Dietary habits in adolescence and midlife and risk of breast cancer in older women

16 Pages·2017·1.47 MB·English

RESEARCHARTICLE
Dietary habits in adolescence and midlife and
risk of breast cancer in older women
AlfheidurHaraldsdottir1,2*,JohannaE.Torfadottir2,UnnurA.Valdimarsdottir2,3,4,Hans-
OlovAdami3,4,ThorAspelund2,LaufeyTryggvadottir5,6,MariannaThordardottir6,Bryndis
E.Birgisdottir1,7,TamaraB.Harris8,LenoreJ.Launer8,VilmundurGudnason6,9,
LaufeySteingrimsdottir1,7
1 FacultyofFoodScienceandHumanNutrition,UniversityofIceland,Reykjavik,Iceland,2 CentreofPublic
HealthSciences,FacultyofMedicine,UniversityofIceland,Reykjavik,Iceland,3 Departmentof
Epidemiology,HarvardT.HChanSchoolofPublicHealth,Boston,Massachusetts,UnitedStatesofAmerica,
a1111111111 4 DepartmentofMedicalEpidemiologyandBiostatistics,KarolinskaInstitutet,Stockholm,Sweden,5 The
a1111111111 IcelandicCancerRegistry,Reykjavik,Iceland,6 FacultyofMedicine,UniversityofIceland,Reykjavik,
Iceland,7 UnitforNutritionResearch,UniversityofIcelandandLandspitaliNationalUniversityHospital,
a1111111111
Reykjavik,Iceland,8 LaboratoryofEpidemiologyandPopulationSciences,IntramuralResearchProgram,
a1111111111
NationalInstituteonAging,Bethesda,Maryland,UnitedStatesofAmerica,9 TheIcelandicHeart
a1111111111 Association,Kopavogur,Iceland
*[email protected]
OPENACCESS Abstract
Citation:HaraldsdottirA,TorfadottirJE,
Recentstudiesindicatethatlifestylefactorsinearlylifeaffectbreastcancerrisk.Wethere-
ValdimarsdottirUA,AdamiH-O,AspelundT,
TryggvadottirL,etal.(2018)Dietaryhabitsin foreexploredtheassociationofhighconsumptionofmeat,milk,andwholegrainproductsin
adolescenceandmidlifeandriskofbreastcancer adolescenceandmidlife,onbreastcancerrisk.Weuseddatafromthepopulationbased
inolderwomen.PLoSONE13(5):e0198017.
AGES-Reykjavikcohort(2002–2006),where3,326womenwithameanageof77years(SD
https://doi.org/10.1371/journal.pone.0198017
6.0)participated.Forfooditemsandprincipalcomponentderiveddietarypatternsweused
Editor:SabineRohrmann,UniversityofZurich,
Coxproportionalmodelstocalculatemultivariatehazardratios(HR)with95%confidence
SWITZERLAND
intervals(95%CI).Duringameanfollow-upof8.8years,97womenwerediagnosedwith
Received:December1,2017
breastcancer.Forbothadolescenceandmidlife,dailyconsumptionofryebreadwasposi-
Accepted:May11,2018 tivelyassociatedwithbreastcancer(HR1.7,95%CI1.1–2.6andHR1.8,95%CI1.1–2.9,
Published:May30,2018 respectively).Incontrast,persistenthighconsumptionofoatmealwasnegativelyassociated
withbreastcancer(0.4,95%CI0.2–0.9).Noassociationwasfoundforotherfooditemsor
Copyright:©2018Haraldsdottiretal.Thisisan
openaccessarticledistributedunderthetermsof dietarypatternsthatincludedryebread.Highryebreadconsumptioninadolescenceand
theCreativeCommonsAttributionLicense,which midlifemayincreaseriskoflate-lifebreastcancerwhilstpersistentconsumptionofoatmeal
permitsunrestricteduse,distribution,and
mayreducetherisk.
reproductioninanymedium,providedtheoriginal
authorandsourcearecredited.
DataAvailabilityStatement:Datacannotbe
madeavailableduetolegalrestrictions;partofthe
datainthisstudywasobtainedfromthenational
Introduction
IcelandicCancerRegistry,wherenationaldata
protectionlawsappliesandforbidsdatafrombeing Duringadolescencethefemalemammarytissueundergoesextensivemodelingorre-modeling.
publiclyavailable.Datacanberequestedthrough Consequently,researchershavehypothesizedthatbreasttissuemaybeparticularlysusceptiblefor
collaborationwiththeIcelandicHeartAssociation
initiationofbreasttumorsduringthisperiod[1,2].Thereisincreasingevidenceontheimpor-
(contact:[email protected])andthe
tanceofadultmidlifedietandriskofbreastcancer[3]whileavailablestudiesontheimpactof
IcelandicCancerRegistry(http://www.
dietduringadolescenceonbreastcancerriskarescarceandsomewhatinconsistent[4].Studying
krabbameinsskra.is/indexen.jsp?id=reference,
contact:[email protected]). earlylifedietcanbechallengingduetopotentialmisclassificationbias,theneedforalongfollow-
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 1/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
Funding:TheAGES-ReykjavikStudywasfunded upandlimitedvariationinfoodintakebetweenparticipants.Interestingly,therewasconsiderable
byNIHcontractN01-AG-12100,theIntramural variabilityindietaryhabitsbetweenresidencyareasinIcelandinearlyandmid20thcenturydue
ResearchProgramoftheNationalInstituteon
torelativeisolationofregionsanddifferencesinfoodaccess.Thisvariationwasobservedtobe
Aging,bytheIcelandicHeartAssociationandthe
strongestforthemostcommonfooditemsconsumedatthattime,ormeat,rye,milkproducts,
IcelandicParliament.Thisworkwassupportedby
fishandfishoil[5].Theseproductsarealsoofinterestinthiscontextbecauseoftheirdiversebio-
theTheIcelandicCentreforResearch,RANNIS
grantnumber:152495051,(http://en.rannis.is/)(A. activecompounds[6,7]aswellastheirwide-spreaduseinmodernWesterndiets.
Haraldsdottir)andthePublicHealthFundofthe IntheNurses’HealthStudy,womeninthehighestquintileofredmeatconsumptionin
IcelandicDirectorateofHealth(A.Haraldsdottir). adolescencehadsignificantlyhigherriskofbreastcancerthanwomeninthelowestquintile
Thefundingagencies(NationalInstituteonAging,
[8–10].Arecentmeta-analysisof14prospectivestudies[11]onadultred-andprocessedmeat
IcelandicHeartAssociationandIcelandic
consumptionreportedaslightlyincreasedbreastcancerriskandsimilarresultswereobserved
Parlament,)fortheAGES-ReykjavikStudy,
RANNIS,orDirectorateofHealthhadnoroleinthe intheNIH-AARPDietandHealthStudy[12].Incontrast,noassociationwitheitherchild-
design,analysisorwritingofthisarticle. hoodoradultmilkconsumptionandriskofbreastcancerhasbeenfound[13–18].Hightotal
dietaryfiberintakeinearlyadulthoodwasassociatedwithsignificantlylowerbreastcancer
Competinginterests:Theauthorshavedeclared
thatnocompetinginterestsexist. risk[19],aswastotalhighadolescentconsumptionoflignans[20],acommonphytoestrogen
commonlyfoundinwholegrains[21].Highconsumptionwholegrainfoodintakeinadoles-
cenceandearlyadulthoodwasassociatedwithlowerriskofpremenopausalbreastcancerrisk
butnotwithpost-menopausalrisk[22].Availablestudiesontotaladultwholegrainconsump-
tionandbreastcancerriskhaveeithersuggestedanegative[22–24]ornoassociation[25,26].
ForthisstudyweuseddatafromthepopulationbasedAGES-ReykjavikStudy,whichis
derivedfromapopulationwithconsiderablevariationindietaryhabitsinadolescence.Using
thesamecohort,wehavepreviouslyobservedapreventiveroleofveryhighfishintakeinado-
lescenceandmidlifeforbreastcancer[27]andalsotheimportanceofearlierdietforprostate
cancerrisk[28–30].Inthepresentstudyouraimwastoexploretheeffectsofhighconsump-
tionofmeat,milk,andwholegrainproductsinadolescenceandmidlifeonbreastcancerrisk
laterinlife,withamainemphasisontheadolescenceperiod.
Materialsandmethods
Studypopulation
WeuseddatafromtheAgeGeneEnvironmentSusceptibility(AGES)—ReykjavikCohortStudy
oftheIcelandicHeartAssociation,asub-cohortofthepopulationbasedReykjavikStudyinitiated
in1967[31].TheAGES-ReykjavikStudyexaminationsbeganin2002andatthattime11,549
ReykjavikStudycohortmemberswerestillalive.Thereof,8,030individualswererandomlyinvited
tothestudyandofthese,5,764individuals(thereof3,326women)participatedbetween2002–
2006(72%responserate).Extensivedatawerecollectedduringclinicalexaminations,including
informationonfoodintakeinadolescence,midlifeandpresentoldage.Forouranalysisweused
datafromthefirstclinicalexamination[32].Forthisstudy,weexcludedwomenwhohadbeen
diagnosedwithbreastcancerpriortostudyentry(n=196),leaving3130womeninourstudy.
Dietaryassessmentinearlylifeandmidlife
Atstudyentry,theparticipantscompletedafoodfrequencyquestionnaire(FFQ)ondietin
adolescence(betweentheagesof14to19years),inmidlife(betweentheages40to50years),
andatstudyentry(latelife).Forthisanalysisweonlyusequestionsfromtheadolescentand
midlifeperiod.
TheFFQwasespeciallydesignedforthisprojectandprovidesinformationonfrequencyof
intakeof10commonfoodsandfoodgroupsconsumedinadolescenceand11inmidlife.For
bothadolescenceandmidlife,thesefoodgroupsweremeat(includingsaltedandsmoked
meat),fish(includingsaltedorsmokedfish),fishliveroil,bloodorliversausage,ryebread,oat-
meal,potatoes,milkandmilkproducts,fruitsandvegetables[33].Anadditionalquestionon
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 2/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
consumptionofwholewheatbreadwasincludedforthemidlifeperiod,butthistypeofbread
wasuncommonintheadolescentperiod.Aspreviouslystated,forthisanalysisourmainfocus
isonmeat,includingsaltedandsmokedmeat,milk,ryebread,oatmealandwholewheatbread.
Wehavepreviouslypublishedanalysisonfishandcodliveroilfromthissamecohort[27]while
analysisonfruitsandvegetableswerenotconductedduetoverylowconsumptiononadaily
basisforbothadolescenceandmidlife.However,allfoodgroups,asdescribedabove,were
includedinthedietarypatternanalysis(seesupportinginformation).
Twoseparatequestionswereaskedregardingmeatconsumption.Oneincludedtotalcon-
sumptionofmeatandgroundmeatasameal(hereafterreferredtoasmeat).Theotherques-
tion(includedintotalmeatconsumption)concernedintakeofcornedmeat,cornedmeat
sausage,oranykindofsalted/smokedmeat(hereafterreferredtoassaltedorsmokedmeat).
Informationonmilkconsumptionincludedfrequencyofintakeofmilkandmilkproducts
(hereafterreferredtoasmilk).Ryebreadconsumptionwasassessedbyonequestiononintake
ofryebreadandflatbreadmadeofrye(hereafterreferredtoasryebread).Formidlife,the
questiononoatmealalsoincludedmuesli,butwillbereferredtoasoatmealinbothperiods.
Formeat,milk,ryebread,oatmeal,andwholewheatbread,thefrequencyofconsumption
wasclassifiedinto;1)never,2)lessthanonceaweek,3)1–2timesaweek,4)3–4timesaweek,
5)5–6timesaweek,6)daily,and7)morethanonceaday.Forsaltedorsmokedmeat,thefol-
lowingresponsecategorieswere:1)never,2)lessthanonceamonth,3)1–3timesamonth,4)
1–2timesaweek5)3–6timesaweek6)dailyormoreoften.
Forbothadolescenceandmidlife,meatconsumptionwasdividedintotwocategories.low(2
timesorlessperweek)andhigh(3timesormoreperweek).Consumptionofsaltedorsmoked
meatwasalsodividedintotwocategories,withlowintakedefinedas3timespermonthorless
andhighasonceperweekormore.Consumptionofmilk,ryebreadandwholewheatbread
(midlifeonly)wasdividedintotwocategories(lessthandailyanddailyormore).Consumption
ofoatmealwasdividedintolow(4timesaweekorless)andhigh(5timesaweekormore).
TheFFQhasbeenvalidatedformidlifeandlatelife[33,34].Inshort,midlifedietaryhabits
werevalidatedbycomparingtheresultsintheAGES-FFQ(n=107)withdetaileddietarydata
gatheredfromthesameindividuals18–19yearspreviouslyinaNationalnutritionsurveycon-
ductedin1990.Themainresultswerethatthecorrelationcoefficientsformostofthefood
itemswerewithinanacceptablerange[33].
Covariateassessment
Informationonpotentialconfounderswasmainlyretrievedfromalifestylequestionnaire,
completedatentrytotheAGES-ReykjavikStudy.Wecollectedinformationonageatentryto
thestudy(continuous),ageatmenarche(continuous),familyhistoryofbreastcancer(mother,
sisterand/ordaughtereverdiagnosedwithbreastcancer),education(primary,secondary,col-
lege/university),useofhormonalreplacementtherapy(never,ever),oralcontraceptive(never,
ever),useofalcoholinmidlife(yes,no),andphysicalactivityinadolescenceandmidlife
(never/rarely,occasionally,moderately/often).Wealsoretrievedinformationoneverbeing
pregnant(y/n)andageatfirstbirth(continuous),andcombinedthisinformationintoone
variable(nobirths,age24andyounger,25andolder).FromtheReykjavikStudyweretrieved
valuesonbodymassindex(BMI)andheightfromthemidlifeperiod(bothcontinuous).Infor-
mationondietarycovariateswasretrievedfromtheAGES-FFQ(Fig1).
Selecteddietarycovariatesonconcurrentconsumptionincluded,milk,meat,saltedand
smokedmeat,oatmeal,andryebread,dependingontheexposure(seecutoffsabove).Wealso
usedinformationonfishconsumption,((cid:20)2portionsp/wvs.>2–4portionsp/wvs.>4por-
tionsp/w)andonsaltedandsmokedfish(3timesamonthorlessvs.onceaweekormore).
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 3/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
Fig1.Timelineofexaminationpointsinthestudy.
https://doi.org/10.1371/journal.pone.0198017.g001
Asresidencedependentvarianceinfoodconsumptionexistedduringtheadolescence
period[5]wealsoaddedinformationonplaceofresidencewhengrowingupinourmodels
(capital,coastalvillage,ruralarea,andcombinationofcoastalvillageandruralarea)collected
intheReykjavikStudy.
Ascertainmentofoutcome
WeascertainedbreastcancerdiagnosisthroughthenationwideIcelandicCancerRegistry
[35].InformationoncauseofdeathwasobtainedfromDirectorateofHealth.Duetothecom-
puterizednationalrosteranduniquepersonalidentificationnumbersforeachperson,follow-
upwasvirtuallycomplete(99%)[35].Participantswerefollowedfromthestudyentry
(between2002and2006)todiagnosisofbreastcancer,death,oruntiltheendoftheobserva-
tionperiod(December31,2014),whicheveroccurredfirst.
Statisticalanalysis
Forbothadolescenceandmidlife,Coxproportionalhazardregressionmodelswereusedto
calculatehazardratio(HR)and95%confidenceinterval(95%CI)forincidentbreastcancer.
Forallexposuresintheadolescentperiodthefirstmodelwasadjustedforage(asacontinuous
variable)atentry.Dependingonexposure,allotherfooditemsunderstudy(meat,saltedand
smokedmeat,milk,ryebread,oatmeal)plusinformationonfishandsaltedandsmokedfish
werethenaddedsimultaneouslytothesecondmodel(seecutoffsaboveincovariatesassess-
ment).Furtheradjustmentswerethenmadeforeducation(3categories),BMI(continuous)in
midlife,ageatfirstchildbirth(3categories)andageatmenarche(continuous).
Aftercombininginformationoneverbeingpregnantandageatfirstbirthintoonevariable
wehadmissingvaluesfor211women.Themissingvalueswereincludedinthemultivariable
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 4/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
analysisasaspecialcategory.Wealsoreplacedthe21missingvaluesforBMIwiththemean
BMIvalueoftheparticipantsofthestudy,or25.The177missingvaluesforageatmenarche
werereplacedwiththemeanageofmenarcheinthecohort,or14years.Wedidnotreplace
missingvaluesforfooditems,neitherforexposuresorcovariatevariables.
Formidlife,thesametypesofmodelswerepresented,exceptheremidlifemeasuresforall
foodconsumptionwereused.Also,informationonconsumptionofalcoholandwholewheat
breadwasaddedtothemultivariatemidlifemodel.
Forbothperiods,furtheradjustmentforfamilyhistoryofbreastcancer,height,andphysical
activity,familyhistoryofbreastcancer,hormonalreplacementtherapyororalcontraceptive
didnotchangeourriskestimates(datanotshown)andwerethereforenotincludedinour
finalstatisticalmodels.
Toassessthepotentialeffectsoflongitudinaldietaryhabitsonbreastcancerriskwepooled
consumptionofeachfooditeminadolescenceandmidlifeintoonevariablewithfourcatego-
ries;1)lowinbothadolescenceandmidlife;2)lowinadolescenceandhighinmidlife;3)high
inadolescenceandlowinmidlife;and4)highinbothadolescenceandmidlife.Forthisanalysis,
thefirstcategory(lowconsumptioninbothperiods)wasusedasareference.Adjustmentswere
madeforsamefactorsasdescribedfortheadolescentperiod.WealsoperformedSpearman´s
correlationtestbetweenconsumptiononfooditemsunderstudyinadolescenceandmidlife.
Tofurtherexaminewhethertheassociationobservedforryebreadandoatmealpersisted
whenotherfooditemscommonlyconsumedwereincludedwealsousedprincipalcomponent
analysistoidentifydietarypatternsfromtheAGES-FFQ,includingalldietarydataavailable.
Thismethodisdatadrivenandformsnewlinearfactors(dietarypatterns)byreducingdata
dimensionandgroupingcorrelatedvariables(foodintake).Foreachpattern,anewvariableis
created,rankingparticipantsontheiradherencetothatparticularpattern[36].Eachvariable/
patternwasfurtherdividedintotertiles,orlow,medium,andhighadherencetoeachpattern.
WeusedCoxproportionalhazardregressiontotestassociationbetweenadherencetoadoles-
cenceandmidlifepatternsandbreastcancerrisk,usingthelowesttertileasareference.For
bothadolescenceandmidlife,adjustmentsweremadeforageatentry,BMI,education,ageat
menarche,andageatfirstchild,usingthesamecutoffsaspreviouslydescribedforindividual
exposuresinadolescenceandmidlife.Wealsotestedfortrendinthehazardratiosforthefirst
category,usingpolynomialcontrasts(datashowninS1andS2Tables).
ForallstatisticalanalysisweusedSPSSsoftware,version24.0(SPSSInc.,Chicago,Illinois;
www.spss.com)andRCoreTeam(2014).R:Alanguageandenvironmentforstatisticalcom-
puting.RFoundationforStatisticalComputing,Vienna,Austria;(http://www.R-project.org/).
ThestudyprotocolwasapprovedbytheIcelandicEthicalReviewBoardandtheIcelandic
DataProtectionAuthority(VSNb2007120014/03-7).
Results
Themeanageatstudyentrywas77.0yearsandstandarddeviation(SD)6.0.Themeanfollow-
uptimewas8.8years(SD=3.1).Duringthattime,97womenwerediagnosedwithbreastcan-
cer,withmeanageatdiagnosis81.4years(SD=6.5).Thedietaryanalyseswerebasedonpar-
ticipantswhoprovidedinformationondietinmidlifeandadolescencewhenenteringthe
AGES-Reykjavikcohortandwerefreeofbreastcancer.
Dietinadolescence
Table1showscharacteristicsofthepopulationthatprovideddataonmeat(n=2,866),milk
(n=2,871),andryebread(n=2,858)consumptioninadolescence.Whencomparedwith
womenwithlowconsumptionofmeatinadolescence,womenwithhighconsumptionofmeat
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 5/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
Table1. CharacteristicsoffemaleparticipantsintheAGES–ReykjavikStudy(2002–2006)accordingtoconsumptionofryebread,meatandmilkinadolescence.
Ryebread Meat Milk
N=2858 N=2866 N=2871
Low High Low High Low High
Lessthan Dailyor 2timesor 3timesor Lessthan Dailyor
daily more lessperweek moreper daily more
week
n=1452 n=1406 Pvalue n=958 n=1908 Pvalue n=695 n=2176 Pvalue
Ageatstudyentry
Mean(SD) 75.2 (5.3) 77.8 (5.7) 0.001 76.1 (5.8) 76.6 (5.5) 0.001 76.3 (5.7) 76.5 (5.6) 0.549
Median 74 78 75 76 75 76
Height(cm)(cid:3)
Mean(SD) 164.4 (5.2) 163.9 (5.5) 0.012 164.3 (5.4) 164.1 (5.4) 0.001 163.9 (5.1) 164.2 (5.5) 0.202
Median 164.5 164 164 164 164 164
BMI(kg/m2)(cid:3)
Mean(SD) 24.9 (3.7) 25.0 (3.8) 0.520 24.8 (3.8) 25.0 (3.8) 0.001 25.3 (3.9) 24.8 (3.7) 0.010
Median 24 24.5 24 24 24.5 24
Education,n(%)
Primary 560 (39) 667 (47) 0.001 358 (37) 874 (46) 0.001 316 (46) 915 (42) 0.232
Secondary 665 (46) 572 (41) 446 (47) 792 (42) 292 (42) 951 (44)
University/College 227 (15) 167 (12) 154 (16) 242 (12) 87 (13) 310 (14)
Birthcohort,n(%)
1908–1919 85 (6) 236 (17) 0.001 104 (11) 216 (11) 0.016 73 (11) 247 (11) 0.386
1920–1924 285 (20) 373 (26) 205 (21) 456 (24) 157 (23) 505 (23)
1925–1929 470 (32) 435 (31) 283 (30) 621 (33) 209 (30) 699 (32)
1930–1935 612 (42) 362 (26) 366 (38) 615 (32) 256 (37) 725 (33)
Locationoffirstresidence,n(%)
Reykjavik 603 (42) 458 (33) 0.001 351 (37) 712 (38) 0.001 283 (41) 783 (37) 0.001
Coastalvillage 515 (36) 417 (30) 267 (28) 666 (36) 270 (39) 664 (31)
Ruralarea 259 (18) 450 (33) 277 (30) 437 (23) 113 (17) 601 (28)
Combinationofcoastalvillageandruralarea 53 (4) 50 (4) 45 (5) 58 (3) 18 (3) 86 (4)
Ageatmenarche,n(%)
(cid:20)13y 669 (46) 609 (43) 0.147 447 (47) 836 (44) 0.147 309 (45) 978 (45) 0.824
(cid:21)14y 782 (54) 794 (57) 510 (53) 1069 (56) 385 (55) 1195 (55)
Ageatfirstbirth,n(%)
Nochildren 101 (7) 111 (8) 0.003 81 (8) 132 (7) 0.017 40 (6) 173 (8) 0.172
(cid:20)24y 927 (64) 804 (57) 544 (57) 1197 (63) 441 (63) 1300 (60)
(cid:21)25y 404 (28) 472 (34) 317 (33) 557 (29) 205 (30) 673 (31)
Physicalactivity
Never 600 (44) 685 (52) 0.001 405 (46) 881 (49) 0.179 310 (47) 977 (48) 0.499
Rarely/occasionally 335 (25) 242 (19) 202 (23) 376 (21) 153 (23) 426 (21)
Moderate/high 426 (31) 377 (29) 281 (32) 526 (30) 194 (30) 615 (31)
Familyhistoryofbreastcancer
Yes 235 (16) 238 (17) 0.593 147 (15) 330 (17) 0.186 121 (17) 355 (16) 0.499
Highintakeofotherfoodgroups
Fish(>4portionsp/w) 532 (37) 606 (43) 0.001 186 (19) 956 (50) 0.001 256 (37) 887 (41) 0.045
Saltedfish(oncep/wormore) 600 (42) 861 (62) 0.001 450 (47) 1008 (53) 0.002 255 (37) 1210 (56) 0.001
Oatmeal((cid:21)5timesp/w) 377 (26) 686 (49) 0.001 353 (37) 710 (38) 0.840 132 (19) 932 (43) 0.001
Saltedorsmokedmeat(oncep/wormore) 356 (25) 575 (41) 0.001 334 (35) 595 (31) 0.053 146 (21) 783 (36) 0.001
Milk(dailyormore) 974 (67) 1185 (84) 0.001 733 (77) 1432 (75) 0.328
(Continued)
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 6/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
Table1. (Continued)
Ryebread Meat Milk
N=2858 N=2866 N=2871
Low High Low High Low High
Lessthan Dailyor 2timesor 3timesor Lessthan Dailyor
daily more lessperweek moreper daily more
week
n=1452 n=1406 Pvalue n=958 n=1908 Pvalue n=695 n=2176 Pvalue
Meat(cid:21)3timesp/w) 944 (65) 953 (68) 0.122 473 (68) 1432 (66) 0.328
Abbreviations:SD,standarddeviation.
(cid:3)Measuredinmidlife—atentrytotheReykjavikStudy
PvaluesarebasedonChi-squaretests,exceptforageatentry,heightandBMI,whereOne-WayANOVAtestwasused
https://doi.org/10.1371/journal.pone.0198017.t001
(3timesormorep/w)wereolderatstudyentry,lesseducated,andolderwhenhavingtheir
firstchild.Theyalsoconsumedmoresaltedorsmokedmeatandmoresaltedandregularfish.
Womenwithhighmilkconsumption(dailyormoreoften)hadlowerBMIinmidlife,were
moreoftenraisedinruralareas,andhadmorefrequentconsumptionofcodliveroil,saltedor
smokedfish,totalfish,meat,andoatmeal.Whencomparedwithlessthandailyryebreadcon-
sumption,womenwhoconsumedryebreaddailyormoreoftenwereolderwhenenteringthe
study,hadlowereducationlevel,weremorecommonlyraisedinruralareas,wereolderupon
firstbirth,andlessphysicallyactive.Theyalsoconsumedfishmorefrequently,particularly
saltedorsmokedfishaswellasmilk,oatmeal,andmeat,includingsaltedorsmokedmeat.
Table2presentshazardratios,with95%CI,forbreastcanceranddietinadolescence.No
statisticallysignificantassociationwasfoundformeat,saltedandsmokedmeatandmilkcon-
sumption.Forwholegrainproducts,wefoundapositiveassociationbetweenhighconsump-
tionofryebread(dailyormoreoften)inadolescenceandbreastcancerrisk(HR1.7,95%CI
Table2. Multivariateanalysisofbreastcancerriskbymeat,milkandwholegrainintakeinadolescence.
Meat Saltedor Milk Ryebread Oatmeal
smokedmeat
N=2866 N=2862 N=2871 N=2858 N=2856
Low High Low High Low High Low High Low High
2timesorless 3timesor 3timesamonth Oncep/wor Lessthan Daily Lessthan Dailyor 4timesorless 5timesor
p/w morep/w orless more daily ormore daily more p/w morep/w
n=958 n=1908 n=1930 n=932 n=695 n=2176 n=1452 n=1406 n=1791 n=1065
Breastcancer 28 68 61 36 29 68 41 55 68 29
(%) (2.9) (3.6) (3.2) (3.9) (4.2) (3.1) (2.8) (3.9) (3.8) (2.7)
Ageadjusted 1.0 1.2 1.0 1.3 1.0 0.8 1.0 1.5 1.0 0.7
HR
(95%CI) (ref.) (0.8–1.9) (ref.) (0.8–1.9) (ref.) (0.50– (ref.) (1.0–2.2) (ref.) (0.5–1.2)
1.2)
Multivariate 1.0 1.3 1.0 1.4 1.0 0.7 1.0 1.7 1.0 0.7
HR(cid:3)
(95%CI) (ref.) (0.8–2.0) (ref.) (0.9–2.2) (ref.) (0.4–1.1) (ref.) (1.1–2.6) (ref.) (0.5–1.2)
Abbreviations:CI,confidenceinterval;HR,hazardratio:p/w,perweek
(cid:3)MultivariateHR;adjustedforageatentry,education,bodymassindexinmidlife,ageatfirstchildandageatmenarche.Allfooditemsunderstudy(meat,saltedand
smokedmeat,milk,ryebread,oatmeal)plusinformationonfishandsaltedandsmokedfishwerethenaddedsimultaneouslytothemultivariatemodel.The
multivariateanalysisincluded2810women,thereof95withbreastcancer.
https://doi.org/10.1371/journal.pone.0198017.t002
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 7/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
1.1–2.6),whencomparedwithlowerconsumption(lessthandaily).Nosignificantassociation
wasobservedforoatmealconsumption.Whenaddingearlyliferesidencetothemultivariate
modeltheriskestimatesforhighryebreadconsumptionincreased(HR1.8,95%CI1.2–2.9).
Nodifferencewasobservedforotherestimates.
Wefurtherdividedryebreadconsumptionintothreegroups(twotimesorlessperweek,
3–6timesperweek,anddailyormore)andexploredtheassociationwithbreastcancerrisk,
wherewefoundasignificanttrendacrossthegroups,HR1.0(95%CI0.5–2.0)andHR1.7
(95%CI0.9–3.2),respectively(P =0.043).
trend
Midlifediet
Formidlife(Table3),nosignificantassociationwasobservedformeatandmilkconsumption,
althoughamarginallypositiveassociationwasobservedforhighconsumption(onceperweek
ormore)ofsaltedorsmokedmeat(HR1.6,95%CI1.0–2.6)comparedwithwomenwithlow
consumption(3timesamonthorless).Forwholegrainproducts,noassociationwasobserved
forwholewheatbreadwhileastatisticallysignificantpositiveassociationwasobservedfor
highconsumption(dailyormore)ofryebread(HR1.8,95%CI1.1–2.9)whencomparedwith
lowerconsumption(lessthandaily).Nosignificantassociationwasobservedforoatmeal.No
differencewasobservedforanyoftheriskestimateswhenearlyliferesidencewasaddedtothe
model.
Ryebreadconsumptioninmidlifewasfurtherdividedintothreegroups(twotimesorless
perweek,3–6timesperweek,anddailyormore).Comparedwiththegroupwiththelowest
consumption,theriskestimatesfor3–6timesperweekanddailyormorewere1.4(95%CI
0.8–2.4)and2.3(95%CI1.3–4.1),respectively(P <0.01).
trend
Longtermconsumption
Table4presentsSpearman’scorrelationbetweendietaryhabitsinadolescenceandmidlife.
Thestrongestcorrelationwasfoundforryebreadconsumption(ρ=0.50,P=0.001)andthe
lowestformeatconsumption(ρ=-0.19,P=0.001).
Table5presentslongtermconsumptionformeat,saltedorsmokedmeat,milk,ryebread,
andoatmealwithlowconsumptioninadolescenceandmidlifeasareferencecategory.Aposi-
tiveassociationwasobservedforhighryebreadconsumptioninadolescenceandmidlife(HR
2.1,95%CI1.2–3.5).Aninverseassociationwasobservedbetweenbreastcancerriskandhigh
consumptionofoatmealinbothadolescenceandmidlife(HR0.4,95%CI0.2–0.9)(P =
trend
0.07).Noassociationwasobservedbetweenbreastcancerandmeat,saltedorsmokedmeat
andmilk.
Supplementarymaterial—dietarypattern
Fourdietarypatternswereextractedfortheadolescentperiod.Factorloadingcoefficientsfor
thosepatternsarepresentedinS1Table.Onepatterncontainingryebreadinadditionto
bloodliversausage,saltedmeat,saltedfish,andoatmealrepresentstraditionalIcelandicdietin
theearlierhalfofthe20thcentury.Highadherencetothispatternwasnotsignificantlyassoci-
atedwithbreastcancerrisk(HR1.395%CI0.7–2.1).Borderlineinverseassociationwas
observedforthehighestadherencetoapatternhighinconsumptionoffish,blood/liversau-
sage,oatmeal,fishoil,andmilkinadolescence(HR0.6,95%CI0.4–1.0)(P =0.049)(S2
trend
Table).
Fourdietarypatternswerealsoextractedforthemidlifeperiod.Noassociationwas
observedforanyofthosepatterns,includingthedietarypatternincludingryebreadconsump-
tion(datanotshown).
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 8/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
Discussion
Inthispopulationbasedcohort,dailyconsumptionofryebreadduringbothadolescenceand
midlifewaspositivelyassociatedwithbreastcancer.Conversely,wefoundreducedriskfor
breastcanceramongwomenwhoconsumedoatmealpersistentlybothinadolescenceand
midlife.However,nodietarypatternineitheradolescenceormidlifethatincludedryebread
wassignificantlyassociatedwithbreastcancerwhileapatterninadolescencethatrepresented
fish,blood/liversausage,oatmeal,fishoil,andmilkseemedtobeprotectiveforbreastcancer.
Wealsoobservedborderlineriskofbreastcancerwithhighmidlifeconsumptionofsaltedor
smokedmeat.Noassociationwasfoundbetweenhighmilkintakeandbreastcancerriskin
eitherperiod.
Tothebestofourknowledge,nostudyhasspecificallyaddressedadolescentconsumptionof
ryebread,acommonwholegrainproductintheNordiccountries,inrelationtolate-lifebreast
cancer.Afewstudieshaveanalyzedtotaladolescentconsumptionoflignans,acommonphyto-
estrogeninwholegraincereals—suchasrye,wheat,oats,andbarley—butalsoinlegumes,oil-
seeds,andvariousfruitsandvegetables[37].ACanadiancasecontrolstudyfoundthathigh
adolescentintakeoflignansreducedtheriskofbreastcancer.However,althoughryebreadwas
includedinthedietassessmentofthestudy,itwasnotcommonlyconsumedandtheseresults
canalsobeconfoundedbyotherhealthyeatinghabits[20].ResultsfromtheNurses’Health
Studysuggestthathighconsumptionofwholegrainsinadolescencecanreducetheriskofpre-
menopausalbreastcancer[22].However,themaindifferencebetweenthisstudyandourswas
ryebreaddoesnotseemtobeincludedintotalwholegrainconsumptionintheNurse´sHealth
Studyquestionnaire.Also,thewomeninourcohortwereallpost-menopausalwithahigh
meanageatdiagnosis,andthereforepossiblydifferentcarcinogenicprocess[38].
Thefewstudiesontotalwholegrainconsumptioninadultsandbreastcancerhaveeither
suggestedaprotective[23,24]ornoassociation[25,26].Themajordifferencebetweenmost
Table3. Multivariateanalysisofbreastcancerriskbymeat,milkandwholegrainintakeinmidlife.
Meat Saltedor Milk Ryebread Oatmeal Wholewheatbread
smokedmeat
N=2871 N=2864 N=2864 N=2875 N=2768 N=2865
Low High Low High Low High Low High Low High Low High
2timesor 3timesor 3timesa Oncep/w Lessthan Dailyor Lessthan Dailyor 4timesor 5timesor Lessthan Dailyor
lessp/w morep/w monthor ormore daily more daily more lessp/w morep/w daily more
less
n=1203 n=1668 n=2210 n=654 n=1247 n=1617 n=1918 n=957 n=2094 n=674 n=1238 n=1627
Breast 41 56 71 26 41 56 57 40 81 16 44 53
cancer (3.4) (3.4) (3.2) (4.0) (3.2) (3.5) (3.0) (4.2) (3.7) (2.3) (3.6) (3.3)
(%)
Age 1.0 1.0 1.0 1.3 1.0 1.1 1.0 1.5 1.0 0.6 1.0 0.9
adjusted
HR
(95%CI) (ref.) (0.6–1.5) (ref.) (0.8–2.1) (ref.) (0.7–1.6) (ref.) (1.0–2.3) (ref.) (0.4–1.1) (ref.) (0.6–1.4)
Multivariate 1.00 1.0 1.0 1.6 1.0 1.1 1.0 1.8 1.0 0.6 1.0 0.8
HR(cid:3)
(95%CI) (ref.) (0.6–1.4) (ref.) (1.0–2.6) (ref.) (0.7–1.7) (ref.) (1.1–2.9) (ref.) (0.4–1.1) (ref.) (0.5–1.3)
Abbreviations:CI,confidenceinterval;HR,hazardratio.
(cid:3)MultivariateHR;adjustedforageatentry,education,bodymassindexinmidlife,alcoholconsumptioninmidlife,ageatfirstchildandageatmenarche.Allfooditems
understudy(meat,saltedandsmokedmeat,milk,ryebread,oatmealandwholewheatbread)plusinformationonfishandsaltedandsmokedfishwerethenadded
simultaneouslytothemultivariatemodel.Themultivariateanalysisincluded2813women,thereof97withbreastcancer.
https://doi.org/10.1371/journal.pone.0198017.t003
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 9/16
Ryebreadintakeinadolescenceandmidlifeandriskofbreastcancer
Table4. Dietaryhabitsamongparticipantsthroughdifferenttimeperiods.
Adolescence Midlife Spearmans´sρ P
n(%) n(%)
Ryebread 0.50 0.001
Lessthandaily 1452(51) 1918(67)
Dailyormore 1406(49) 957(31)
Milkandmilkproducts 0.44 0.001
Lessthandaily 695(24) 1247(44)
Dailyormore 2176(76) 1617(56)
Meat -0.19 0.001
2timesorlessp/w 958(33) 1203(42)
3timesormorep/w 1908(67) 1668(58)
Saltedorsmokedmeat 0.36 0.001
3timesamonthorless 1930(67) 2210(77)
Onceormorep/w 932(33) 654(23)
Oatmeal 0.38 0.001
4timesorlessp/w 1791(63) 2715(76)
5timesormorep/w 1065(37) 690(24)
Vegetables 0.38 0.001
Never 585(20) 163(6)
6timesp/worless 2175(76) 2845(87)
Daily 105(4) 213(7)
Fruit 0.24 0.001
Never 1013(36) 114(4)
6timesp/worless 1784(62) 2501(87)
Daily 68(2) 256(9)
https://doi.org/10.1371/journal.pone.0198017.t004
oftheseandthepresentstudyisthatwedisentangledwholegrainconsumptionintorye,
wholewheatbreadandoatmeal.Differenttypesofwholegrainscontaindissimilarbioactive
compounds[7]whichmayactdifferentlyonbreastcancergenesis.
ADanishstudy[25]reportednoassociationbetweenneithertotaladultconsumptionof
wholegrainsnorwhenanalyzedseparately(ryebread,wholegrainbread,andoatmeal)with
breastcancer.However,thewomenintheDanishcohortwereyoungeratstudyentryandat
diagnosis(averageage56years)thanparticipantsintheAGES-ReykjavikStudy(averageage
77years).Thismightreflectapreviouslyreporteddifferenceincharacteristicsofdiagnosed
breasttumorsinolderwomen(over70)comparedwiththoseinyoungerwomen[38].Indeed,
wefoundnoassociationforryebreadwhenaseparateanalysiswasmadeforwomenwho
werealreadydiagnosedwithbreastcanceratstudyentry(meanageatdiagnosis64years,data
notshown).
Prolongedexposuretoryebreadmightthereforepossiblyexplaintheobservedriskfound
inourcohort.Ryebreadisrichinlignans,whichcanbeconvertedintoenterolignansbythe
gutmicrobiome.Enterolignanshavebeensuggestedtoenhancebreastcancerriskreduction
[7,39,40].However,theycanalsoexpressweakestrogenicaffinityandcanactbothasagonists
orantagonistsinbreasttumors.althoughthismechanismisconsideredcomplex[7,41–43].
Nevertheless,longitudinalexposuretoestrogens,exemplifiedbyearlymenarche,latemeno-
pause,anduseofhormonalreplacementtherapy,areconsideredasamajorriskfactorsfor
breastcancer[44].Therefore,itmaybehypothesizedthatlongtermexposuretophytoestro-
gensviaryebreadconsumptionmayhaveasimilareffect.Indeed,ouranalysisondietary
PLOSONE|https://doi.org/10.1371/journal.pone.0198017 May30,2018 10/16