Tyrrell, J., Richmond, R. C., Palmer, T. M., Feenstra, B., Rangarajan, J.,Metrusky, S., ... Freathy, R. M. (2016). Genetic Evidence for CausalRelationships Between Maternal Obesity-Related Traits and Birth Weight.JAMA - Journal of the American Medical Association, 315(11), 1129-1140.https://doi.org/10.1001/jama.2016.1975

Peer reviewed version

Link to published version (if available):10.1001/jama.2016.1975

Link to publication record in Explore Bristol ResearchPDF-document

University of Bristol - Explore Bristol ResearchGeneral rights

This document is made available in accordance with publisher policies. Please cite only the publishedversion using the reference above. Full terms of use are available:http://www.bristol.ac.uk/pure/about/ebr-terms

1

Dateofrevision:11February2016

Wordcount(text):3837

Geneticevidenceforcausalrelationshipsbetweenmaternalobesity-related

traitsandbirthweight

JessicaTyrrell,PhD1,2;RebeccaC.Richmond,PhD3,4,12;TomM.Palmer,PhD5,6*;BjarkeFeenstra,

PhD7;JananiRangarajan,MS8;SarahMetrustry,MSc9;AlanaCavadino,MSc10,11;LaviniaPaternoster,

PhD12;LorenL.Armstrong,PhD13;N.ManekaG.DeSilva,PhD12;AndrewR.Wood,PhD1;Momoko

Horikoshi,MD,PhD14,15;FrankGeller,MSc7;RonnyMyhre,PhD16;JonathanP.Bradfield,BS17;Eskil

Kreiner-Møller,MD18;VilleHuikari,MSc19;JodieN.Painter,PhD20;Jouke-JanHottenga,PhD21,22;

CatherineAllard,BSc23,24;DianeJ.Berry,PhD11;LuigiBouchard,PhD,MBA24-26;ShiktaDas,PhD27;

DavidM.Evans,PhD3,12,28;HakonHakonarson,MD,PhD17,29,30;M.GeoffreyHayes,PhD13;Jani

Heikkinen,MSc31;AlbertHofman,PhD32;BridgetKnight,PhD1;PenelopeA.Lind,PhD20;MarkI.

McCarthy,MD,PhD14,15,33;GeorgeMcMahon,PhD3;SarahE.Medland,PhD20;MadsMelbyeMD,

DMSc7,34;AndrewP.Morris,PhD15,35;MichaelNodzenski,MS8;ChristophReichetzeder,MD36,37;

SusanM.Ring,PhD3,12;SylvainSebert,PhD19,38;VerenaSengpiel,PhD39;ThorkildI.A.Sørensen,

MD12,40,41;GonnekeWillemsen,PhD21,22;EcoJ.C.deGeus,PhD21,22;NicholasG.Martin,PhD20;Tim

D.Spector,MD9;ChristinePower,PhD11;Marjo-RiittaJärvelin,MD,PhD19,38,42-44;HansBisgaard,MD,

DMSci18;StruanF.A.Grant,PhD17,29,30;EllenA.Nohr,PhD45;VincentW.Jaddoe,PhD4,32,46;Bo

Jacobsson,MD,PhD16,39;JeffreyC.Murray,MD47;BertholdHocher,MD,PhD36,48;AndrewT.

Hattersley,DM1;DeniseM.Scholtens,PhD8;GeorgeDaveySmith,DSc3,12;Marie-FranceHivert,

MD49-51;JanineF.Felix,PhD4,32,46;ElinaHyppönen,PhD11,52,53;WilliamL.Lowe,Jr.,MD13;TimothyM.

Frayling,PhD1*;DebbieA.Lawlor,PhD3,12*;andRachelM.Freathy,PhD1,12*fortheEarlyGrowth

Genetics(EGG)Consortium

1. InstituteofBiomedicalandClinicalScience,UniversityofExeterMedicalSchool,RoyalDevonand

ExeterHospital,BarrackRoad,Exeter,EX25DW,UK.2. EuropeanCentreforEnvironmentandHumanHealth,UniversityofExeter,TheKnowledgeSpa,

Truro,TR13HD.3. SchoolofSocialandCommunityMedicine,UniversityofBristol,OakfieldHouse,OakfieldGrove,

Bristol,BS82BN,UK.4. TheGenerationRStudyGroup,ErasmusMC,UniversityMedicalCenterRotterdam,P.O.Box2040,

3000CA,Rotterdam,theNetherlands.5. DivisionofHealthSciences,WarwickMedicalSchool,UniversityofWarwick,Coventry,UK.6. DepartmentofMathematicsandStatistics,LancasterUniversity,Lancaster,UK.7. DepartmentofEpidemiologyResearch,StatensSerumInstitut,Copenhagen,Denmark.

2

8. DepartmentofPreventiveMedicine,NorthwesternUniversityFeinbergSchoolofMedicine.9. DepartmentofTwinResearch,King'sCollegeLondon,St.Thomas'Hospital,London,UK.10. CentreforEnvironmentalandPreventiveMedicine,WolfsonInstituteofPreventiveMedicine,

BartsandtheLondonSchoolofMedicineandDentistry,QueenMaryUniversityofLondon.11. Population,PolicyandPractice,UCLInstituteofChildHealth,UniversityCollegeLondon,UK.12. MedicalResearchCouncilIntegrativeEpidemiologyUnitattheUniversityofBristol,UK.13. DivisionofEndocrinology,MetabolismandMolecularMedicine,NorthwesternUniversity

FeinbergSchoolofMedicine14. OxfordCentreforDiabetes,EndocrinologyandMetabolism,UniversityofOxford,UK.15. WellcomeTrustCentreforHumanGenetics,UniversityofOxford,Oxford,UK.16. DepartmentofGenesandEnvironment,DivisionofEpidemiology,NorwegianInstituteofPublic

Health,Oslo,Norway.17. CenterforAppliedGenomics,TheChildren’sHospitalofPhiladelphia,Philadelphia,Pennsylvania,

USA.18. CopenhagenProspectiveStudiesonAsthmainChildhood(COPSAC),FacultyofHealthSciences,

UniversityofCopenhagen,Copenhagen,Denmark&DanishPediatricAsthmaCenter,CopenhagenUniversityHospital,Gentofte,Denmark.

19. InstituteofHealthSciences,UniversityofOulu,Oulu,Finland20. QIMRBerghoferMedicalResearchInstitute,LockedBag2000,RoyalBrisbaneHospital,Herston,

Qld4029,Australia.21. EMGOInstituteforHealthandCareResearch,VUUniversityMedicalCenter,Amsterdam,The

Netherlands.22. DepartmentofBiologicalPsychology,VUUniversityAmsterdam,VanderBoechorststraat1,1081

BTAmsterdam,TheNetherlands.23. DepartmentofMathematics,UniversitedeSherbrooke,QC,Canada.24. CentrederechercheduCentreHospitalierUniversitairedeSherbrooke,Sherbrooke,QC,Canada.25. ECOGENE-21andLipidClinic,ChicoutimiHospital,Saguenay,QC,Canada.26. DepartmentofBiochemistry,UniversitédeSherbrooke,Sherbrooke,QC,Canada.27. DepartmentofPrimaryCareandPublicHealth,ImperialCollegeLondon.28. UniversityofQueenslandDiamantinaInstitute,TranslationalResearchInstitute,Brisbane,

Queensland,Australia.29. DivisionofHumanGenetics,TheChildren’sHospitalofPhiladelphia,Philadelphia,Pennsylvania,

USA.30. DepartmentofPediatrics,PerelmanSchoolofMedicine,UniversityofPennsylvania,Philadelphia,

Pennsylvania,USA.31. FIMMInstituteforMolecularMedicineFinland,HelsinkiUniversityHelsinki,FI-00014,Finland.32. DepartmentofEpidemiology,ErasmusMC,UniversityMedicalCenterRotterdam,P.O.Box2040,

3000CA,Rotterdam,theNetherlands.33. OxfordNationalInstituteforHealthResearch(NIHR)BiomedicalResearchCentre,Churchill

Hospital,Oxford,UK.34. DepartmentofMedicine,StanfordUniversitySchoolofMedicine,Stanford,California,USA.35. DepartmentofBiostatistics,UniversityofLiverpool,LiverpoolL693GA,UK.36. InstituteofNutritionalScience,UniversityofPotsdam,Germany37. CenterforCardiovascularResearch/Charité,Berlin,Germany.38. DepartmentofEpidemiologyandBiostatistics,SchoolofPublicHealth,MedicalResearch

Council-HealthProtectionAgencyCentreforEnvironmentandHealth,FacultyofMedicine,ImperialCollegeLondon,UK

39. DepartmentofObstetricsandGynecology,SahlgrenskaAcademy,SahgrenskaUniversityHospital,Gothenburg,Sweden.

40. InstituteofPreventiveMedicine,BispebjergandFrederiksbergUniversityHospital,CapitalRegion,Copenhagen,Denmark

41. NovoNordiskFoundationCenterforBasicMetabolicResearchandDepartmentofPublicHealth,FacultyofHealthandMedicalSciences,UniversityofCopenhagen,Copenhagen,Denmark

42. DepartmentofChildrenandYoungPeopleandFamilies,NationalInstituteforHealthandWelfare,Aapistie1,Box310,FI-90101Oulu,Finland.

43. BiocenterOulu,UniversityofOulu,Oulu,Finland.

3

44. UnitofPrimaryCare,OuluUniversityHospital,Kajaanintie50,P.O.Box20,FI-90220Oulu,90029OYS,Finland.

45. ResearchUnitofObstetrics&Gynecology,InstituteofClinicalResearch,UniversityofSouthernDenmark,Odense,Denmark.

46. DepartmentofPediatrics,ErasmusMC,UniversityMedicalCenterRotterdam,P.O.Box2040,3000CA,Rotterdam,theNetherlands.

47. DepartmentofPediatrics,UniversityofIowa,IowaCity,Iowa,USA.48. TheFirstAffiliatedHospitalofJinanUniversity,Guangzhou,510630,China.49. DepartmentofPopulationMedicine,HarvardPilgrimHealthCareInstitute,HarvardMedical

School,Boston,MA.50. DiabetesCenter,MassachussettsGeneralHospital,Boston,MA.51. DepartmentofMedicine,UniversitedeSherbrooke,QC,Canada.52. CentreforPopulationHealthResearch,SchoolofHealthSciences,andSansomInstitute,

UniversityofSouthAustralia,Adelaide,Australia.53. SouthAustralianHealthandMedicalResearchInstitute,Adelaide,Australia.

*Theseauthorsjointlydirectedthiswork

Correspondingauthors:Dr.RachelM.FreathyUniversityofExeterMedicalSchoolRoyalDevonandExeterHospitalBarrackRoadExeter,UKTel:+44(0)1392408238Email:r.freathy@ex.ac.ukProf.DebbieA.LawlorMRCIntegrativeEpidemiologyUnitattheUniversityofBristolOakfieldHouse,OakfieldRoad,Bristol,UKTel:+44(0)1173310096Email:d.a.lawlor@bristol.ac.ukProf.TimothyM.FraylingUniversityofExeterMedicalSchoolRoyalDevonandExeterHospitalBarrackRoadExeter,UKTel:+44(0)1392408256Email:t.m.frayling@ex.ac.uk

4

Structuredabstract

Importance:Neonatesborntooverweight/obesewomenarelargerandathigherriskofbirth

complications.Manymaternalobesity-relatedtraitsareobservationallyassociatedwithbirth

weight,butthecausalnatureoftheseassociationsisuncertain.

Objective:Totestforgeneticevidenceofcausalassociationsofmaternalbodymassindex(BMI)and

relatedtraitswithbirthweight.

Design,SettingandParticipants:WeusedMendelianrandomizationtotestwhethermaternalBMI

andobesity-relatedtraitsarecausallyrelatedtooffspringbirthweight.Mendelianrandomization

makesuseofthefactthatgenotypesarerandomlydeterminedatconceptionandarethusnot

confoundedbynon-geneticfactors.Datawereanalysedon30,487womenfrom18studies.

ParticipantswereofEuropeanancestryfrompopulation-orcommunity-basedstudieslocatedin

Europe,NorthAmericaorAustraliaandparticipatingintheEarlyGrowthGenetics(EGG)

Consortium.Live,term,singletonoffspringbornbetween1929and2013wereincluded.Wetested

associationsbetweenageneticscoreof30BMI-associatedsinglenucleotidepolymorphisms(SNPs)

and(i)maternalBMIand(ii)birthweight,toestimatethecausalrelationshipbetweenBMIandbirth

weight.Analyseswererepeatedforotherobesity-relatedtraits.

Exposures:GeneticscoresforBMI,fastingglucoselevel,type2diabetes,systolicbloodpressure

(SBP),triglyceridelevel,HDL-cholesterollevel,vitaminDstatusandadiponectinlevel.

MainOutcome(s)andMeasure(s):Offspringbirthweightmeasuredbytrainedstudypersonnel(n=2

studies),frommedicalrecords(n=10studies)orfrommaternalreport(n=6studies).

Results:Amongthe30,487newbornsthemeanbirthweightinthevariouscohortsrangedfrom

3325gto3679g.ThegeneticscoreforBMIwasassociatedwitha2g(95%CI:0,3g)higheroffspring

birthweightpermaternalBMI-raisingallele(P=0.008).Thematernalgeneticscoresforfasting

glucoseandSBPwerealsoassociatedwithbirthweightwitheffectsizesof8g(95%CI:6,10g)per

5

glucose-raisingallele(P=7x10-14)and-4g(95%CI:-6,-2g)perSBP-raisingallele(P=1x10-5),

respectively.A1standarddeviation(1SD≈4kg/m2)geneticallyhighermaternalBMIwasassociated

witha55g(95%CI:17,93g)higherbirthweight.A1-SDgeneticallyhighermaternalfastingglucose

(≈0.4mmol/L)orSBP(10mmHg)wereassociatedwitha114g(95%CI:80,147g)higheror-208g(95%

CI:-394,-21g)lowerbirthweight,respectively.ForBMIandfastingglucosethesegenetic

associationswereconsistentwiththeobservationalassociations,butforSBP,thegeneticand

observationalassociationswereinoppositedirections.

ConclusionsandRelevance:InthisMendelianrandomizationstudyofmorethan30,000women

withsingletonoffspringfrom18studies,geneticallyelevatedmaternalBMIandbloodglucoselevels

werepotentiallycausallyassociatedwithhigheroffspringbirthweight,whereasgeneticallyelevated

maternalsystolicbloodpressurewasshowntobepotentiallycausallyrelatedtolowerbirthweight.

Ifreplicated,thesefindingsmayhaveimplicationsforcounselingandmanagingpregnanciestoavoid

adverseweight-relatedbirthoutcomes.

6

Introduction

Neonatesborntooverweightorobesewomenaremorelikelytobelargeforgestationalage.1The

precisemechanismsunderlyingthisassociationandtheextenttowhichconfoundingfactors

contributearepoorlyunderstood.Itisimportanttounderstandwhichmaternaltraitsarecausally

associatedwithbirthweightbecausethismay(i)facilitatetargeteddevelopmentofinterventionsto

betestedinrandomizedcontrolledtrials,and(ii)enableclear,evidence-basedrecommendationsin

pregnancy.

Maternaloverweightandobesityarekeyriskfactorsforgestationaldiabetes.2Evenintheabsence

ofdiabetes,obesewomenhavehigherglucoselevelsthannormalweightwomen,despitea

controlleddiet.3Theassociationbetweengestationaldiabetesandhigherbirthweightiswell

documented4,andmaternalglucoselevelsbelowthosediagnosticofdiabetesalsoshowstrong

associationswithbirthweight.5

Thefetusofanoverweightorobesewomanmaybeexposedtotheconsequencesofhigher

maternaltriglyceridelevelsandbloodpressure,lowerlevelsofHDL-cholesterol(HDLc)and

adiponectinandlowervitaminDstatus1,6,7(Box1).Thesematernalobesity-relatedtraitshavebeen

variablyassociatedwithbirthweightinobservationalstudies:highertriglyceridesandlowerHDLc

withhigherbirthweight8,9;higherbloodpressurewithlowerbirthweight10;lowervitaminDstatus

withlowerbirthweight11;andloweradiponectinwithhigherbirthweight12.However,associations

arenotalwaysconsistentlyobservedandmaybeconfounded,forexamplebymaternal

socioeconomicstatusandassociatedbehaviourssuchassmokinganddiet.Furthermore,thehigh

inter-correlationofobesity-relatedtraitscomplicatesdeterminationofcausalrelationshipsinan

observationalsetting.

7

MaternalgenotypesmaybeusedinaMendelianrandomization13,14approachtoprovideevidenceof

apotentialcausalassociationbetweenmaternaltraitsandbirthoutcomes(Figure1).Mendelian

randomizationisanalogoustoarandomizedcontrolledtrial:genotypes,whicharerandomly

allocatedatconception,arelargelyfreefromconfoundingandcanbeusedtoestimatethepossible

causaleffectsofmaternaltraits.Inthisstudy,geneticvariantswereselectedtocalculategenetic

scoresrepresentingmaternalBMIandeachof7obesity-relatedmaternaltraits.Thepotentialcausal

relationshipbetweenmaternalBMIandeachrelatedtraitwasestimatedbytestingassociations

betweenmaternalgeneticriskscoresandoffspringbirthweights.

8

Methods

Studyparticipants

Singlenucleotidepolymorphism(SNP)genotypedatawereusedfromatotalof30,487womenfrom

18population-orcommunity-basedstudieslocatedinEurope,NorthAmericaorAustralia.Thebirth

weightofonechildpermotherwasincluded(seeeTable1forfulldetailsofparticipant

characteristicsandeTable2forgenotypinginformation).Birthweightwasmeasuredbytrained

studypersonnel(n=2studies),frommedicalrecords(n=10studies)orfrommaternalreport(n=6

studies).Theoffspringyearsofbirthwerefrom1929to2013.Multiplebirths,stillbirths,congenital

anomalies,birthsbefore37weeksgestationandindividualsofnon-Europeanancestrywere

excluded.Informedconsentwasobtainedfromallparticipants,andstudyprotocolswereapproved

bythelocalregionalorinstitutionalethicscommittees.

Selectionofmaternalobesity-relatedtraitsandSNPs

InadditiontoBMI,traitswereselectedthatareassociatedwithmaternalobesityandmayaffect

fetalgrowththroughtheintrauterineenvironment.Theireffectsweremodelledinthedirections

hypothesisedbytheirrelationshipstomaternalBMI(Box1)

SNPsknowntoberobustlyassociated(P<5x10-8)withBMIandeachobesity-relatedtraitwere

selected.FulldetailsoftheselectedSNPsareprovidedineTable3.SNPsassociatedwith(i)fasting

glucoseand(ii)type2diabeteswereusedtorepresentmaternalglycemia.TheType2diabetesSNPs

wereconsideredtorepresentexposuretomaternaldiabetesinpregnancy,includinggestational

diabetes,givenoverlapbetweentype2andgestationaldiabetesgeneticsusceptibilityvariants.15For

bloodpressure,SNPswereselectedthatareprimarilyassociatedwithsystolicbloodpressure(SBP),

thoughallalsoshowstrongevidenceofassociationwithdiastolicbloodpressure.ForvitaminD

status,twoSNPswithhypothesisedrolesinvitaminDsynthesiswereusedtorepresent25(OH)D

9

levels(anindicatorofoverallvitaminDstatus),aspreviouslyrecommended.16,17Furtherdetailsof

SNPselectionareprovidedintheeMethods.

Aweightedgeneticscorewascalculatedforeachmaternaltrait(seeeMethodsforfulldetails).Very

fewoftheselectedSNPshavebeentestedinpregnancy.Geneticscoreswerevalidatedby

confirmingthateachwasassociatedwithitsrespectivematernaltrait,measuredduringpregnancy

(withtheexceptionofBMI,forwhichthepre-pregnancyvaluewasused).Maternalpre-pregnancy

BMIwasavailablefromregistrydata(N=2studies)orcalculatedfromself-reportedweightand

height(N=3studies).IntheAvonLongitudinalStudyofParentsandChildren(ALSPAC)study,the

self-reportwasvalidatedwithaclinicmeasure18.Detailsoftraitsmeasuredinpregnancyandtheir

sourcesaregivenineTable4.Ineachavailablestudy,linearregressionofthematernaltrait(e.g.

BMI)againstthegeneticscorewasperformed,adjustingformaternalage.Toconfirmthat

associationsbetweeneachgeneticscoreanditsrespectivematernaltraitweresimilarinthesame

individualsduringandafterpregnancy,availabledatawereusedfromtwolongitudinalstudies(the

AvonLongitudinalStudyofParentsandChildren[ALSPAC]andtheExeterFamilyStudyofChildhood

Health[EFSOCH]).TocheckthatthestrategyforSNPselectionhadresultedingeneticscoresthat

werespecifictoeachmaternaltrait,theassociationwastestedbetweeneachofthe8geneticscores

andtheothermaternaltraits,andindicatorsofmaternalsocio-economicstatusandsmoking.

TestingthehypothesisthatmaternalBMIandobesity-relatedtraitsareassociatedwithbirth

weightthroughtheintra-uterineenvironment.

ForBMIandeachrelatedmaternaltrait,twoMendelianrandomizationapproacheswereusedto

testthehypothesis.First,associationsweretestedbetweengeneticscoresrepresentingmaternal

traitsandoffspringbirthweightusingthemaximumnumberofparticipants(i.e.foreachtrait,those

withgeneticscoreandoffspringbirthweightdataavailable,irrespectiveofwhethertheyhadthe

maternaltraitmeasured).Anassociationofthegeneticscorewithbirthweightwouldsupporta

10

possiblecausalrelationshipbetweenthetrait(e.g.pre-pregnancyBMI)andbirthweight,butwould

notprovideinformationonthesizeofthatassociation.Second,weperformedanalysesinthosewith

themeasuredtraitthatenabledanestimateofthesizeofapossiblecausalrelationship.The

analysestookintoaccounttheassociationbetweeneachgeneticscoreandthematernaltraitit

represented(e.g.BMI),inadditiontotheassociationbetweenthesamegeneticscoreandbirth

weight.Thesetworesultswereusedtocalculateanassociationbetweenthematernaltrait(e.g.

BMI)andbirthweightthatwasfreefromconfounding.Thissecondapproachmeasuresthe

relationshipbetweenvariationinmaternalBMI(orBMI-relatedtrait)andbirthweightthatis

attributableonlytogeneticfactors(seeFigure1foranexplanationofthemethod).Foreach

approachmeta-analysiswasusedtocombinedatafromindividualstudies(seeeMethods).

Usingthefirstapproach,weinvestigatedtheassociationbetweeneachgeneticscoreand(i)birth

weightand(ii)ponderalindex(anindexofneonatalleanness,measuredinkg/m3).Withineach

study,birthweightorponderalindexZ-scoreswereregressedagainsteachmaternalgeneticscore,

adjustedforoffspringsexandgestationalage.Analysesusingthetype2diabetesgeneticscorewere

repeatedafterexcludingparticipantswithpre-existingandgestationaldiabetes.Analysesusingthe

SBPgeneticscorewererepeatedafterexcludingparticipantswithpre-eclampsiaandexistingor

gestationalhypertension.

Thegeneticestimateoftheassociationbetweeneachmaternaltraitandbirthweight/ponderal

indexfromthesecondapproachwascomparedwiththecorrespondingobservationalassociation.

Toobtaintheobservationalestimateslinearregressionwasperformedusingbirthweightor

ponderalindexasthedependentvariable,andeachof7maternaltraitsasindependentvariables,

adjustingforsexandgestationalage.Therewasinsufficientinformationonmaternaltype2diabetes

prevalence,soitwasnotpossibletoestimatethecausalrelationshipforthattrait.Fulldetailsofthe

analysisareprovidedintheeMethods.

11

EstimatinghowmuchoftheassociationbetweenmaternalBMIandbirthweightismediatedby

fastingglucose

Availabledatawereusedtoestimatetheapproximatecausalrelationshipbetweena1SD(≈4kg/m2)

highermaternalBMIand(i)fastingglucoseand(ii)SBP.Usingeachofthoseestimates,theresultsof

theMendelianrandomizationanalyseswererescaledtorepresenttheeffectsoffastingglucoseand

SBPthatcouldbedirectlycomparedwiththecausalrelationshipbetweena1SDhigherBMIand

birthweight(seeeMethodsforadetaileddescriptionofthemethod).

Correctingfordirectfetalgenotypeeffects

Genotypesofmaternal-fetalpairswereavailableinupto8studies(N=upto11,494).Analyseswere

repeatedincludingthefetalgenotypeateachSNPinthemodel,tocorrectforpotentialconfounding

causedbydirecteffectsofthefetalgenotype.

12

Results

Thecharacteristicsofincludedparticipantsfromthe18contributingstudiesareshowninTable1.

Amongthe30,487newbornsthemeanbirthweightrangedfrom3325gto3679g.Themeanpre-

pregnancyBMIwasavailablein11studiesandrangedfrom22.78kg/m2to24.83kg/m2.Themean

maternalageatdelivery,availablein16studies,rangedfrom24.5yearsto31.5years.

Therewasevidenceofassociationbetweeneachgeneticscoreanditscorrespondingmaternaltrait

measuredinpregnancy(P≤0.003;Table2).ForBMI,fastingglucoseandSBP,datafrommultiple

studiesweremeta-analysed,withsimilareffectestimatesbetweenstudiesforBMIandfasting

glucose(Phet>0.05)andevidenceofheterogeneityforSBP(Phet=0.04).Theeffectsizesofassociations

betweenmaternaltraitsandtheirrespectivegeneticscoreswereverysimilarwhencomparedinthe

sameindividualsduringandoutsidepregnancy,withtheexceptionoftheSBPgeneticscorewhich

hadaweakereffectduringpregnancy(eTable5).Therewasnoevidenceofassociationbetweenany

geneticscoreandpotentiallyconfoundingvariables.Noindividualgeneticscorewasassociatedwith

anyoftheothermaternaltraits,exceptforthegeneticscoreforBMI,whichwaspositively

associatedwithSBP(P<0.003Bonferroni-correctedfor15tests;eTable6).

GeneticevidenceforapossiblecausalassociationbetweenhighermaternalBMIandhigherbirth

weight

ThematernalBMIgeneticscorewasassociatedwithhigherbirthweight(Table3)andponderal

index(eTable7)withsimilareffectsizesbeforeandafteradjustingforpossibleeffectsoffetal

genotype.Usingthegeneticscoretoquantifythepossiblecausalassociation,a1SDgenetically

highermaternalBMI(equivalentto4kg/m2)wasassociatedwitha55g(95%CI:17,93)higher

offspringbirthweight.Afteradjustingforfetalgenotype,theestimatedeffectwas104g(95%CI:32,

176)(Table4).TheseMendelianrandomizationcausalestimatesweresimilartotheobservational

13

associationof62g(95%CI:56,70)per1SD(4kg/m2)highermaternalBMI(Figure2).Similarresults

wereobtainedforponderalindex(eTable8andeFigure1).

Geneticevidenceforapossiblecausalassociationbetweenhighermaternalfastingglucoseand

higherbirthweight,butnoassociationwithmaternallipidsoradiponectin

Thematernalfastingglucoseandtype2diabetesgeneticscoreswereassociatedwithhigherbirth

weight(Table3)andponderalindex(eTable7)withsimilareffectsizeestimatesbeforeandafter

adjustingforfetalgenotype,andbeforeandafterexcludingpre-existingandgestationaldiabetes.

Usingthegeneticscoretoestimatethepossiblecausalrelationship,a1SD(0.4mmol/L)genetically

highermaternalglucosewasassociatedwitha114g(95%CI:80,147)higherbirthweight.After

adjustingforfetalgenotype,theassociationwas145g(95%CI:91,199)(Table4).Thesegenetic

estimatesweresimilartotheobservationalassociationof92g(95%CI:80,104)per1SD(0.4mmol/L)

highermaternalglucose(Figure2).Similarresultswereobtainedforponderalindex(eTable8and

eFigure1).

Thematernaltriglyceridegeneticscorewasnotassociatedwithoffspringbirthweight(Table3)or

ponderalindex(eTable7).Usingthegeneticscoretoestimatethepossiblecausalrelationship,a

geneticallyhighermaternaltriglyceridelevelwasnotassociatedwithoffspringbirthweightandthe

95%confidenceintervalsaroundthegeneticestimateexcludedtheobservationalassociation

betweenmaternaltriglyceridesandbirthweight(P=0.007testingdifferencebetweengeneticand

observationalassociation;Table4;Figure2).Likewise,thegeneticestimateofthepossibleeffectof

maternaladiponectinlevelsonoffspringbirthweightwasdifferentfromtheobservational

association(P=0.002).ThegeneticscoreforHDLcwasnotassociatedwithbirthweightorponderal

indexandtheanalysiswasconsistentwithnocausalrelationship,howeverthiscouldnotbe

distinguishedfromthenegativeobservationalassociationbetweenmaternalHDLcandbirthweight.

14

Geneticevidenceforapossiblecausalassociationbetweenhighersystolicbloodpressureand

lowerbirthweight

ThematernalSBPgeneticscorewasassociatedwithlowerbirthweight(Table3)andponderalindex

(eTable7)withsimilareffectsizeestimatesbeforeandafteradjustingforfetalgenotype,andbefore

andafterexcludingmaternalpre-eclampsiaandhypertension.Usingthegeneticscoretoestimate

thepossiblecausalrelationship,a1SD(10mmHg)geneticallyhighermaternalSBPwasassociated

witha-208g(95%CI;-394,-21)loweroffspringbirthweight.Afteradjustingforfetalgenotype,the

estimatedeffectwas-151g(95%CI:-390,89)(Table4).Thegeneticestimateoftheassociation

betweenmaternalSBPandbirthweightinthefullsampleofwomenwasintheoppositedirectionto

theobservationalassociation(P=0.01fordifferencebetweengeneticandobservationalassociations;

Table4;Figure2).Similarresultswereobtainedforponderalindex(eTable8andeFigure1).

ThematernalgeneticscoreforlowervitaminDstatuswasassociatedwithlowerbirthweight

(P=0.03;Table3).However,theestimatedcausalrelationshipwasnotsignificantlydifferentfrom

zero(theestimatedchangeinbirthweightfora10%geneticallylowermaternal25[OH]Dlevelwas-

26g(95%CI:-54,2);Table4,Figure2).

Associationsbetweenthegeneticscoresandbirthweightwereconsistentacrossstudies

Associationsbetweenmaternalgeneticscoresandoffspringbirthweightweresimilarbetween

studiesinthemeta-analysis(Table3;Phet>0.05).Wheredatawerecombinedfromobservational

analyses,theassociationsbetweenmaternalfastingglucoseorSBPandbirthweightweresimilar

(Phet>0.05),andtherewasevidenceofheterogeneityfortheBMI-birthweightobservational

association(Table4;Phet=0.03).

Exposureofthefetustohighermaternalfastingglucoseisunlikelytoexplainalloftheassociation

betweenhighermaternalBMIandhigheroffspringbirthweight

15

ToestimatehowmuchoftheassociationbetweenmaternalBMIandbirthweightmightbe

mediatedbyfastingglucose,theBMIandfastingglucosegeneticscoreswereused:a1SD(≈4

kg/m2)geneticallyhighermaternalBMIwasassociatedwitha0.34SD(≈0.14mmol/L)higher

maternalfastingglucose.FromtheMendelianrandomizationanalyses,1SD(≈0.4mmol/L)

geneticallyhighermaternalfastingglucosewasassociatedwitha114g(95%CI:80,147)higherbirth

weight.Consequently,itwaspredictedthata0.34SDhigherfastingglucosewouldbeassociated

witha114g×0.34=39g[95%CI:27,50]higherbirthweight.Thisapproximationisbroadlysimilarto

thetotalestimatedeffectofa1SDhigherBMIonbirthweight(55g[95%CI:17,93]).However,using

thesamemethodwiththeBMIandSBPgeneticscoresweestimatedthata1SDhighermaternalBMI

wouldbeassociatedwitha-40g[95%CI:-75,-4]lowerbirthweightviaitsassociationwithmaternal

SBP(eFigure2),whichwouldopposethepositiveassociationwithmaternalfastingglucose.

16

Discussion

ThisstudyprovidesevidenceforapossiblecausalassociationbetweenmaternalBMIandoffspring

birthweight.A4kg/m2geneticallyhighermaternalBMI(a1SDrise)wasassociatedwitha55g(95%

CI:17,93)higheroffspringbirthweight.Inaddition,a0.4mmol/l(1SD)geneticallyhighercirculating

maternalfastingglucosewasassociatedwitha114g(95%CI:80,147)higherbirthweight,whilea10

mmHggeneticallyhighermaternalSBPwasassociatedwitha-208g(95%CI:-394,-21)lowerbirth

weight.TheseresultsprovideevidencethatgeneticallyelevatedmaternalglucoseandSBPhave

directionallyoppositecausalassociationswithbirthweight.Theestimatedassociationsbetween

thesematernaltraitsandbirthweight(eitherincreasedorreduced)aresubstantialandofclinical

importance.Theysupporteffortstomaintainhealthygestationalglucoseandbloodpressurelevels

toensurehealthyfetalgrowth.ThepositiveassociationbetweenmaternalBMIandbirthweight

maybepartiallymediatedbytheeffectofhigherBMIoncirculatingmaternalfastingglucose.There

wasnoevidenceofassociationwithageneticscoreformaternaltriglycerides,whichhavealsobeen

hypothesisedtobeimportantcontributorstohigherbirthweightinoverweightorobesewomen.

Otherlipids,orspecificsubclassesoftriglycerides,mightbeimportantbutrequirefurtherstudy.

Theseresultsprovidegeneticevidenceofacausalassociationbetweenmaternalglycemiaandbirth

weightandponderalindex,eveninwomenwithnopre-existingorgestationaldiabetes,whichis

consistentwithpublishedobservationaldata.5Apossibleexplanationforthisfindingisthatwomen

withahighergeneticscorefortype2diabeteshaverelativelyhigherglucoselevelsinpregnancy,as

aresultofinadequatebetacellcompensationinresponsetogestationalinsulinresistance,19,20

leadingtoincreasedplacentalglucosetransferandfetalinsulinsecretion,21andconsequentlyhigher

birthweight.

Thesedatadidnotsupportacausalassociationbetweenmaternaltriglyceride,HDLcoradiponectin

levelsandbirthweightorponderalindex.Thegeneticassociationsbetweenmaternaltriglycerides

17

andadiponectinandbirthweightwerenull,incontrasttotheobservationalassociations,suggesting

thattheobservationalassociationsseenhere,andinotherpublishedstudies8,9,12,areconfounded.

TheMendelianrandomizationanalysisshowedthatthepositiveobservationalassociationbetween

SBPandbirthweightisconfounded,mostlikelybyBMI,whichisbothanimportantriskfactorfor

higherSBPinpregnancyandpositivelyassociatedwithbirthweight.1Usinggeneticvariantsthatare

independentofconfoundingbyBMI,itwasdemonstratedthatgeneticallyhighermaternalSBPis

associatedwithlowerbirthweight,evenafterexcludingpre-eclampsiaandhypertension.The

precisionofourestimateofthechangeinbirthweightper1SDinmaternalSBPcouldbeaffectedby

theheterogeneitybetweenstudiesinthegeneticscore-SBPassociation(P=0.04,I2=76.0%;Table2).

However,associationsbetweentheSBPgeneticscoreandbirthweightwereconsistentacrossall13

meta-analyzedstudies(P=0.14,I2=30.4%;Table3)andsupportiveofacausalassociationbetween

highermaternalSBPandlowerbirthweight.Thesefindingssupportobservationalassociations

betweenmaternalSBPandbirthweightthatwereadjustedforawiderangeofconfounders,22and

areconsistentwithlaboratoryandpopulationstudiessuggestingalinkbetweenhypertensive

disordersofpregnancyandimpairedfetalgrowthduetoplacentalpathology.23Thereareincreasing

concernsabouttheeffecttheobesityepidemicmighthaveonbirthsize,viagreatermaternalBMI.

However,thefocusofthatconcernhasbeenlargelyonincreasedbirthsizeasaresultofgreater

maternalglucoseandotherfetalnutrients.Ourfindingssuggestthatthereareopposingeffectsof

maternalbloodpressureandglucose.

PublishedMendelianrandomizationanalysesprovideevidencethathigherBMIiscausallyassociated

withlowervitaminDstatus,6andevidencefrommultipleobservationalstudiessuggeststhatlower

maternalvitaminDisassociatedwithlowerbirthweight.11,24OuranalysisofthevitaminDgenetic

scoreprovidedsomeevidencetosupportapossiblecausalassociationwithbirthweight,butthis

requiresfurtherexplorationinlargernumbersofpregnancies.

18

Socio-economicfactorsandrelatedbehaviourssuchassmokingarekeyconfoundersof

observationalassociationsbetweenmaternalBMI(orBMI-relatedtraits)andoffspringbirthweight,

sincetheyareassociatedwithbothvariables(seeeTable9forademonstrationoftheseassociations

intheALSPACstudy).Thegeneticscoresusedinouranalyseswerenotassociatedwithsocio-

economicfactorsorsmoking,andthisillustratesakeystrengthoftheMendelianrandomization

approach:sincegenotypesaredeterminedatconception,suchconfoundingisavoided.

Therearesomelimitationstothisstudy.Despiteattemptstomaximisespecificityofthegenetic

scores,wecannotfullyexcludethepossibilitythattheselectedgeneticvariantsactonmorethan

onematernaltrait.Althoughallavailableinformationwasused,therewaslimitedpowertodetect

associationsbetweenthegeneticscoresandothertraits.Forexample,theknownassociation

betweenBMI-associatedvariantsandtriglyceridelevelswasnotdetected.25Withthepotentialfor

high-throughputmetabolomicstudiesandagrowingpublicdatabaseofgeneticassociations,26-28

futurestudieswillimprovethespecificity(fordifferentlipidsub-fractions)ofselectedgenetic

variants.

Despitethelargesampleinthisstudy,statisticalpowertodetectcausalrelationshipswaslimitedfor

somematernaltraits(seeeMethodsandeTable10forpowercalculations).Thetotalsample

provided>99%powertodetectassociationsatP<0.05betweenbirthweightandgeneticscoressuch

asfastingglucoseandsystolicbloodpressurethatexplainatleast0.1%varianceinbirthweight.

However,largersamples(N>80,000)willbeneededtoconfidentlydetectorruleout(i)the

associationwithvitaminDstatussuggestedbyourdata,or(ii)smallerpositiveornegativecausal

associationsbetweenmaternaltriglycerides,HDLcoradiponectinandbirthweight.

19

Whileadjustingforthefetalgeneticscoreswasnecessarytoseparatematernaleffectsfromthe

directeffectsofgeneticvariantsinthefetus,thiscouldpotentiallyintroducebiasviaassociation

withpaternalgenotypes.AssortativematingforBMIcouldadditionallyresultinacorrelation

betweenmaternalandpaternalgenotypes,leadingtosimilarbias.However,afather’sgeneticscore

wouldonlyconfoundtheMendelianrandomizationestimatesifthefather’sphenotypewererelated

tobirthweight,andwefoundonlyveryweakassociationsoffathersBMIandrelatedtraitswith

offspringbirthweight(eTable11).Anotherpotentialbiascouldbeinducedbytheuseofthegenetic

scoreforSBP,whichwasderivedfromagenome-wideassociationstudyofbloodpressure

conditionalonBMI.SinceBMIisalsoassociatedwithbirthweight,thiscouldbiastheresults.

However,similarresultswereobtainedusinganalternativegeneticscorethatwasunadjustedfor

BMI(eMethods).

InMendelianrandomizationanalysis,aweakstatisticalassociationbetweenageneticscoreanda

maternaltrait(duetolowvarianceexplainedand/orsmallsamplesize)hasthepotentialtocause

weakinstrumentbiastowardstheobservationalresults.29Theproportionsofmaternaltraitvariance

explainedbythegeneticscoresaremodestinourstudy(Table2).However,thelargeoverallsample

sizeensuredthatthepossiblecausalassociationsidentifiedareunlikelytobeduetoweak

instrumentbias(seeeMethods).

OuranalysesassumethatmaternalBMIandrelatedtraitsarelinearlyassociatedwithoffspringbirth

weight.Wehavenottestedfornon-linearassociationswhich,inaMendelianrandomizationdesign,

wouldrequireverylargenumbers30.However,formaternalBMI,fastingglucoseandsystolicblood

pressure,thereisobservationalevidenceofsuchlinearassociationsacrossthedistribution,withno

evidenceofthresholdorcurvilinearassociations5,10,31.

20

Conclusions

InthisMendelianrandomizationstudyofmorethan30,000womenwithsingletonoffspringfrom18

studies,geneticallyelevatedmaternalBMIandbloodglucoselevelswerepotentiallycausally

associatedwithhigheroffspringbirthweight,whereasgeneticallyelevatedmaternalsystolicblood

pressurewasshowntobepotentiallycausallyrelatedtolowerbirthweight.Ifreplicated,these

findingsmayhaveimplicationsforcounselingandmanagingpregnanciestoavoidadverseweight-

relatedbirthoutcomes.

21

REFERENCES

1. LawlorDA,ReltonC,SattarN,NelsonSM.Maternaladiposity--adeterminantofperinatalandoffspringoutcomes?NatRevEndocrinol.2012;8(11):679-688.

2. ShinD,SongWO.Prepregnancybodymassindexisanindependentriskfactorforgestationalhypertension,gestationaldiabetes,pretermlabor,andsmall-andlarge-for-gestational-ageinfants.JMaternFetalNeonatalMed.2014:1-8.

3. HarmonKA,GerardL,JensenDR,etal.Continuousglucoseprofilesinobeseandnormal-weightpregnantwomenonacontrolleddiet:metabolicdeterminantsoffetalgrowth.DiabetesCare.2011;34(10):2198-2204.

4. LandonMB,MeleL,SpongCY,etal.Therelationshipbetweenmaternalglycemiaandperinataloutcome.ObstetGynecol.2011;117(2Pt1):218-224.

5. MetzgerBE,LoweLP,DyerAR,etal.Hyperglycemiaandadversepregnancyoutcomes.NEnglJMed.2008;358(19):1991-2002.

6. VimaleswaranKS,BerryDJ,LuC,etal.CausalrelationshipbetweenobesityandvitaminDstatus:bi-directionalMendelianrandomizationanalysisofmultiplecohorts.PLoSMed.2013;10(2):e1001383.

7. GaillardR,DurmusB,HofmanA,MackenbachJP,SteegersEA,JaddoeVW.Riskfactorsandoutcomesofmaternalobesityandexcessiveweightgainduringpregnancy.Obesity(SilverSpring).2013;21(5):1046-1055.

8. MisraVK,TrudeauS,PerniU.Maternalserumlipidsduringpregnancyandinfantbirthweight:theinfluenceofprepregnancyBMI.Obesity(SilverSpring).2011;19(7):1476-1481.

9. KulkarniSR,KumaranK,RaoSR,etal.Maternallipidsareasimportantasglucoseforfetalgrowth:findingsfromthePuneMaternalNutritionStudy.DiabetesCare.2013;36(9):2706-2713.

10. Macdonald-WallisC,TillingK,FraserA,NelsonSM,LawlorDA.Associationsofbloodpressurechangeinpregnancywithfetalgrowthandgestationalageatdelivery:findingsfromaprospectivecohort.Hypertension.2014;64(1):36-44.

11. LeffelaarER,VrijkotteTG,vanEijsdenM.MaternalearlypregnancyvitaminDstatusinrelationtofetalandneonatalgrowth:resultsofthemulti-ethnicAmsterdamBornChildrenandtheirDevelopmentcohort.BrJNutr.2010;104(1):108-117.

12. LoweLP,MetzgerBE,LoweWL,Jr.,etal.Inflammatorymediatorsandglucoseinpregnancy:resultsfromasubsetoftheHyperglycemiaandAdversePregnancyOutcome(HAPO)Study.JClinEndocrinolMetab.2010;95(12):5427-5434.

13. DaveySmithG,EbrahimS.'Mendelianrandomization':cangeneticepidemiologycontributetounderstandingenvironmentaldeterminantsofdisease?IntJEpidemiol.2003;32(1):1-22.

14. LawlorDA,HarbordRM,SterneJA,TimpsonN,DaveySmithG.Mendelianrandomization:usinggenesasinstrumentsformakingcausalinferencesinepidemiology.StatMed.2008;27(8):1133-1163.

15. KwakSH,KimSH,ChoYM,etal.Agenome-wideassociationstudyofgestationaldiabetesmellitusinKoreanwomen.Diabetes.2012;61(2):531-541.

16. BerryDJ,VimaleswaranKS,WhittakerJC,HingoraniAD,HypponenE.EvaluationofgeneticmarkersasinstrumentsforMendelianrandomizationstudiesonvitaminD.PLoSOne.2012;7(5):e37465.

17. VimaleswaranKS,CavadinoA,BerryDJ,etal.AssociationofvitaminDstatuswitharterialbloodpressureandhypertensionrisk:amendelianrandomisationstudy.LancetDiabetesEndocrinol.2014;2(9):719-729.

18. LawlorDA,FraserA,LindsayRS,etal.Associationofexistingdiabetes,gestationaldiabetesandglycosuriainpregnancywithmacrosomiaandoffspringbodymassindex,waistandfatmassinlaterchildhood:findingsfromaprospectivepregnancycohort.Diabetologia.2010;53(1):89-97.

22

19. HayesMG,UrbanekM,HivertMF,etal.IdentificationofHKDC1andBACE2asgenesinfluencingglycemictraitsduringpregnancythroughgenome-wideassociationstudies.Diabetes.2013;62(9):3282-3291.

20. FreathyRM,HayesMG,UrbanekM,etal.HyperglycemiaandAdversePregnancyOutcome(HAPO)study:commongeneticvariantsinGCKandTCF7L2areassociatedwithfastingandpostchallengeglucoselevelsinpregnancyandwiththenewconsensusdefinitionofgestationaldiabetesmellitusfromtheInternationalAssociationofDiabetesandPregnancyStudyGroups.Diabetes.2010;59(10):2682-2689.

21. PedersenJ.Thepregnantdiabeticandhernewborn:problemsandmanagement.Baltimore:Williams&Wilkins;1977.

22. LawlorDA,DaveySmithG,EbrahimS.Birthweightofoffspringandinsulinresistanceinlateadulthood:crosssectionalsurvey.BMJ.2002;325(7360):359.

23. Rich-EdwardsJW,FraserA,LawlorDA,CatovJM.Pregnancycharacteristicsandwomen'sfuturecardiovascularhealth:anunderusedopportunitytoimprovewomen'shealth?EpidemiolRev.2014;36(1):57-70.

24. BodnarLM,CatovJM,ZmudaJM,etal.Maternalserum25-hydroxyvitaminDconcentrationsareassociatedwithsmall-for-gestationalagebirthsinwhitewomen.JNutr.2010;140(5):999-1006.

25. FallT,HaggS,MagiR,etal.Theroleofadiposityincardiometabolictraits:aMendelianrandomizationanalysis.PLoSMed.2013;10(6):e1001474.

26. WurtzP,WangQ,KangasAJ,etal.Metabolicsignaturesofadiposityinyoungadults:mendelianrandomizationanalysisandeffectsofweightchange.PLoSMed.2014;11(12):e1001765.

27. ShinSY,FaumanEB,PetersenAK,etal.Anatlasofgeneticinfluencesonhumanbloodmetabolites.NatGenet.2014;46(6):543-550.

28. KettunenJ,TukiainenT,SarinAP,etal.Genome-wideassociationstudyidentifiesmultiplelociinfluencinghumanserummetabolitelevels.NatGenet.2012;44(3):269-276.

29. BurgessS,ThompsonSG,CollaborationCCG.AvoidingbiasfromweakinstrumentsinMendelianrandomizationstudies.IntJEpidemiol.2011;40(3):755-764.

30. SilverwoodRJ,HolmesMV,DaleCE,etal.Testingfornon-linearcausaleffectsusingabinarygenotypeinaMendelianrandomizationstudy:applicationtoalcoholandcardiovasculartraits.IntJEpidemiol.2014;43(6):1781-1790.

31. GroupHSCR.HyperglycaemiaandAdversePregnancyOutcome(HAPO)Study:associationswithmaternalbodymassindex.BJOG.2010;117(5):575-584.

32. CatalanoPM,McIntyreHD,CruickshankJK,etal.Thehyperglycemiaandadversepregnancyoutcomestudy:associationsofGDMandobesitywithpregnancyoutcomes.DiabetesCare.2012;35(4):780-786.

33. FraserA,Macdonald-WallisC,TillingK,etal.CohortProfile:theAvonLongitudinalStudyofParentsandChildren:ALSPACmotherscohort.IntJEpidemiol.2013;42(1):97-110.

34. SchlemmL,HaumannHM,ZiegnerM,etal.Newevidenceforthefetalinsulinhypothesis:fetalangiotensinogenM235Tpolymorphismisassociatedwithbirthweightandelevatedfetaltotalglycatedhemoglobinatbirth.JHypertens.2010;28(4):732-739.

35. PowerC,ElliottJ.Cohortprofile:1958Britishbirthcohort(NationalChildDevelopmentStudy).IntJEpidemiol.2006;35(1):34-41.

36. ZhaoJ,LiM,BradfieldJP,etal.Examinationoftype2diabeteslociimplicatesCDKAL1asabirthweightgene.Diabetes.2009;58(10):2414-2418.

37. BisgaardH.TheCopenhagenProspectiveStudyonAsthmainChildhood(COPSAC):design,rationale,andbaselinedatafromalongitudinalbirthcohortstudy.AnnAllergyAsthmaImmunol.2004;93(4):381-389.

23

38. NohrEA,TimpsonNJ,AndersenCS,DaveySmithG,OlsenJ,SorensenTI.Severeobesityinyoungwomenandreproductivehealth:theDanishNationalBirthCohort.PLoSOne.2009;4(12):e8444.

39. OlsenJ,MelbyeM,OlsenSF,etal.TheDanishNationalBirthCohort--itsbackground,structureandaim.ScandJPublicHealth.2001;29(4):300-307.

40. KnightB,ShieldsBM,HattersleyAT.TheExeterFamilyStudyofChildhoodHealth(EFSOCH):studyprotocolandmethodology.PaediatrPerinatEpidemiol.2006;20(2):172-179.

41. LacroixM,BattistaMC,DoyonM,etal.Loweradiponectinlevelsatfirsttrimesterofpregnancyareassociatedwithincreasedinsulinresistanceandhigherriskofdevelopinggestationaldiabetesmellitus.DiabetesCare.2013;36(6):1577-1583.

42. JaddoeVW,vanDuijnCM,FrancoOH,etal.TheGenerationRStudy:designandcohortupdate2012.EurJEpidemiol.2012;27(9):739-756.

43. MagnusP,IrgensLM,HaugK,etal.Cohortprofile:theNorwegianMotherandChildCohortStudy(MoBa).IntJEpidemiol.2006;35(5):1146-1150.

44. RantakallioP.Groupsatriskinlowbirthweightinfantsandperinatalmortality.ActaPaediatrScand.1969;193:Suppl193:191+.

45. BoomsmaDI,deGeusEJ,VinkJM,etal.NetherlandsTwinRegister:fromtwinstotwinfamilies.TwinResHumGenet.2006;9(6):849-857.

46. MedlandSE,NyholtDR,PainterJN,etal.CommonvariantsinthetrichohyalingeneareassociatedwithstraighthairinEuropeans.AmJHumGenet.2009;85(5):750-755.

47. MoayyeriA,HammondCJ,ValdesAM,SpectorTD.CohortProfile:TwinsUKandhealthyageingtwinstudy.IntJEpidemiol.2013;42(1):76-85.

48. MoayyeriA,HammondCJ,HartDJ,SpectorTD.TheUKAdultTwinRegistry(TwinsUKResource).TwinResHumGenet.2013;16(1):144-149.

49. SpeliotesEK,WillerCJ,BerndtSI,etal.Associationanalysesof249,796individualsreveal18newlociassociatedwithbodymassindex.NatGenet.2010;42(11):937-948.

50. DupuisJ,LangenbergC,ProkopenkoI,etal.Newgeneticlociimplicatedinfastingglucosehomeostasisandtheirimpactontype2diabetesrisk.Nat.Genet.2010;42(2):105-116.

51. MorrisAP,VoightBF,TeslovichTM,etal.Large-scaleassociationanalysisprovidesinsightsintothegeneticarchitectureandpathophysiologyoftype2diabetes.NatGenet.2012;44(9):981-990.

52. TeslovichTM,MusunuruK,SmithAV,etal.Biological,clinicalandpopulationrelevanceof95lociforbloodlipids.Nature.2010;466(7307):707-713.

53. EhretGB.Genome-wideassociationstudies:contributionofgenomicstounderstandingbloodpressureandessentialhypertension.CurrHypertensRep.2010;12(1):17-25.

54. YaghootkarH,LaminaC,ScottRA,etal.Mendelianrandomizationstudiesdonotsupportacausalroleforreducedcirculatingadiponectinlevelsininsulinresistanceandtype2diabetes.Diabetes.2013;62(10):3589-3598.

55. FreathyRM,Mook-KanamoriDO,SovioU,etal.VariantsinADCY5andnearCCNL1areassociatedwithfetalgrowthandbirthweight.Nat.Genet.2010;42(5):430-435.

24

ARTICLEINFORMATION

AuthorContributionsDr.FreathyandProfsLawlorandFraylinghadfullaccesstoallofthedataintheALSPAC,EFSOCHandHAPO(non-GWAS)studiesandaccesstosummarydatafromallcontributingstudiesandtakeresponsibilityfortheintegrityofthedataandaccuracyofthedataanalysis.Studyconceptanddesign:J.Tyrrell,D.A.Lawlor,T.M.Frayling&R.M.FreathyAnalysisandinterpretationofdata:J.Tyrrell,R.C.Richmond,S.Metrustry,B.Feenstra,A.Cavadino,E.Hyppönen,M-FHivert,J.F.Felix,W.L.Lowe,D.A.Lawlor,T.M.Frayling&R.M.FreathyDraftingmanuscript:J.Tyrrell,R.C.Richmond,S.Metrustry,B.Feenstra,A.Cavadino,E.Hyppönen,M-FHivert,J.F.Felix,W.L.Lowe,D.A.Lawlor,T.M.Frayling&R.M.Freathy

Criticalrevisionofmanuscriptforimportantintellectualcontent:Allauthors

Statisticalanalysis:N.M.G.DeSilva,A.R.Wood,G.McMahon,D.M.Evans,E.Kreiner-Møller,F.Geller,C.Allard,J-J.Hottenga,J.P.Bradfield,M.G.Hayes,D.M.Scholtens,L.L.Armstrong,J.Rangaraja,M.Horikoshi,M.I.McCarthy,M.Nodzenski,L.Paternoster,S.Das,V.Huikari,J.N.Painter,S.E.Medland,P.A.Lind,D.J.Berry,R.Myhre,V.Sengpiel,J.Tyrrell,R.C.Richmond,S.Metrustry,B.Feenstra,A.Cavadino,J.F.Felix,D.A.Lawlor&R.M.Freathy

Genotyping:D.M.Evans,S.M.Ring,J.C.Murray,L.Bouchard,J.F.Felix,J-J.Hottenga,H.Hakonarson,T.I.A.Sørensen,N.G.Martin,E.A.Nohr,T.D.Spector,S.F.Grant,D.A.Lawlor,T.M.Frayling&R.M.Freathy

Individualstudydesign:T.D.Spector,B.Jacobsson,C.Power,N.G.Martin,S.Serbert,T.I.A.Sørensen,M.Järvelin,B.Hocher,M.G.Hayes,W.L.Lowe,S.F.Grant,H.Hakonarson,J.P.Bradfield,E.J.C.deGeus,V.W.Jaddoe,A.Hofman,M.Melbye,J.C.Murray,H.Bisgaard,G.DaveySmith,A.T.Hattersley,E.Hyppönen,M-FHivert,J.F.Felix,W.L.Lowe,D.A.Lawlor,&T.M.Frayling

Phenotypingincontributingstudies:T.D.Spector,B.Jacobsson,C.Power,E.Hyppönen,N.G.Martin,T.I.A.Sørensen,E.A.Nohr,C.Reichetzeder,B.Hocher,S.F.Grant,H.Hakonarson,J.P.Bradfield,G.Willemsen,V.W.Jaddoe,M-FHivert,M.Melbye,F.Geller,B.Feenstra,E.Kreiner-Møller,H.Bisgaard,G.DaveySmith,S.M.Ring,D.A.Lawlor,B.Knight,A.T.Hattersley

25

ConflictsofInterestandFinancialDisclosuresNoconflictsofinterestwerereported.Funding/SupportFunding/supportofauthorsisasfollows(fundingdetailsforindividualstudiesarereportedintheSupplementaryMaterial).T.M.FandA.R.WaresupportedbytheEuropeanResearchCouncilgrant:323195SZ-24550371-GLUCOSEGENES-FP7-IDEAS-ERC;A.T.H.andM.I.M.areWellcomeTrustSeniorInvestigators;M.I.M.isanNIHRSeniorInvestigator.R.M.F.isaSirHenryDaleFellow(WellcomeTrustandRoyalSocietygrant:104150/Z/14/Z);J.T.isfundedbytheERDFandaDiabetesResearchandWellnessFoundationFellowship;RCRisfundedbytheWellcomeTrust4-yearstudentship(GrantCode:WT083431MF).D.A.L.,G.D-S.,D.M.E.andS.R.allworkinaUnitthatreceivesfundingfromtheUniversityofBristolandtheUKMedicalResearchCouncil(MC_UU_1201/1/5,MC_UU_1201/1andMC_UU_1201/4).D.A.L.issupportedbyawardsfromtheWellcomeTrust(WT094529MAandWT088806),USNationalInstituteofHealth(R01DK10324)andisanNIHRSeniorInvestigator(NF-SI-0611-10196).D.M.EandS.E.M.weresupportedbyAustralianResearchCouncilFutureFellowship(FT130101709andFT110100548).B.FissupportedbyanOakFoundationScholarship.L.B.isajuniorresearchscholarfromtheFondsdelarechercheensantéduQuébec(FRQS)andamemberoftheFRQS-fundedCentrederechercheduCHUS.M.F.H.isaFRQSresearchscholarsandwasawardedaClinicalScientistAwardbytheCanadianDiabetesAssociationandtheMaudMentenAwardfromtheInstituteofGenetics–CanadianInstituteofHealthResearch(CIHR).C.A.wasawardedtheCIHR-FrederickBantingandCharlesBestCanadaGraduateScholarships.V.W.J.issupportedbytheNetherlandsOrganizationforHealthResearchandDevelopment(ZonMw–VIDI016.136.361).A.P.M.isaWellcomeTrustSeniorResearchFellow(grantnumberWT098017).T.S.isholderofanERCAdvancedPrincipalInvestigatoraward.RoleoftheSponsorsThefundingagencieshadnoroleinthedesignandconductofthestudy;collection,management,analysisandinterpretationofdata;preparation,review,approvalofmanuscript;ordecisiontosubmitmanuscriptforpublication.PreviousPresentationsThisworkwaspresentedattheDiabetesUKAnnualProfessionalConference2014,5-7March,Liverpool,UK.AdditionalContributionsWeareextremelygratefultotheparticipantsandfamilieswhocontributedtoallofthestudiesandtheteamsofinvestigatorsinvolvedineachone.Theseincludeinterviewers,computerandlaboratorytechnicians,clericalworkers,researchscientists,volunteers,managers,receptionistsandnurses.Foradditionalstudy-specificacknowledgements,pleaseseeSupplementaryMaterial.ThisresearchhasbeenconductedusingtheUKBiobankresource.

26

FIGURE/BOXLEGENDS

Box1.Thematernalobesity-relatedtraitshypothesizedtocauseincreasedordecreasedfetalgrowth,

based on observational associations with birth weight: body mass index (BMI)1; fasting glucose5;

gestationalortype2diabetes32;triglycerides9;HDL-cholesterol8;systolicbloodpressure10;vitaminD

status(asindicatedby25-hydroxyvitaminD,25[OH]Dlevel)11;adiponectin12.

Figure1

PrincipleofMendelianrandomization:Ifamaternaltraitcausallyinfluencesoffspringbirthweight,

thenariskscoreofgeneticvariantsassociatedwiththattraitwillalsobeassociatedwithbirth

weight.Sincegenotypeisdeterminedatconception,itshouldnotbeassociatedwithfactorsthat

normallyconfoundtheassociationbetweenmaternaltraitsandbirthweight(e.g.socio-economic

status).Estimatesofthegeneticscore-maternalphenotypeassociation(w)andthegeneticscore-

birthweightassociation(x)maybeusedtoestimatetheassociationbetweenthematernaltrait

variationthatisduetogeneticscore,andbirthweight(y=x/w),whichisexpectedtobefreefrom

confounding.Iftheestimatedcausalrelationship,y,isdifferentfromtheobservationalassociation

betweenthemeasuredmaternalphenotypeandbirthweight,thiswouldsuggestthatthe

observationalassociationisconfounded(assumingthattheassumptionsoftheMendelian

randomizationanalysesarevalid).14Thedashedlineconnectingmaternaltraitwithfetalgrowthhas

noarrow,toindicatethatthecausalnatureoftheassociationisuncertain.Itisimportanttoadjust

forpossibledirecteffectsoffetalgenotype(z).

Figure2.Comparisonoftheobservationalwiththegeneticchangeinbirthweight(ingrams)fora1

standarddeviation(SD)changeineachmaternalobesity-relatedtrait.For25[OH]Dandadiponectin,

wepresentthechangeinbirthweightfora10%changeinmaternaltraitlevelbecausethese

variableswereloggedforanalysis.ThegeneticchangewasestimatedfromMendelian

randomizationanalysis,inwhichageneticscorewasusedtoestimatethepossiblecausal

27

relationshipbetweenthematernaltraitandbirthweight.Thegeneticestimateispresentedtwice:in

thesecondcaseitwasadjustedforfetalgenotypeusingasubsetofavailablestudies.Theerrorbars

representthe95%confidenceintervalsaroundtheeffectsizeestimates.Formaternalpre-

pregnancyBMIandfastingglucose,the95%confidenceintervalsforboththeobservationaland

geneticapproachesexcludethenull,suggestingpositivepossiblecausalrelationshipsbetween

maternalBMIandfastingglucoseandbirthweight.FormaternalSBP,theobservationalanalysis

suggestedaweakpositiveassociationwithbirthweight,whereasthegeneticanalysisshowed

evidenceofanegativepossiblecausalrelationship.Observationalanalysessuggestedthathigher

maternaltriglyceridelevels,lowermaternaladiponectinandlowermaternalHDL-cholesterollevels

wereassociatedwithhigherbirthweight,whilelowermaternalvitaminDstatuswasassociatedwith

lowerbirthweight,butnoneoftheseweresupportedbythegeneticanalyses.

28Table1.Keycharacteristicsofparticipantsbystudy(forfulldetails,seeeTable1)

Abbreviatedstudyname*

Country(samplesource)

Offspringyearsofbirth

Nwomenwithbirthweightofonechild

Noffspringwith

genotype

Meanmaternalageat

deliveryinyears(SD)

Meanmaternal

pre-pregnancyBMI(SD)in

kg/m2

Meanoffspring

birthweight(SD)ingrams

ALSPACmothers33 UK 1991-1992 7304 4913 28.5(4.8) 22.93(3.73) 3481(475)

BBCmothers34 Germany 2000-2004 1357 1357 30.1(5.4) 22.78(3.93) 3472(511)

B58C-WTCCC35 UK 1972-2000 855 NA 26.2(5.2) NA 3325(483)

B58C-T1DGC35 UK 1972-2000 836 NA 26.1(5.4) NA 3379(469)

CHOPmothers36 USA 1987-

present 312 NA NA NA 3440(562)

COPSAC-2000mothers37 Denmark 1998-2001 282 282 30.4(4.3) NA 3560(505)

DNBC-GOYA38 Denmark 1996-2002 1805 NA 29.2(4.2) 23.57(4.27) 3643(495)

DNBC-PTB-CONTROL39 Denmark 1987-2009 1649 975 29.9(4.2) 23.57(4.27) 3595(497)

EFSOCHmothers40 UK 2000-2004 746 332‡ 30.5(5.3) 24.07(4.42) 3512(480)

GEN-3Gmothers41 Canada 2010-2013 676 NA 28.4(4.4) 24.83(5.63) 3448(433)

GenerationRmothers42 TheNetherlands 2002-2006 3810 2196 31.2(4.5)† 23.12(3.92) 3528(494)

HAPOmothers(GWAS)5

UK,Canada,Australia 2000-2006 1380 1300 31.5(5.3)† 24.5(5.0) 3557(517)

HAPOmothers(non-

GWAS)5

USA,UK,Canada,Australia

2000-2006 3590 2318 30.4(5.4)† 24.63(5.33) 3526(463)

MoBamothers43 Norway 1999-2008 650 350 28.5(3.3) 23.93(3.94) 3679(430)

NFBC196644 Finland 1987-2001 2035 NA 26.5(3.7) NA 3525(461)

NTR45 TheNetherlands 1946-2003 706 NA 27.1(3.7) NA 3469(529)

QIMR46 Australia 1929-1990 892 NA Q24.5(4.0)R 22.79(5.13) 3344(532)

TwinsUK47,48 UK NA 1602 NA NA NA 3365(581)

*Expandedstudynames:ALSPAC,AvonLongitudinalStudyofParentsandChildren;BBC,BerlinBirthCohort;B58C-WTCCC,1958BritishBirthCohort-WellcomeTrustCaseControlConsortium;B58C-T1DGC,1958BritishBirthCohort-Type1DiabetesGeneticsConsortium;CHOP,Children’sHospitalOfPhiladelphia;DNBC-GOYA,DanishNationalBirthCohort-GeneticsofObesityinYoungAdultsstudy;DNBC-PTB-CONTROLS,DanishNationalBirthCohortPretermBirthstudyControls;EFSOCH,ExeterFamilyStudyOfChildhoodHealth;GEN-3G,GeneticsofGlycemicregulationinGestationandGrowth;HAPO,HyperglycemiaandAdversePregnancyOutcomestudy(GWAS,Genome-WideAssociationStudy);MoBa,theNorwegianMotherandBabyCohort;NFBC1966,theNorthernFinland1966BirthCohort;NTR,NetherlandsTwinRegistry;QIMR,QueenslandInstituteofMedicalResearch.†InGenerationR,maternalagewasrecorded,onaverage,at14.4weeksofgestation;inHAPO,maternalagewasrecorded,onaverage,at28weeksofgestation.‡FetalgenotypeinEFSOCHavailableonlyforthefastingglucosegeneticscore.NA,notavailable.

29Table2.Associationsbetweenmaternalgeneticscoresandmaternalobesity-relatedtraits

Maternalobesity-relatedtrait

NumberofSNPsusedtoconstructgeneticscore

ReferenceforprimaryGWAS

paperforeachsetofSNPs*

Estimateof%variancein

maternaltraitexplainedby

geneticscoreinpregnantwomen†

TotalNwomenwithtraitmeasured

duringpregnancy(exceptBMI,for

whichtheappropriate

measurementispre-pregnancy)

Nstudies

Estimatedchangeinmaternaltraitperaverageweighted

trait-raising/lowering‡allele(95%CI)

Pvalue

HeterogeneityPvalue;I2%,

whereresultsfrom>1studyweremeta-analysed

Higherpre-pregnancyBodyMassIndex

(BMI)30

Speliotesetal.,2010,NatGenet49

1.8%inALSPAC11,822 5 0.145(0.126,0.164)

kg/m2 <2x10-16 0.18;35.8

Higherfastingglucose§ 13

Dupuisetal.2010,NatGenet50

5%inEFSOCH5,402 3 0.029(0.025,0.032)

mmol/L <2x10-16 0.70;0

Higheroddsofgestationaldiabetesandexistingdiabetes(SNPsassociatedwith

type2diabetes)

55

Morrisetal.2012,NatGenet51

1.4%inALSPAC6,606

(54Cases,6,552controls)

1 Oddsratio:1.08(1.03,1.14) 0.003 -

Highertriglycerides 17 Teslovichetal.2010,Nature52

3%inEFSOCH 663 1 0.055(0.032,0.078)mmol/L 3x10-6 -

LowerHDL-cholesterol 4 Teslovichetal.2010,Nature52

3%inEFSOCH 733 1 -0.050(-0.072,-0.027)mmol/L 1x10-5 -

Highersystolicbloodpressure 33 Ehretetal.2010,

Nature531%inALSPAC 8,450 2 0.186(0.140,0.231)

mmHg <2x10-16 0.04;76.0

LowervitaminDstatus,ln[25(OH)D] 2(“Synthesis”score)¶

Vimaleswaranetal.2013,PloSMed6

0.2%inALSPAC4,767 1 -0.024(-0.039,-

0.009)onlogscale 0.002 -

Loweradiponectin,ln[adiponectin] 3

Yaghootkaretal.2013,Diabetes54

2%inHAPO1,376 1 -0.17(-0.23,-0.11)

onlogscale 1x10-8 -

*Thereferencesincolumn3indicatethepublishedgenome-wideassociationstudiesthatoriginallyidentifiedtheSNPsusedinthegeneticscores(studiesofnon-pregnantindividuals).†Toestimatethevarianceineachtraitexplainedbyitsrespectivegeneticscoreinpregnantwomen(column4),thelargestavailablestudywasused.Abbreviations:ALSPAC,AvonLongitudinalStudyofParentsandChildren33.EFSOCH,ExeterFamilyStudyOfChildhoodHealth40.HAPO,HyperglycemiaandAdversePregnancyOutcomestudy5.FurtherdetailsabouttheincludedstudiescanbefoundineTable4.‡Thedecisiontomodeltheassociationinrelationtothetrait-raisingortrait-loweringalleledependedontheknowndirectionofassociationofeachtraitwithhigherBMI(seeBox1).Column1specifieseachofthesedirectionsofassociation.§Removingtheonestudyinwhichthers10830963SNPwaspoorlyimputed(r2<0.8),weobtainedverysimilarresults(n=4026;effectsize=0.028(95%CI:0.024,0.032);P<2x10-16;Phet=0.46;I2=0%).

30¶ThetwoSNPsselectedforthevitaminDgeneticscorehaveahypothesizedroleinthesynthesisofvitaminD(asopposedtoitsmetabolism)andarerecommendedforuseinMendelianrandomizationstudies.16,17Levelsof25(OH)Dandadiponectinlevelswerelog-transformedtoachievenormalitybeforeanalyses.

31Table3.Associationsbetweenmaternalgeneticscoresandbirthweightofoffspring

Maternaltraitforwhichgeneticscorewas

constructedNstudies TotalN

women

Estimatedchangeinoffspringbirthweight(grams)permaternaltrait-

raising/lowering*allele(95%CI),tothenearest1

gram†

Pvalue

HeterogeneityPValue;I2%frommeta-analysis

Nstudieswithfetalgenotypes

TotalNoffspringwith

genotypesavailable

Estimatedchangeinbirthweight(grams)permaternaltrait-raising/loweringallele*(95%CI),tothenearest1gram†,adjustedforfetal

genotypes

Pvalue(adjustedforfetal

genotypes)

HeterogeneityPValue;I2%frommeta-analysis

(adjustedforfetal

genotypes)

Higherpre-pregnancyBMI 16 25,265 2(0,3) 0.008 0.84;0 7 10,964 4(1,6) 0.004 0.20;30.5

Higherfastingglucose 15 23,902 8(6,10) 7x10-14 0.11;33.3 8 11,493 11(7,14) 7x10-9 0.26;21.6

Higheroddsoftype2Diabetes 12 18,670 2(0,2) 0.06 0.22;23.1 5 7,769 4(2,6) 0.0004 0.81;0

Higheroddsoftype2Diabetes(excludingpre-existingandgestational

diabetes)

6 13,029 2(1,3) 0.02 0.92;0 4 6,210 4(1,6) 0.006 0.93;0

Highertriglycerides 15 24,985 -2(-4,0) 0.12 0.83;0 6 11,031 -2(-7,1) 0.21 0.86;0

LowerHDL-cholesterol 15 22,167 0(-3,3) 1 0.52;0 6 9,176 0(-5,5) 0.98 0.85;0

Highersystolicbloodpressure 13 20,062 -4(-6,-2) 1x10-5 0.14;30.4 5 7,790 -3(-6,0) 0.09 0.50;0

Highersystolicbloodpressure(excludingpre-

eclampsiaandhypertension)

7 13,271 -5(-7,-3) 6x10-6 0.18;32 4 5,488 -4(-8,0) 0.04 0.16;41.2

LowervitaminDstatus 18 30,340 -6(-12,0) 0.03 0.13;37.1 3 9,510 -14(-25,3) 0.01 0.77;0

Loweradiponectin 9 14,920 -2(-16,12) 0.76 0.90;0 5 7,820 7(-16,30) 0.55 0.71;0

*Thedecisiontomodeltheassociationinrelationtothetrait-raisingortrait-loweringalleledependedontheknowndirectionofassociationofeachtraitwithhigherBMI(seeBox1).Column1specifieseachofthesedirectionsofassociation.Resultsareperaverageweightedallele,adjustedforsexandgestationalage.†StandarddeviationofbirthweightaveragedoveranumberofEuropeanstudies(=484g)55wasusedtogeneratetheseestimatesfromz-scores.Weconsidereda2-tailedPvalueof<0.05toprovideevidenceagainstthenullhypothesis.

32

33Table4.Acomparisonoftheobservationalwiththegeneticassociationbetweeneachmaternaltraitandoffspringbirthweight

Maternaltrait Valueofa1SDchangeinthetraitwithunits

Study/ies*usedforobservationalestimates

[TotalNwomen]

Nwomenforobservationalestimates

Observationalestimateofthechangeinbirth

weight(g)per1SD(or10%†)changein

maternaltrait,adjustedforsexand

gestationalage(95%CI)

Geneticestimateofthechangeinbirthweight(g),adjusted

forsexandgestationalage,per1SD(or10%†)changeinmaternaltrait,unadjustedfor

fetalgenotype(95%CI)[NwomenasinTables

1and2]

Pvalue‡comparing

observationalwithgeneticbirth

weightassociations(unadjustedforfetalgenotype)

Geneticestimateofthechangeinbirthweight(g),adjusted

forsex,gestationalageandfetalgenotype,per1SD(or10%†)changeinmaternaltrait(95%CI)[N

offspringasinTables1and2]

Pvalue‡comparing

observationalwithgeneticbirth

weightassociations(adjustedfor

fetalgenotype)

Higherpre-pregnancyBMI 4kg/m2

ALSPACMothers,EFSOCHMothers,HAPO

Mothers11,969 62(56,70) 55(17,93) 0.70 104(32,176) 0.28

Higherfastingglucose 0.4mmol/L EFSOCHMothers,HAPO

Mothers 6,008 92(80,104) 114(80,147) 0.28 145(91,199) 0.09

Highertriglycerides 0.7mmol/L EFSOCHMothers 930 32(7,56) -24(-55,8) 0.007 -33(-86,20) 0.03

LowerHDL-cholesterol 0.5mmol/L EFSOCHMothers 927 30(3,58) 0(-33,34) 0.17 -1(-55,54) 0.32

Highersystolicbloodpressure 10mmHg ALSPACMothers,HAPO

Mothers 12,077 24(15,34) -208(-394,-21) 0.01 -151(-390,89) 0.14

LowervitaminDstatus 10%† ALSPACMothers 4,710 -4(-7,-2) -26(-54,2) 0.13 -56(-112,1) 0.07

Loweradiponectin 10%† HAPOMothers(GWASonly) 1,376 14(9,18) -1(-9,7) 0.002 4(-9,17) 0.19

*Heterogeneitystatisticsfromthemeta-analysesofobservationalassociationswere:Phet=0.03andI2=67.7%forBMI;Phet=0.09andI2=59.1%forfastingglucose;Phet=0.54andI2=0%forSBP.† For25[OH]Dandadiponectin,wepresenttheestimatedchangeinbirthweightper10%reductioninmaternaltraitlevelbecausethesevariableswereloggedforanalysis.‡P-values<0.05areconsideredtoindicateevidencethatthegeneticeffectsizeestimateisdifferentfromtheobservationalestimate,suggestingthattheobservationalestimateissubjecttoconfoundingorbias.ALSPAC:AvonLongitudinalStudyofParentsAndChildren33;EFSOCH:ExeterFamilyStudyofChildhoodHealth40;HAPO:HyperglycaemiaandAdversePregnancyOutcomes5;SDstandarddeviation