CHR Loca ion Coe SE P alue FDR P alue Nea es gene
Analysis B = associa ion o co d blood DNA me hyla ion wi h la e childhood BMI (5–10 yea s)
cg05937453 10 99531765 0.96288 0.16871 1.15 × 10
−8
0.0049 SFRP5
Analysis C = c oss-sec ional associa ion o whole blood DNA me hyla ion wi h childhood BMI (2–10 yea s)
cg25212453 17 1509953 0.31925 0.05978 9.27 × 10
−8
0.02075 SLC43A2
cg03500056 16 8814507 0.30577 0.05767 1.15 × 10
−7
0.02075 ABAT
cg05281708 3 44690673 0.65856 0.12614 1.78 × 10
−7
0.02075 ZNF35
cg15125798 5 122621645 0.49705 0.09548 1.93 × 10
−7
0.02075 –
cg04456029 12 113496126 0.27587 0.05358 2.63 × 10
−7
0.0226 DTX1
cg03431111 11 62621406 0.19261 0.03791 3.77 × 10
−7
0.0270 SNORD30;SNORD22;SNORD29;
SNORD31;SNHG1
cg26889953 15 22915992 0.31743 0.06391 6.81 × 10
−7
0.0304 CYFIP1
cg19743522 12 113495566 0.33854 0.0682 6.92 × 10
−7
0.0304 DTX1
cg25877069 8 95003236 −0.45126 0.09092 6.94 × 10
−7
0.0304 –
cg13931559 20 33146515 −0.84718 0.17082 7.07 × 10
−7
0.0304 MAP1LC3A
Analysis D = c oss-sec ional associa ion o whole blood DNA me hyla ion wi h adolescen BMI (12–18 yea s)
cg10040131 2 73178866 0.32434 0.0566 1.00 × 10
−8
0.0043 SFXN5
Coe icien s (Coe ) and s anda d e o s (SE) a e p esen ed pe 10% inc ease in he me hyla ion le el
Analyses B was adjus ed o ma e nal age, educa ional le el, smoking s a us, p e-p egnancy o ea ly p egnancy BMI, pa i y, ges a ional age, ba ch, and es ima ed
cell ype p opo ions. Analysis C was addi ionally adjus ed o child co a ia es bi h weigh and b eas eeding, whe eas analysis D was adjus ed o he same
co a ia es as analysis C plus adolescen sex, age, smoking, and pube y s a us
Vehmeije e al. Genome Medicine (2020) 12:105 Page 7 o 15
we e all excluded, esul s emained s ongly co ela ed
o hose om he main models. Pea son co ela ion
coe icien s o he e ec es ima es ac oss all CpG si es
we e 0.64–0.97 and 0.95–0.99 ac oss op CpG si es (P
alue < 1 × 10–4) o all models (Addi ional ile 9:
Table S7A-D).
Compa ison wi h p e ious indings
Mos CpGs iden i ied o be associa ed wi h BMI in p e-
ious candida e gene s udies o smalle EWASs (N<
1000) did no eplica e in ou esul s (Addi ional ile 10:
Table S8) [7,15,17,18,20,21,23,24,55]. When com-
pa ing he genome-wide signi ican indings om he
la ges BMI EWASs (N> 1000) in adul s o ou mos sig-
ni ican indings ac oss he ou age anges, we ound an
inc easing o e lap wi h age (Table 3and Add-
i ional ile 11: Table S9) [5,6,9,25,26,56–59]. We used
wo cu o s o selec he mos signi ican indings in ou
esul s: a P alue < 1 × 10
−5
, o iden i y “sugges i e” ind-
ings, and a less s ingen , nominal P alue < 0.05. The
numbe o CpGs ha me hese c i e ia a e p o ided in
Table 3. Fi s , we examined he absolu e numbe o
o e lapping CpGs be ween he s udies in adul s and ou
indings wi h a P alue < 1 × 10
−5
and calcula ed en ich-
men . Wi h ad ancing age ac oss childhood and adoles-
cence, we obse ed inc easing en ichmen o he 187
CpGs p e iously epo ed o be associa ed wi h adul
BMI in he la ges s udy o da e (N= 10,261) [5]. Fo he
wo co d blood models, he e was no o e lap wi h he
adul indings (P
en ichmen
= 1), o he c oss-sec ional
model in childhood 2/187 adul hi s o e lapped,
(P
en ichmen
= 0.0002), and o he c oss-sec ional
model in adolescence 3/187 o e lapped (P
en ichmen
=
2.10 × 10
−7
)(Table3and Addi ional ile 11:Table
S9). Using he less s ingen cu o (P alue < 0.05),
his end was e en clea e . The o e lap be ween he
187 CpGs om he adul EWAS and he CpGs in
ou da a wi h a P alue < 0.05 was 8/187 CpGs
(P
en ichmen
= 0.77, analysis A) o he associa ion o
co d blood DNA me hyla ion and ea ly childhood BMI
and 11/187 CpGs (P
en ichmen
= 0.30, analysis B) o he as-
socia ion o co d blood DNA me hyla ion and la e child-
hood BMI. Fo he c oss-sec ional model in childhood, he
o e lap was 61/187 CpGs (P
en ichmen
= 1.97 × 10
−20
,ana-
lysis C), and in adolescence, he o e lap was 77/187 CpGs
(P
en ichmen
= 1.68 × 10
−44
, analysis D) (Table 3and Add-
i ional ile 11: Table S9). Twen y-se en CpGs we e among
he en iched CpGs in bo h he childhood and he adoles-
cen model. This o e lap was no signi ican (P= 0.88).
Co ela ion coe icien s be ween he e ec es ima es o
he 187 hi s and he e ec es ima es o hose CpGs in
he ou models inc eased wi h age (analysis A = −0.186
(P= 0.01), analysis B = −0.013 (P= 0.86), analysis C =
0.604 (P= 5.31 × 10
−20
), and analysis D = 0.816 (P=
7.89 × 10
−46
). The di e ence in co ela ion coe icien s
was signi ican o all compa isons (P’s o compa ison
be ween co ela ion coe icien s < 0.01) excep o he
compa ison be ween models A and B (P= 0.09).
E ec sizes o he associa ions o hese 187 adul BMI
CpGs in ou analyses inc eased wi h ad ancing age o chil-
d en in ou analyses (Addi ional ile 5: Supplemen a y In-
o ma ion, Fig. S5). We ound simila ends o
en ichmen o CpGs om o he EWASs in adul s and ad-
olescen s (Table 3)[6,9,25,26,56–59]. O hose indings
om adul s udies ha had a nominal P alue (< 0.05) in
ou models, 17–35%we e epo edbymo e hanone
adul s udy. Mos o hese we e ound in wo o h ee
s udies, bu ou , cg06500161, cg19750657, cg12593793,
and cg18181703, we e epo ed in six o se en p e ious
analyses.
Func ional analyses
A unc ional en ichmen analysis using genes linked o he
CpGs wi h P alues < 1 × 10
−4
in each o he models
showed no unc ional en ichmen o Gene On ology (GO)
e ms o Kyo o Encyclopedia o Genes and Genomes
(KEGG) e ms (FDR < 0.05) (Addi ional ile 12: Table S10).
Discussion
In his la ge me a-analysis o EWASs o childhood and
adolescen BMI, we ound li le e idence o an associa ion
be ween DNA me hyla ion and childhood o adolescen
BMI. DNA me hyla ion a h ee di e en CpGs, each one
in a di e en age ange, was associa ed wi h BMI in ea ly
li e. Wi h he ad ancing age o child en in ou analyses,
we obse ed inc easing en ichmen o CpGs p e iously
iden i ied o hei ela ion wi h adolescen o adul adi-
posi y. In addi ion, o he 187 CpGs iden i ied in he la -
ges p e ious s udy o adul BMI, we ound inc easing
e ec sizes and inc easing co ela ions be ween he adul
e ec sizes and hose in ou analyses, wi h age.
In e p e a ion o main indings
Childhood obesi y is a majo public heal h p oblem and
associa ed wi h sho - and long- e m mo bidi y and mo -
ali y [61]. Al hough he e is some e idence om p e ious
s udies ha DNA me hyla ion may media e associa ions
o p egnancy- ela ed exposu es wi h o sp ing adiposi y,
only ew speci ic CpG si es ha e been iden i ied [4,27].
Thus a , mos o he e idence o associa ions o DNA
me hyla ion wi h adiposi y s ems om adul s udies.
In his s udy, we ound li le e idence o an associa ion
be ween DNA me hyla ion and childhood o adolescen
BMI. DNA me hyla ion a h ee CpGs (cg05937453,
cg25212453, and cg10040131), each in a di e en age
ange, was associa ed wi h BMI a Bon e oni signi i-
cance, P< 1.06 × 10
−7
. Howe e , we did obse e inc eas-
ing en ichmen and inc easing poin es ima es o CpGs
Vehmeije e al. Genome Medicine (2020) 12:105 Page 8 o 15
Table 3 Absolu e numbe o o e lapping CpGs and P alues o he en ichmen o signi ican CpGs om p e ious EWASs (N> 1000)
in ou da a
P e ious s udy
(Nsi es associa ed
wi h BMI)
Signi icance
le el
Analysis A: associa ion
o co d blood DNA
me hyla ion wi h ea ly
childhood BMI
(2–5 yea s)
Analysis B: associa ion
o co d blood DNA
me hyla ion wi h la e
childhood BMI
(5–10 yea s)
Analysis C: c oss-
sec ional analysis o
whole blood DNA
me hyla ion wi h
childhood BMI
(2–10 yea s)
Analysis D: c oss-sec ional
analysis o whole blood
DNA me hyla ion wi h
adolescen BMI (12–18
yea s)
1×10
−5
N=7 N=8 N=51 N=26
0.05 N= 22,687 N= 20,645 N= 37,074 N= 25,292
Ali e al. [56] (3 CpGs) 1 × 10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
0.05 0
P
en ichmen
=1
0
P
en ichmen
=1
1/3
P
en ichmen
= 0.24
0
P
en ichmen
=1
Aslibekyan e al. [6](8
CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
0.05 0
P
en ichmen
=1
0
P
en ichmen
=1
1/8
P
en ichmen
= 0.51
2/8
P
en ichmen
= 0.08
Campanella e al. [57]
(26 CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
1/26
P
en ichmen
= 0.002
0.05 3/26
P
en ichmen
= 0.16
1/26
P
en ichmen
= 0.72
6/26
P
en ichmen
= 0.02
11/26
P
en ichmen
= 1.006 × 10
−7
Geu s e al. [58] (310
CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
2/310
P
en ichmen
= 0.0006
2/310
P
en ichmen
= 0.0002
0.05 12/310
P
en ichmen
= 0.90
13/310
P
en ichmen
= 0.73
103/310
P
en ichmen
= 3.92 × 10
−34
125/310
P
en ichmen
= 6.63 × 10
−70
Mendelson e al. [9]
(83 CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
2/83
P
en ichmen
= 4.67 × 10
−5
3/83
P
en ichmen
= 1.81 × 10
−8
0.05 4/83
P
en ichmen
= 0.64
8/83
P
en ichmen
= 0.045
28/83
P
en ichmen
= 1.36 × 10
−10
45/83
P
en ichmen
= 3.02 × 10
−33
Sayols-Baixe as e al. [59]
(96 CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
0.05 8/96
P
en ichmen
= 0.13
9/96
P
en ichmen
= 0.04
24/96
P
en ichmen
= 1.53 × 10
−6
30/96
P
en ichmen
= 1.85 × 10
−14
Sun e al. [26]black
pa icipan s (36 CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
1/36
P
en ichmen
= 0.002
0.05 3/36
P
en ichmen
= 0.30
6/36
P
en ichmen
= 0.007
13/36
P
en ichmen
= 4.98 × 10
−6
22/36
P
en ichmen
= 1.50 × 10
−18
Sun e al. [26]whi e
pa icipan s (349 CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
0.05 12/349
P
en ichmen
= 0.959
22/349
P
en ichmen
= 0.12
86/349
P
en ichmen
= 4.13 × 10
−19
116/349
P
en ichmen
= 1.75 × 10
−54
Wahl e al. [5] (187
CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
2/187
P
en ichmen
= 0.0002
3/187
P
en ichmen
= 2.10 × 10
−7
0.05 8/187
P
en ichmen
= 0.77
11/187
P
en ichmen
= 0.29
61/187
P
en ichmen
= 1.97 × 10
−20
77/187
P
en ichmen
= 1.68 × 10
−44
Wang e al. [25] (54
CpGs)
1×10
−5
0
P
en ichmen
=1
0
P
en ichmen
=1
0
P
en ichmen
=1
1/54
P
en ichmen
= 0.003
0.05 2/54
P
en ichmen
= 0.79
4/54
P
en ichmen
= 0.26
23/54
P
en ichmen
= 2.49 × 10
−11
33/54
P
en ichmen
= 3.98 × 10
−27
NCpGs in ≥2 adul
s udies
0.05 9/52 (17.3%) 23/75 (30.7%) 98/347 (28.2%) 163/465 (35.1%)
Two cu o s we e used o selec he signi ican indings in ou esul s: a P alue < 1 × 10
−5
, o iden i y “sugges i e” indings, and a less s ingen , nominal P alue
< 0.05, o iden i y any ends. We used a hype geome ic es o calcula e en ichmen wi h he phype unc ion in he R S a s package in R. Resul s in bold a e
nominally signi ican . O hose indings om adul s udies ha had a nominal P alue (< 0.05) in ou models, 17–35% we e epo ed by mo e han one adul s udy
Vehmeije e al. Genome Medicine (2020) 12:105 Page 9 o 15