Breast-Feeding and Childhood-Onset Type 1 Diabetes A...

Breast-Feeding and Childhood-Onset Type 1

Diabetes A pooled analysis of individual

participant data from 43 observational studies

Chris R Cardwell, Lars C Stene, Johnny Ludvigsson, Joachim Rosenbauer, Ondrej Cinek,

Jannet Svensson, Francisco Perez-Bravo, Anjum Memon, Suely G Gimeno, Emma J K

Wadsworth, Elsa S Strotmeyer, Michael J Goldacre, Katja Radon, Lee-Ming Chuang, Roger

C Parslow, Amanda Chetwynd, Kyriaki Karavanaki, Girts Brigis, Paolo Pozzilli, Brone

Urbonaite, Edith Schober, Gabriele Devoti, Sandra Sipetic, Geir Joner, Constantin Ionescu-

Tirgoviste, Carine E de Beaufort, Kirsten Harrild, Victoria Benson, Erkki Savilahti, Anne-

Louise Ponsonby, Mona Salem, Samira Rabiei and Chris C Patterson

Linköping University Post Print

N.B.: When citing this work, cite the original article.

Original Publication:

Chris R Cardwell, Lars C Stene, Johnny Ludvigsson, Joachim Rosenbauer, Ondrej Cinek,

Jannet Svensson, Francisco Perez-Bravo, Anjum Memon, Suely G Gimeno, Emma J K

Wadsworth, Elsa S Strotmeyer, Michael J Goldacre, Katja Radon, Lee-Ming Chuang, Roger C

Parslow, Amanda Chetwynd, Kyriaki Karavanaki, Girts Brigis, Paolo Pozzilli, Brone

Urbonaite, Edith Schober, Gabriele Devoti, Sandra Sipetic, Geir Joner, Constantin Ionescu-

Tirgoviste, Carine E de Beaufort, Kirsten Harrild, Victoria Benson, Erkki Savilahti, Anne-

Louise Ponsonby, Mona Salem, Samira Rabiei and Chris C Patterson, Breast-Feeding and

Childhood-Onset Type 1 Diabetes A pooled analysis of individual participant data from 43

observational studies, 2012, Diabetes Care, (35), 11, 2215-2225.

http://dx.doi.org/10.2337/dc12-0438

Copyright: American Diabetes Association

http://www.diabetes.org/

Postprint available at: Linköping University Electronic Press

http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-86376

http://dx.doi.org/10.2337/dc12-0438

http://www.diabetes.org/

http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-86376

http://twitter.com/?status=OA Article: Breast-Feeding and Childhood-Onset Type 1 Diabetes A pooled analysis of individu... http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-86376 via @LiU_EPress %23LiU

Breast-Feeding and Childhood-OnsetType 1 DiabetesA pooled analysis of individual participant data from 43observational studies

CHRIS R. CARDWELL, PHD1

LARS C. STENE, PHD2,3

JOHNNY LUDVIGSSON, MD, PHD4

JOACHIM ROSENBAUER, MD, DIPL MATH5

ONDREJ CINEK, MD, PHD6

JANNET SVENSSON, PHD7

FRANCISCO PEREZ-BRAVO, PHD8

ANJUM MEMON, MBBS, DPHIL9

SUELY G. GIMENO, DPH10

EMMA J.K. WADSWORTH, PHD11

ELSA S. STROTMEYER, PHD12

MICHAEL J. GOLDACRE, FFPHM13

KATJA RADON, PHD14

LEE-MING CHUANG, MD, PHD15

ROGER C. PARSLOW, PHD16

AMANDA CHETWYND, PHD17

KYRIAKI KARAVANAKI, MD, PHD18

GIRTS BRIGIS, MD, PHD19

PAOLO POZZILLI, MBBS, MD20

BRONE URBONAIT _E, MD, PHD21

EDITH SCHOBER, MD22

GABRIELE DEVOTI, MD23

SANDRA SIPETIC, MD, PHD24

GEIR JONER, MD, PHD25,26

CONSTANTIN IONESCU-TIRGOVISTE, MD, PHD27

CARINE E. DE BEAUFORT, MD, PHD28

KIRSTEN HARRILD, MSC29

VICTORIA BENSON, PHD30

ERKKI SAVILAHTI, MD, PHD31

ANNE-LOUISE PONSONBY, MBBS, PHD32

MONA SALEM, MD33

SAMIRA RABIEI, MSC34

CHRIS C. PATTERSON, PHD1

OBJECTIVEdTo investigate if there is a reduced risk of type 1 diabetes in children breastfed orexclusively breastfed by performing a pooled analysis with adjustment for recognized confounders.

RESEARCHDESIGNANDMETHODSdRelevant studies were identified from literaturesearches using MEDLINE, Web of Science, and EMBASE. Authors of relevant studies were askedto provide individual participant data or conduct prespecified analyses. Meta-analysis techniqueswere used to combine odds ratios (ORs) and investigate heterogeneity between studies.

RESULTSdData were available from 43 studies including 9,874 patients with type 1 diabetes.Overall, there was a reduction in the risk of diabetes after exclusive breast-feeding for.2 weeks(20 studies; OR = 0.75, 95% CI 0.64–0.88), the association after exclusive breast-feeding for.3months was weaker (30 studies; OR = 0.87, 95%CI 0.75–1.00), and no association was observedafter (nonexclusive) breast-feeding for.2 weeks (28 studies; OR = 0.93, 95% CI 0.81–1.07) or.3 months (29 studies; OR = 0.88, 95% CI 0.78–1.00). These associations were all subject tomarked heterogeneity (I2 = 58, 76, 54, and 68%, respectively). In studies with lower risk of bias,the reduced risk after exclusive breast-feeding for .2 weeks remained (12 studies; OR = 0.86,95% CI 0.75–0.99), and heterogeneity was reduced (I2 = 0%). Adjustments for potential con-founders altered these estimates very little.

CONCLUSIONSdThe pooled analysis suggests weak protective associations between exclu-sive breast-feeding and type 1 diabetes risk. However, these findings are difficult to interpretbecause of the marked variation in effect and possible biases (particularly recall bias) inherent inthe included studies.

Diabetes Care 35:2215–2225, 2012

Childhood type 1 diabetes is causedby the autoimmune destruction ofthe pancreatic b-cells. The increases

in type 1 diabetes incidence in recent dec-ades (1) and the concordance rate of,50% in monozygotic twins (2) sug-gests a role for environmental factors inthe development of diabetes. The propor-tional greater relative increases in inci-dence in children ,5 years old (1)indicates that environmental exposuresin early life, such as infant feeding, couldplay an important role.

In 1994, a meta-analysis (3) con-cluded that breast-feeding for a short pe-riod reduced the risk of type 1 diabetes by40%, but another meta-analysis in 1996(4), largely containing the same studies,concluded that the weak reduction in di-abetes risk after breast-feeding was smalland may have methodological explana-tions. However, both these meta-analyseshad limitations. The associations withtype 1 diabetes (extracted from publishedstudies) were based upon categorizeddurations of breast-feeding, which dif-fered between studies. Furthermore, inthese meta-analyses, it was not possibleto adjust the association between breast-feeding and type 1 diabetes consistentlyfor potential confounders (factors such asmaternal age, birth weight, and maternaldiabetes that were associated with bothdiabetes risk and the propensity to breast-feed). Also, studies that recorded breast-feeding results but did not report them insufficient detail were excluded, potentiallyexaggerating any breast-feeding associationthrough reporting bias. These weaknesseswere largely because both were literature-based meta-analyses (based only upon

c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c

From the 1School ofMedicine andDentistry, Queen’sUniversity Belfast, Belfast, U.K.; the 2Division ofEpidemiology, Norwegian Institute of PublicHealth, Oslo, Norway; the 3Oslo Diabetes Re-search Centre, Oslo University Hospital, Oslo,Norway; the 4Department of Clinical ExperimentalMedicine, Division of Pediatrics, Linkoping Uni-versity Hospital, Linkoping, Sweden; the 5Instituteof Biometrics and Epidemiology, German Diabe-tes Centre, Leibniz Institute at Duesseldorf Uni-versity, Duesseldorf, Germany; the 6Department of

Paediatrics, 2nd Faculty of Medicine, CharlesUniversity in Prague and University HospitalMotol, Motol, Czech Republic; the 7Pediatric De-partment, Herlev University Hospital, Herlev,Denmark; the 8Department of Nutrition, Faculty ofMedicine, Laboratory of Nutritional Genomics,University of Chile, Santiago,Chile; the 9DivisionofPrimary Care and Public Health, Brighton andSussex Medical School and NHS Brighton andHove, Brighton, U.K.; the 10Preventive MedicineDepartment, Federal University of São Paulo, São

Paulo, Brazil; the 11Cardiff Work EnvironmentResearch Centre, Cardiff University, Cardiff, U.K.;the 12Department of Epidemiology, GraduateSchool of Public Health, Center for Aging andPopulation Health, University of Pittsburgh,Pittsburgh, Pennsylvania; the 13Department ofPublic Health, Oxford University, Oxford, U.K.;the 14Institute and Outpatient Clinic for Occu-pational and Environmental Medicine, Hospitalof the Ludwig Maximilian University Munich,Munich, Germany; the 15Department of Internal

care.diabetesjournals.org DIABETES CARE, VOLUME 35, NOVEMBER 2012 2215

E p i d e m i o l o g y / H e a l t h S e r v i c e s R e s e a r c hO R I G I N A L A R T I C L E

reported associations in published stud-ies) rather than collaborative meta-analysesor pooled analyses (using individualparticipant data from included studiesafter contacting authors of articles con-taining relevant data) (5). Since the publi-cation of these two meta-analyses, manyadditional observational studies have in-vestigated breast-feeding and type 1 dia-betes, allowing a new meta-analysis to beconducted on an independent group ofmore recent studies.

The aim of this study was to assess theevidence of an association between dura-tion of breast-feeding (and exclusivebreast-feeding) and type 1 diabetes usingcollaborative meta-analysis techniquesfrom observational studies publishedsince 1996. Importantly, this pooledanalysis includes individual participantdata, allowing consistent categorization ofbreast-feeding across studies and adjust-ment for potential confounders such asmaternal age, birth weight, and maternaldiabetes and will also include studiesinvestigating breast-feeding but not re-porting detailed results.

RESEARCH DESIGN ANDMETHODS

Literature searchThe main literature search was conductedusing MEDLINE through Ovid Online.The search was adapted from a previousstudy (6) and comprised search terms forinfant feeding consisting of text words(including breast fed, breast feed, breastfeeds, breast feeding, breast milk, bottlefed, bottle feed, bottle feeds, bottle feed-ing, infant feed, infant feeds, infant feed-ing, infant nutrition, formula fed, formulafeed, formula feeds, formula feeding, in-fant diet, infant diets, dried milk, andearly nutrition) or MEDLINE subjectheading key words (including breast

feeding, bottle feeding, milk, human, in-fant food, and infant formula) and termsfor type 1 diabetes including the textwords (IDDM or [diabetes and type 1])or the MEDLINE subject heading keyword “diabetes mellitus, type 1.” Similarsearches were conducted on Web of Sci-ence and EMBASE. The searches werelimited to studies on humans publishedafter January 1996 [to avoid overlap withtwo previous meta-analyses (3,4)] up to 1May 2011. No language restrictions wereapplied. Titles and then abstracts werescreened independently by two investigators(C.R.C. and C.C.P.) to establish if thestudies were likely to provide relevantdata based on the following inclusion cri-teria: 1) they identified a group with type1 diabetes and a group without type 1diabetes, and 2) they recorded breast-feeding in these groups. Studies were ex-cluded if they contained ,20 patientswith diabetes or if they were family-based(because the association between infantfeeding and type 1 diabetes may be differ-ent in individuals with a strong geneticsusceptibility). The corresponding au-thor of each included article was askedif they were aware of any additionalstudies.

The corresponding author of eachstudy was contacted to provide individualparticipant data to allow breast-feeding tobe consistently categorized across studies,standardize adjustment of the associationwith breast-feeding for potential con-founders (including maternal diabetes,birth weight, gestational age, maternalage, birth order, Caesarean section, andsocioeconomic status), and include datafrom studies that did not report detailedfindings. Authors were requested to pro-vide raw data or to provide adjustedestimates of the association betweenbreast-feeding and type 1 diabetes afterconducting specified additional analyses.

Details of each study were extractedby one reviewer (C.R.C.) and agreed withthe study author.

Statistical analysisA two-stage technique was used to calcu-late pooled estimates of the associationbetween breast-feeding and diabetes, be-fore and after adjustment for potentialconfounders (7). Odds ratios (ORs) andSEs were calculated for the association be-tween diabetes and breast-feeding withineach study. Unconditional and condi-tional logistic regression was used to cal-culate the ORs and SEs for the unmatchedandmatched case-control studies, respec-tively. In cohort studies with varyinglength of participant follow-up, rate ratiosand their SEs were used instead of ORs,which were not directly calculable. Astype 1 diabetes is a rare disease, thesemeasures should be approximately equal.In one study (8) with a zero cell count foran exposure category, a correction wasadded to obtain an estimate and SE foruse in themeta-analysis. In another study,an additional adjustment was alwaysmade for the perinatal cohort entry scoreto reflect the study design (9). Finally, instudies for which no author could be con-tacted or data were not available, if possi-ble, required estimates were extractedfrom published reports. Meta-analysistechniques were then applied to these es-timates. Tests for heterogeneity betweenstudies were conducted and random-effects models used to calculate pooledORs. Random-effects models were used,instead of fixed-effects models, becausebetween-study heterogeneity was antici-pated from these observational studies.The I2 statistic was calculated to quantifythe degree of heterogeneity between stud-ies. Small study effects (possibly due topublication bias) were investigated bychecking for asymmetry in funnel plots

c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c

Medicine, National Taiwan University Hospital,Taipei, Taiwan; the 16Paediatric EpidemiologyGroup, University of Leeds, Leeds, U.K.; the17Mathematics and Statistics Department, LancasterUniversity, Lancaster, U.K.; the 18SecondUniversityDepartment of Pediatrics, National and Kapodis-trian University of Athens School of Medicine,Athens, Greece; the 19Department of Public Healthand Epidemiology, Riga Stradins University, Riga,Latvia; the 20University CampusBio-Medico, Rome,Italy; the 21Institute of Endocrinology of Lithua-nian University of Health Science, Kaunas, Lith-uania; the 22Department of Paediatrics, MedicalUniversity of Vienna, Vienna, Austria; the 23De-partment of Social Sciences and Communication,University of Salento, Lecce, Italy; the 24School ofMedicine, Institute of Epidemiology, Belgrade

University, Belgrade, Serbia; the 25Faculty ofMedicine, Institute of Health and Society, Uni-versity of Oslo, Oslo, Norway; the 26Departmentof Paediatrics, Oslo University Hospital, Oslo,Norway; the 27Nutrition and Metabolic DiseasesClinic, N. Paulescu Institute of Diabetes, Bu-charest, Romania; the 28Clinique PédiatriqueLuxembourg, Luxembourg City, Luxembourg;the 29Medical Statistics Team, University ofAberdeen, Aberdeen, U.K.; the 30Cancer Epide-miologyUnit, University of Oxford, Oxford, U.K.; the 31Hospital for Children and Adolescents,University of Helsinki, Helsinki, Finland; the32Murdoch Children’s Research Institute, TheRoyal Children’s Hospital, Melbourne, Australia;the 33Paediatrics Department, Ain Shams University,Cairo, Egypt; and the 34Faculty of Nutrition

Sciences and Food Technology, Shahid BeheshtiUniversity of Medical Sciences and Health Ser-vices, Tehran, Iran.

Corresponding author: Chris R. Cardwell, [email protected].

Received 5 March 2012 and accepted 9 May 2012.DOI: 10.2337/dc12-0438This article contains Supplementary Data online at

http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc12-0438/-/DC1.

A slide set summarizing this article is available online.© 2012 by the American Diabetes Association.

Readers may use this article as long as the work isproperly cited, the use is educational and not forprofit, and the work is not altered. See http://creativecommons.org/licenses/by-nc-nd/3.0/ fordetails.

2216 DIABETES CARE, VOLUME 35, NOVEMBER 2012 care.diabetesjournals.org

Breast-feeding and childhood type 1 diabetes

mailto:[email protected]

mailto:[email protected]

http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc12-0438/-/DC1


http://creativecommons.org/licenses/by-nc-nd/3.0/

http://creativecommons.org/licenses/by-nc-nd/3.0/

of the study ORs against one over the SEof the logarithm of the ORs.

Similar methods were conducted topool adjusted estimates. First, adjustedestimates and SEs were calculated withineach study using regression models ap-propriate to the study design includingdiabetes as the outcome variable and breast-feeding and the potential confounder(s)of interest as explanatory variable(s).Meta-analysis techniques were then applied tothese adjusted estimates.

The analysis was initially conductedbased upon a priori categorizations of non-exclusive and exclusive breast-feeding $2weeks versus breast-feeding ,2 weeksand was repeated for nonexclusive breast-feeding $3 months versus breast-feeding,3 months in which these data or similarwere available. Subgroup analyses werealso conducted including only studieswith a low risk of bias based upon highresponse rates (.70%) in both cases andcontrols and the use of randomly selectedpopulation-based controls. Subgroup anal-ysis was also performed by age at diagnosisof diabetes, geographic region, and inci-dence rate based upon a published sum-mary of worldwide incidence (10).

All statistical analyses were per-formed using STATA 11.0 (Stata, CollegeStation, TX).

RESULTS

Search resultsThe results from the literature search areshown in Supplementary Fig. 1. Briefly,initial searches identified 238 articlesfrom MEDLINE, 393 from Web of Sci-ence, and 609 from EMBASE. Of these,40 articles contained relevant data from47 studies, as information from 7 centerswas taken from 1 article (11) and informa-tion from 2 centers was taken from another(12). An investigator from each of the 47studies was invited to provide raw data (orestimates from prespecified analyses). Esti-mates were obtained from 43 of the 47identified studies covering 95% (9,874 of10,390) of potential patients with type 1diabetes. Individual patient data were ob-tained from 27 of these studies, 8 suppliedprespecified estimates, in 8 studies datacould be extracted from published reports,and 4 studies could not be included.

Characteristics of included studiesTable 1 contains the characteristics of the43 studies from which relevant data wereobtained (see Supplemental Data for fulllist of included articles). Of these, 40 were

case-control studies, and 3 were cohortstudies. Twenty-eight studies were fromEurope, and in 24 studies, cases wereidentified using a national or local diabe-tes register. Thirty-seven studies ascertainedbreast-feeding data using questionnairesor interviews, and in the majority,breast-feeding data were recalled manyyears after the birth of the child.

Overall findings for any(nonexclusive) breast-feedingThe associations between any (nonexclu-sive) breast-feeding and type 1 diabetesfrom the 43 studies (including 9,874patients with type 1 diabetes) are shownin Fig. 1 and Table 2. Initially, an analysiswas conducted investigating any recordedduration of breast-feeding (the breast-feeding duration used in this analysisfor each study is shown in Fig. 1).Overall, in studies which investigatedany measure of breast-feeding, there wasa reduction in the risk of type 1 diabetes inbreast-fed children of 18% (OR = 0.81,95% CI 0.72–0.92; P , 0.001). Therewas, however, marked heterogeneity be-tween studies (I2 = 63%; heterogeneityP , 0.001). Restricting this analysis tostudies with a low risk of bias (due tohigh response rates and randomly selectedpopulation-based controls, as shown inTable 1), there was no evidence of an as-sociation (OR = 1.00, 95% CI 0.89–1.11;P = 0.93) and little evidence of heteroge-neity (I2 = 1%; heterogeneity P = 0.51).

Analyses were also conducted in sub-groups of studies recording breast-feedingof ;2 weeks’ duration or more (see Fig. 1for exact duration of breast-feeding used ineach study) and ;3 months duration ormore (see Table 1 for exact duration ofbreast-feeding used in each study). In 28studies, there was little evidence of anydifference in the risk of type 1 diabetesin children breast-fed for ;2 weeks ormore compared with ,2 weeks (OR =0.93, 95% CI 0.81–1.07; P = 0.32), butagain there was heterogeneity betweenstudies (I2 = 54%; heterogeneity P ,0.001). This heterogeneity was reducedin 15 studies with a low risk of bias (I2 =8%; heterogeneity P = 0.37), but the over-all finding was little altered (OR = 1.00,95% CI 0.87–1.15; P = 0.99). In 29 stud-ies, there was also little evidence of a re-duction in the risk of type 1 diabetes inchildren breast-fed for ; 3 months ormore compared with children breast-fedfor ,3 months (OR = 0.88, 95% CI0.78–1.00; P = 0.05), but the heterogene-ity between studies persisted (I2 = 68%;

heterogeneity P , 0.001). Repeating thisanalysis in 15 studies with a low risk ofbias showed no evidence of an association(OR = 0.99, 95% CI 0.86–1.14; P = 0.92)and slightly less, though still significant,heterogeneity (I2 = 48%; heterogeneityP = 0.02). Funnel plots (not shown) pro-vided little evidence of asymmetry (poten-tially caused by publication bias).

Finally, an analysis was conducted inthe following approximate categories of(nonexclusive) breast-feeding:,2 weeks,2 weeks to 4 months, 4–6 months, and.6 months. In 24 studies with availabledata, when compared with childrenbreast-fed for ,2 weeks children breast-fed for 2 weeks to 4 months had a 4%reduction in diabetes risk (OR = 0.96,95% CI 0.83–1.12), children breast-fedfor 4–6 months had a 5% reduction in di-abetes risk (OR = 0.95, 95%CI 0.78–1.15),and children breast-fed for .6 monthshad a 6% reduction in diabetes risk (OR =0.94, 95% CI 0.76–1.17), though none ofthese differences were significant. Thesefindings were similar when restricted to13 studies with a low risk of bias.

Additional analyses (shown in Sup-plementary Table 2) in subgroups of stud-ies defined by geographic region(European, non-European) and low andhigh incidence rate countries (,15 casesof type 1 diabetes per 100,000 person-years and.15 per 100,000 person-years,respectively) did not reveal any markeddifferences in association.

There was little evidence of a differ-ence in the association between child-hood type 1 diabetes and (nonexclusive)breast-feeding in early diagnosed diabetes(i.e., ,5 years old) and later diagnoseddiabetes (i.e., between 5 and 15 yearsold) in studies in which both age groupswere available (shown in SupplementaryTable 3). Specifically, in 22 studies, theOR for breast-feeding for ;2 weeks ormore was 0.84 (95% CI 0.65–1.08) inearly diagnosed disease and 0.95 (95%CI 0.82–1.11) for later diagnosed disease.Similarly, in 23 studies, the OR for breast-feeding for ;3 months or more was 0.86(95% CI 0.69–1.08) in early diagnoseddisease and 0.90 (95% CI 0.79–1.02) forlater diagnosed disease.

Supplementary Table 1 shows thefindings for (nonexclusive) breast-feedingfor;2weeks ormore after adjustment forpotential confounders. The associationbetween type 1 diabetes and breast-feedingwas little altered after individual adjustmentfor maternal diabetes, birth weight, ges-tational age, maternal age, birth order,


Cardwell and Associates







Table

1dCha

racteristics

ofinclud

edstud

iespu

blished20

04–20

11investigatingtheassociationbetweenbreast-feeding

andchild

hood-onset

type

1diabetes,

orderedby

publicationda

te

Firstauthor,

year

(reference

a )Design

Cou

ntry

Typ

e1diabetes

Con

trols

Metho

dof

data

collectionc

(delay

years)

Cutoffu

sed

in3months

breast-feeding

analysis

Cutoffu

sed

in3mon

thexclusivebreast-

feedinganalysise

Ascertainment

metho

d(yearcases

diagnosed)

Age

atdx

(years)

nbRespo

nse

rate(%

)Source

(matching

criteria,w

here

used)

nbRespo

nse

rate(%

)

Ponson

by,

2011

(R1)

Coh

ortAustralia

Tasmaniandiabetes

register

(198

8–20

04)

0–16

2396

Tasmanianinfanthealth

survey

9,85

696

I(0)

3mon

ths

ex3mon

ths

Rabiei,20

11(R2)

C-C

Iran

IranianDiabetesAssociation(?)

0–15

100

100

Prim

aryandguidance

schoo

l20

010

0Q

(0–15

)3mon

ths

ex3mon

ths

Borras,2

011

(R3)

C-C

Spain

Catalan

register

ofdiabetes

(198

7–20

03)

0–10

349

88Registerof

newbo

rnscreeningformetabolic

diseases

(birth

year)

1,22

410

0R

Salem,2

010

(R4)

C-C

Egypt

Clin

icattendance

(200

5to

2006

)0–

1845

7?

Unk

nown

400

??

ex3mon

ths

Rob

ertson

,20

10(R5)

C-C

Scotland

Scottish

diabetes

register

(198

4–20

05)

0–14

5416

Aberdeenmaternity

records(birth

year)

154

16R

Bener,20

09(R6)

C-C

Qatar

Hospitalo

utpatient’s

clinic(?)

0–16

170

78Attendingprim

aryhealth

care

physician(age,sex,

ethn

icity

)17

075

Q(0–16

)

Mem

on,2

009

(R7)

C-C

Kuwait

Kuw

aitchild

hood

diabetes

register

(198

3–20

03)

0–19

376

94Attendingprim

arycare/

generalpractice

(birth

year,sex,n

ationality,

area)

376

92Q

(0–19

)3mon

ths

ex3mon

ths

Savilahti,2

009

(R8)

Coh

ortFinland

Nationalhealth

institu

tetreatm

entd

ata(1995–2006)

0–12

3175

Trialbirthcoho

rtused

d4,60

475

Q(0)

4mon

ths

ex4mon

ths

Karavanaki,

2008

(R9)

C-C

Greece

Hospitaladm

ission

(199

9to

2000

)0–

1512

195

Hospitaladm

ission

sexclud

ingchronic

diseases

150

100

Q(2–15

)4mon

ths

Benson

,200

8(R10

)C-C

Canada

Prince

Edw

ardIslanddiabetes

register

(200

1–20

04)

0–23

5286

ProvincialMedicare

Registry(age,sex)d

9573

I(1–23

)3mon

ths

Rosenbauer,

2008

(R11

)C-C

Germany

Nationw

ideho

spital-based

surveillance(199

2–19

95)

0–4

760

71Lo

calregistrationoffices

(age,sex,area)

1,86

843

Q(0–4)

7weeks

ff7weeks

Ievins,2

007

(R12

)Coh

ortEng

land

Hospitaladm

ission

[ICD

diabetescode](1963–1999)

0–14

288

fOxfordshire/W

est

Berkshirematernity

records

17,567

2f

R

Tenconi,

2007

(R13

)C-C

Italy

Paviatype1diabetes

register

(198

8–20

00)

0–19

9685

Hospital(age,sex,week)

202

98I(0–19

)ex

3mon

ths

Tzeng

,200

7(R14

)C-C

Taiwan

Hospitalattendance

(?)

7–14

54?

Scho

olmates

ofcases

(age,sex)

194

?Q

(2–25

)Yg

reda,20

06(R15

)C-C

Peru

Hospitalattendance

(?)

0–18

84?

Hospital

82?

?

Contin

uedon

p.2219



Tab

le1d

Con

tinu

ed

Firstauthor,

year

(reference

a )Design

Cou

ntry

Typ

e1diabetes

Con

trols

Metho

dof

data

collectionc

(delay

years)

Cutoffu

sed

in3months

breast-feeding

analysis

Cutoffu

sed

in3mon

thexclusive

breast-

feedinganalysise

Ascertainment

metho

d(yearcases

diagnosed)

Age

atdx

(years)

nbRespo

nse

rate(%

)Source

(matching

criteria,w

here

used)

nbRespon

serate(%

)

Wei,2

006

(R16

)C-C

Taiwan

Scho

ol-based

urinescreening

program

andqu

estion

naire

(199

2–19

97)

0–18

269

87Rando

mlyselected

negativesfrom

screeningprogram

d49

388

I(6–18

)ex

2mon

ths

Hatho

ut,

2006

(R17

)C-C

UnitedStates

Attendingho

spital(?)

0–22

102

?Hospitalw

ell-child

clinic

300

?I(0–22

)Malcova,2

006

(R18

)C-C

Czech

Repub

licCzech

child

hood

diabetes

register

(199

6–20

00)

0–14

738

76Scho

olmates

ofcases(age)

1,34

673

Q(0–18

)3mon

ths

ex3mon

ths

Radon

,200

5(R19

)C-C

Germany

Pediatricreferralcenter

admission

(?)

1–14

276

91Su

rgerypatients(hospital)

247

91Q

(6–16

)ff4mon

ths

Svensson

,20

05(R20

)C-C

Denmark

Danishregister

ofchild

hood

diabetes

(199

6–19

99)

0–14

666

81Danishpo

pulationregister

(age,sex)

471

48Q

(0–18

)4mon

ths

ff4mon

ths

Sipetic,2

005

(R21

)C-C

Serbia

BelgradeHospitaladmission

(199

4–19

97)

0–16

105

89Hospitalo

utpatientswith

skindisease(age,sex,area)

210

100

Q(0–16

)3mon

ths

ex3mon

ths

Strotm

eyer,

2004

(R22

)C-C

China

14region

altype

1diabetes

incidenceregisters

(198

5–19

98)

0–15

292

90Friends

ofcases(area)

496

?Q

(0–15

)4mon

ths

Marshall,20

04(R23

)C-C

England

Morecom

beBay/East

Lancashirediabetes

clinics(199

8)0–

1519

683

HealthAuthorities

(sex,

birthdate)

381

53Q

(0–15

)4mon

ths

ex4mon

ths

Sadauskaite-

Kuehne,

2004

(R24

)

C-C

Sweden

SoutheastSw

eden

type

1diabetes

register

(199

5–20

00)

0–15

509

100

Popu

lation

register

d1,12

294

Q(0–15

)3mon

ths

ex3mon

ths

C-C

Lithuania

Lithuanian

type1diabetes

register

(199

6–20

00)

0–15

283

100

Outpatient

clinicd

809

73Q

(0–15

)3mon

ths

ex3mon

ths

Stene,20

03(R25

)C-C

Norway

NorwegianChildhoo

dDiabetesRegistry

(199

7–20

00)

0–14

491

73Norwegianpo

pulation

registry

1,55

756

Q(0–14

)3mon

ths

ex3mon

ths

Perez-Bravo,

2003

(R26

)C-C

Chile

Childhoo

ddiabetes

register

(199

9–20

01)

0–14

410

?Scho

ols(socioecon

omic

cond

itions)

204

?Q

(0–14

)3mon

ths

Visalli,

2003

(R27

)C-C

Italy

Laziotype

1diabetes

register

(198

9–19

95)

0–14

130

100

Scho

ols(age)d

703

91Q

(0–14

)3mon

ths

DeStefano

,20

01(R28

)C-C

UnitedStates

Healthmaintenance

organization[ICDcode]

(198

8–19

97)

0–10

203

100

Healthmaintanence

organization(sex,

dateof

birth)

d66

010

0R

4mon

ths

Esferjani,200

1(R29

)C-C

Iran

Hospitalattendance(?)

0–14

54?

Hospitalo

utpatient

(age,sex,location,

socialclass)

51?

Q(0–14

)ff3weeks

Contin

uedon

p.2220



Tab

le1d

Con

tinu

ed

Firstauthor,

year

(reference

a )Design

Cou

ntry

Typ

e1diabetes

Con

trols

Metho

dof

data

collectionc

(delay

years)

Cutoffu

sed

in3months

breast-feeding

analysis

Cutoffu

sed

in3mon

thexclusive

breast-

feedinganalysise

Ascertainment

metho

d(yearcases

diagnosed)

Age

atdx

(years)

nbRespo

nse

rate(%

)Source

(matching

criteria,w

here

used)

nbRespon

serate(%

)

Stene,20

00(R30

)C-C

Norway

NorwegianChildhoo

dDiabetesRegistry

(198

9–19

98)

0–14

8294

Norwegianpo

pulation

registry

d1,05

573

Q(1–17

)3mon

ths

ex3mon

ths

Hyp

pönen,

2000

(R31

)C-C

Finland

Nationald

iabetesregister

(198

6–19

89)

0–14

435

94Finn

ishPopu

lation

Registry(birth

date,sex)d

386

85Q

(0–14

)ff3mon

ths

McK

inney,

1999

(R32

)C-C

England

YorkshireChildho

odDiabetes

Register(199

3to

1994

)0–

1522

092

Generalpractice

(age,sex)d

433

82Q

(0–15

)3mon

ths

ex3mon

ths

C-C

Austria

Vienn

atype1diabetesregister

(198

9–19

94)

0–14

101

89Scho

ols(age,sex)d

349

80Q

(0–14

)3mon

ths

ex3mon

ths

EURODIAB,

1999

(R33

)C-C

Bulgaria

W.B

ulgariatype

1diabetes

register

(199

1–19

94)

0–14

125

73Scho

olsand

policlin

ics(age)d

435

79Q

(0–14

)3mon

ths

ex3mon

ths

C-C

Latvia

Latviantype

1diabetesregister

(198

9–19

94)

0–14

140

99Po

pulation

register

(age)d

317

79Q

(0–14

)3mon

ths

ex3mon

ths

C-C

Lithuania

Lithuanian

type1diabetes

register

(198

9–19

94)

0–14

112

94Po

liclin

ics(age)d

262

73Q

(0–14

)3mon

ths

ex3mon

ths

C-C

Luxembo

urg

Luxembo

urg

type

1diabetes

register

(198

9–19

95)

0–14

5810

0Preschoo

lsand

scho

ols(age)d

173

95Q

(0–14

)3mon

ths

ex3mon

ths

C-C

Northern

Ireland

NorthernIrelandtype

1diabetes

register

(199

0–19

92)

0–14

182

79Generalpractice

459

64Q

(0–14

)3mon

ths

ex3mon

ths

C-C

Rom

ania

Bucharesttype1diabetes

register

(198

9–19

94)

0–14

8174

Preschoo

lsandscho

ols

(age)d

277

81Q

(0–14

)3mon

ths

ex3mon

ths

Gim

eno,

1997

(R34

)C-C

Brazil

Diabetesassociation/hospital

admission

(199

5)0–

1934

491

Unclear

(neigh

borhoo

d,sex,age)

345

100

I(0–19

)ex

4mon

ths

Wadsw

orth,

1997

(R35

)C-C

United

Kingd

omBritish

PediatricAssociatio

nSu

rveillance

Unit(199

2)0–

521

689

HealthAuthority

Immun

izationRegister

(age,sex)d

320

70Q

(0–5)

4mon

ths

ex4mon

ths

Melon

i,19

97(R36

)C-C

Sardinia

Hospitalattendance(?)

0–15

100

?Hospital(age,sex)

100

?Q

(0–15

)3mon

ths

ff3mon

ths

C-C,case-con

trol;dx

,diagno

sis;ex,exclusivebreast-feeding;

ff,form

ula-fed;I,interview

question

naire;Q,po

stal

questio

nnaire;

R,medical

ormaternity

records;Resp,respon

se.a Sup

plem

entalDatacontains

acorrespo

ndingreferencelistfor

includ

edstudies.bNum

berincludedinanalysisofbreast-feeding.

c Metho

dofbreast-feedingdatacollection(range

ofdelayfrom

birthtobreastfeedingdatacollection).dLo

wrisk

ofbias

study.eApp

roximatecutoffs

forother

analyses

used

ineach

study

show

nin

Figs.1

and2.

f Not

applicable.



Figure 1dPooled analysis of association between (nonexclusive) breast-feeding and childhood-onset type 1 diabetes in studies investigating ;2weeks (nonexclusive) breast-feeding and studies investigating any measure of breast-feeding.



Caesarean section delivery, or socio-economic status, where possible. Therewas also little change in the estimatewhen adjustment for all available variablessimultaneously was conducted.

Overall findings for exclusivebreast-feedingSimilar analyses were conducted for the as-sociations between exclusive breast-feedingand type 1 diabetes (investigated by 33studies including 7,621 patients with type1 diabetes) that are shown in Fig. 2 (whichcontains the duration of exclusive breast-feeding used in each study) and Table 1.Overall, in studies which investigated anymeasure of exclusive breast-feeding, therewas a reduction in the risk of type 1 di-abetes in exclusively breast-fed childrenof 26% (OR = 0.74, 95% CI 0.64–0.84;P, 0.001), but there was marked hetero-geneity between studies (I2 = 67%; hetero-geneity P , 0.001). When restricting thisanalysis to 13 studies with a low riskof bias, the association was no longer sig-nificant (OR = 0.89, 95% CI 0.78–1.02;P = 0.09), and there was less heterogeneity(I2 = 6%; heterogeneity P = 0.38).

As before, analysis was conducted insubgroups of studies recording;2 weeksor more (see Fig. 1 for definition) and;3months ormore of exclusive breast-feeding(see Table 1 for definition). In 20 stud-ies, children exclusively breast-fed for ;2weeks or more had a 25% reduction intheir risk of type 1 diabetes (OR = 0.75,95% CI 0.64–0.88; P = 0.001) comparedwith children exclusively breast-fed for,2 weeks, but there was marked hetero-geneity between studies (I2 = 58%; het-erogeneity P , 0.001). This associationwas slightly attenuated but remained signif-icant in 12 studies with a low risk of bias(OR = 0.86, 95% CI 0.75–0.99; P = 0.04),but in these studies, there was less hetero-geneity (I2 = 0%; heterogeneity P = 0.44).There was less evidence of a reduction indiabetes risk (in 30 studies) in children ex-clusively breast-fed for;3months ormore(OR = 0.87, 95% CI 0.75–1.00; P = 0.06)compared with children exclusively breast-fed for ,3 months but there was markedheterogeneity (I2 = 76%; heterogeneityP , 0.001). In 13 studies with a lowrisk of bias, the heterogeneity remained(I2 = 43%; heterogeneity P = 0.03), andthere was little evidence of an association(OR = 1.13, 95%CI 0.96–1.33; P = 0.15).Insufficient data were available to conductan analysis of exclusive breast-feeding infiner duration categories. Funnel plots forthese associations (not shown) providedT

able

2dPo

oled

analyses

oftheassociationbetweenbreast-feeding

andchild

hood-onset

type

1diab

etes

inallstud

iesan

din

stud

ieswithalower

risk

ofbias

Allstudies

Stud

ieswith

lowrisk

ofbias

d

Nb

Cases

OR(95%

CI)

Pvalue

Heterogeneity

x2(P

value)

I2in

%(95%

CI)

Nb

Cases

OR(95%

CI)

Pvalue

Heterogeneity

x2(P

value)

I2in

%(95%

CI)

Non

exclusive

breast-feedinga

Any

vs.n

onec

439,87

40.81

(0.72–

0.92

)0.00

111

2.16

(,0.00

1)63

(48–

73)

162,91

81.00

(0.89–

1.11

)0.93

15.20(0.51)

1(0–53

)$2vs.,

2wks

e28

6,79

80.93

(0.81–

1.07

)0.32

58.32(,

0.00

1)54

(29–

70)

152,34

31.00

(0.87–

1.15

)0.99

15.16(0.37)

8(0–44

)$3vs.,

3mos

e29

6,68

30.88

(0.78–

1.00

)0.05

87.10(,

0.00

1)68

(53–

78)

152,33

40.99

(0.86–

1.14

)0.92

27.00(0.02)

48(6–71

),2wks

e24

6,18

51.00

(ref.cat.)

132,04

51.00

(ref.cat.)

2wks–4mos

0.96

(0.83–

1.12

)0.60

38.11(0.03)

40(2–63

)1.03

(0.86–

1.22

)0.77

13.06(0.37)

8(0–46

)4–

6mos

0.95

(0.78–

1.15

)0.59

43.55(0.006

)47

(15–

67)

1.09

(0.90–

1.32

)0.38

7.02

(0.86)

0(0–57

).6mos

0.94

(0.76–

1.17

)0.57

69.28(,

0.00

1)67

(49–

78)

1.09

(0.83–

1.44

)0.52

26.75(0.01)

55(16–

76)

Exclusive

breast-feedinga

Any

vs.n

onec

337,62

10.74

(0.64–

0.84

),0.00

196

.22(,

0.00

1)67

(52–

77)

132,18

70.89

(0.78–

1.02

)0.09

12.84(0.38)

6(0–59

)$2vs.,

2wks

f20

4,38

80.75

(0.64–

0.88

)0.00

145

.10(0.001

)58

(30–

74)

121,91

80.86

(0.75–

0.99

)0.04

11.07(0.44)

0(0–58

)$3vs.,

3mos

f30

7,31

20.87

(0.75–

1.00

)0.06

122.64

(,0.00

1)76

(66–

83)

132,19

01.13

(0.96–

1.33

)0.15

23.00(0.03)

43(0–70

)

ref.cat.,reference

category;w

ks,w

eeks;m

os,m

onths.

a For

approxim

atecategories

used

ineach

study,see

Fig.1(for

anyno

nexclusive

breast-feedingandno

nexclusive

breast-feedingfor2weeks)andFig.2(for

any

exclusivebreast-feedingandexclusivebreast-feedingfor2

weeks)and

Tables1and2(for

anyno

nexclusive

for3

mon

thsor

exclusivebreast-feedingfor3

mon

ths).bNum

bero

fstudies.cAny

measureofbreast-feedingor

exclusivebreast-feeding

versus

none.dStud

ieswith

lowrisk

ofbias

indicatedby

footno

tein

Table1.

e Lessthan

2weekcategory

and,3mon

thcategory

includ

esno

breast-feeding.fLessthan

2weekcategory

and,3

mon

thcategory

includ

esno

exclusivebreast-feeding.



Figure 2dPooled analysis of association between exclusive breast-feeding and childhood-onset type 1 diabetes in studies investigating ;2 weeksexclusive breast-feeding and studies investigating any measure of exclusive breast-feeding.



little evidence of asymmetry (potentiallycaused by publication bias).

Additional analyses (shown in Sup-plementary Table 2) in subgroups ofstudies defined by geographic region (Eu-ropean, non-European) and incidencerates of type 1 diabetes (low incidence,high incidence) did not reveal anymarked differences in association.

As before, a comparison of the asso-ciations with exclusive breast-feeding byearly diagnosed diabetes (i.e.,,5 years old)and later diagnosed diabetes (i.e., between5 and 15 years old) showed little evidenceof a difference in studies in which both agegroups were available (shown in Supple-mentary Table 3). Specifically, in 18 stud-ies, the OR for exclusive breast-feeding for;2weeks ormorewas 0.73 (95%CI 0.58–0.91) in early diagnosed disease and 0.75(95% CI 0.60–0.95) for later diagnoseddisease. Similarly, in 24 studies, the ORfor exclusive breast-feeding for;3 monthsor more was 0.86 (95% CI 0.71–1.03) inearly diagnosed disease and 0.90 (95% CI0.76–1.05) for later diagnosed disease.

Supplementary Table 1 shows thefindings for exclusive breast-feedingfor ;2 weeks or more after adjustmentfor potential confounders. The asso-ciation between type 1 diabetes and ex-clusive breast-feeding was little alteredafter adjustment for any of these con-founders individually, where possible, orwhen adjustments were conducted for allavailable confounders simultaneously.

CONCLUSIONSdThis pooled analysissuggests that breast-feeding exclusivelyin the early weeks of life could reducethe risk of childhood onset type 1 diabetesby 15%, based upon the highest qualitystudies. This pooled analysis provideslittle evidence that longer exclusive ornonexclusive breast-feeding has a protec-tive effect. However, firm conclusions aredifficult to reach because of the markedheterogeneity in the observed associationsbetween studies and the weaknesses in-herent inmany of the included studies. Theobserved association between exclusivebreast-feeding and type 1 diabetes couldnot be explained by confounding fromestablished risk factors for diabetes in-cluding maternal diabetes, birth weight,gestational age, maternal age, birth order,or Caesarean section delivery.

This is, to our knowledge, the firstpooled analysis of individual participantdata on breast-feeding and diabetes. Themain strengths of using individual partic-ipant data in this study was that it allowed

analysis to be conducted in similar cate-gories of breast-feeding duration be-tween studies and enabled adjustment ofassociations for various confounders.This analysis contains data from 43 stud-ies and includes 9,874 patients with type1 diabetes allowing the power to identifyassociations of relatively small magni-tude. Also, the analyses benefited fromcoverage of 95% of eligible cases fromstudies identified by the comprehensivesearch strategies. However, it remainspossible that studies with relevant datawere not identified by the searches. Aweakness of this pooled analysis was thatthe majority of studies were questionnaire-based case-control studies, and theirfindings are based upon breast-feedinginformation usually recalled over manyyears, potentially resulting in recall andother biases. There were only two cohortstudies (8,9) that prospectively recordedbreast-feeding at 2 weeks and 3 months,but these contained relatively few patientswith type 1 diabetes, 31 and 24 cases, re-spectively. Consequently, there is a needfor further studies prospectively orroutinely recording breast-feeding inearly life. Although attempts were madeto identify studies with lower risk ofbias (with higher response rates and ran-domly selected population-based con-trols), even in these studies, responserates varied, and in many, recall bias ispossible. As with all observational stud-ies, it is impossible to rule out the effectof unmeasured confounding, for example,by other infant food intake.

The association between breast-feedingand type 1 diabetes differed markedlybetween studies. This could reflect realdifferences in the association in differentpopulations or biases specific to eachstudy. In studies with a low risk of bias,less heterogeneity was observed, allowingmore confidence to be placed on theresulting pooled estimates.

The findings of our pooled analysis ofindividual participant data differ fromtwo published literature-based meta-analyses (3,4). The first meta-analysis(3) (containing studies conducted priorto 1994) demonstrated a reduction in di-abetes risk of 30%with breast-feeding.3months, and a 35% reduction with exclu-sive breast-feeding for .3 months. Thesecond meta-analysis (containing studiesconducted prior to 1996, many of whichwere in the first meta-analysis) demon-strated a reduction in diabetes risk of20% with breast-feeding for 3 monthsand 40% for exclusive breast-feeding for

.3 months (4). In our pooled analysis,there was little evidence of any reductionin the risk of diabetes after 3 months ofbreast-feeding (exclusive or nonexclusive).Unfortunately, neither of the two earliermeta-analyses presented estimates forbreast-feeding in the first 2 weeks of lifeto allow comparison with our results.Both these meta-analyses were largelybased upon questionnaire-based case-control studies, and the second meta-analysis (4) highlighted that the observedassociation could reflect methodologicalweaknesses, particularly recall bias. Theauthors suggested that mothers of chil-dren with diabetes may more critically re-member occasions when they deviatedfrom the breast-feeding recommendationsof health care personnel (4). As previouslydiscussed, themajority of studies includedin this pooled analysis were questionnaire-based case-control studies and would alsobe subject to this bias.

In 2007, the Agency for HealthcareResearch and Quality reviewed the evi-dence of the effect of breast-feeding ontype 1 diabetes (13). However, the reportdid not contain a meta-analysis becausetheir searches identified only 6 studiesafter 1996 (up to 2007) when the pre-vious meta-analyses were conducted, incontrast to the 43 studies included in ourpooled analysis between 1996 and 2011.The estimated OR of 0.75 for any breast-feeding for 3 months from the Agency forHealthcare Research and Quality report(13) (based solely on the two previousmeta-analyses) has since been used to es-timate the cost-effectiveness of breast-feeding in the U.S. (14). Our results wouldsuggest that this OR of 0.75 may overesti-mate any protective effect against type 1diabetes and that consequently the cost-effectiveness of breast-feeding with re-spect to type 1 diabetes could also beexaggerated. A recent narrative review (15)commenting on the effects of breast-feedingand complementary feeding noted thatthe associations were controversial andthat contradictory outcomes had been ob-served. Studies have also been conductedintob-cell autoimmunity andbreast-feeding,but the results have been mixed, withsome studies observing protective effectsof breast-feeding, whereas others have ob-served no association [see Knip et al. (15)for references].

This pooled analysis suggests thatinitial exclusive breast-feeding may re-duce diabetes risk, indicating that theintroduction of formula feeding may ex-plain any association. The responsible








mechanism is unclear, but researchershave speculated (15) that breast-feedingcould protect against type 1 diabetes riskby increasing gut permeability; previousstudies have shown that gut permeabilitydecreases faster in breastfed children (16)or by reducing the risk of enterovirus in-fections (17), which have been associatedwith increased diabetes risk (18). Alterna-tively, any increased risk of diabetes asso-ciated with formula feeding could reflectearly exposure to a variety of cow’s milkproteins [studies have shown childrenwith type 1 diabetes have increased con-centrations of antibodies to dietary anti-gens (19)], modification of the gutmicroflora (15), and increased weightgain in early life [although one study sug-gested that any effect of breast-feedingwas independent of weight gain (20)].In our analysis, an association was onlyobserved for initial exclusive breast-feeding (for $2 weeks) but not for ex-clusive breast-feeding for $3 months.This distinction may be real or couldreflect the relatively accurate maternalrecall of initial breast-feeding but lessaccurate recall of the duration of feed-ing, as previously suggested (3). If earlyintroduction of formula feeding is re-sponsible then ongoing efforts to inves-tigate alternative formula, such as theTrial to Reduce Insulin-dependent Diabe-tes Mellitus in the Genetically at Risk (21),may be productive.

In conclusion, our pooled analysissuggests that children who are initiallyexclusively breast-fed have a small reduc-tion in their risk of type 1 diabetes. Thisfinding is difficult to interpret because theassociations varied markedly betweenstudies and because the results are basedprimarily upon questionnaire-based case-control studies, which could have beenaffected by recall and other biases.

AcknowledgmentsdThis work was sup-ported by National Institutes of Health grantsR01-DK-46498 and R01-DK-42316, the Chi-nese Foundation of Health, a grant (DOH91-TD1167) from the Department of Health,Executive Yuan, Republic of China, theMinistryfor Science and Technology of the Republicof Serbia through Contract 175042 (2011–2014), a grant from the German ResearchFoundation (HE 234/1-1), and the ResearchCouncil of Norway (148359/330).No potential conflicts of interest relevant to

this article were reported.C.R.C., L.C.S., J.L., J.R., O.C., J.S., F.P.-B.,

A.M., S.G.G., E.J.K.W., E.S.S., M.J.G., K.R.,L.-M.C., R.C.P., A.C., K.K., G.B., P.P., B.U.,

E.S., G.D., S.S., G.J., C.I.-T., C.E.d.B., K.H.,V.B., E.S., A.-L.P., M.S., S.R., and C.C.P. re-searched data. C.R.C., L.C.S., J.R., K.R., A.C.,S.S., E.S., and M.S. were responsible for sta-tistical analysis. C.R.C. wrote the manuscript.C.R.C., L.C.S., J.L., J.R., O.C., J.S., F.P.-B.,A.M., S.G.G., E.J.K.W., E.S.S., M.J.G., K.R.,L.-M.C., R.C.P., A.C., K.K., G.B., P.P., B.U.,E.S., G.D., S.S., G.J., C.I.-T., C.E.d.B., K.H.,V.B., E.S., A.-L.P., M.S., S.R., and C.C.P. re-viewed and edited the manuscript. C.R.C. isthe guarantor of this work and, as such, hadfull access to all the data in the study and takesresponsibility for the integrity of the data andthe accuracy of the data analysis.The authors thank A. Pezic (Murdoch

Children’s Research Institute, Parkville,Victoria, Australia), R. Portuesi (Department ofEndocrinology and Diabetes, University Cam-pus Bio-Medico, Rome, Italy), E. Weintraub(Centers for Disease Control and Prevention,Atlanta, GA), R. Rastmanesh (Clinical Nutritionand Dietetics Department, Shahid BeheshtiUniversity of Medical Sciences, Tehran, Iran),J. Dorman (Department of Epidemiology,Graduate School of Public Health, Universityof Pittsburgh, Pittsburgh, PA), and M. Yang(Department of Epidemiology, Graduate Schoolof Public Health, University of Pittsburgh,Pittsburgh, PA) for support. The authors alsothank G. Soltész (University of Pecs, Pecs,Hungary) and G. Dahlquist (Umea University,Umea, Sweden), coordinators of the EURODIABSubstudy.

References1. Patterson CC, Dahlquist GG, Gyürüs E,

Green A, Soltész G; EURODIAB StudyGroup. Incidence trends for childhoodtype 1 diabetes in Europe during 1989-2003 and predicted new cases 2005-20:a multicentre prospective registrationstudy. Lancet 2009;373:2027–2033

2. Hirschhorn JN. Genetic epidemiology oftype 1 diabetes. Pediatr Diabetes 2003;4:87–100

3. Gerstein HC. Cow’s milk exposure andtype I diabetes mellitus. A critical over-view of the clinical literature. DiabetesCare 1994;17:13–19

4. Norris JM, Scott FW. A meta-analysis ofinfant diet and insulin-dependent di-abetes mellitus: do biases play a role?Epidemiology 1996;7:87–92

5. Stewart LA, Tierney JF. To IPD or not toIPD? Advantages and disadvantages ofsystematic reviews using individual patientdata. Eval Health Prof 2002;25:76–97

6. Owen CG, Whincup PH, Kaye SJ, et al.Does initial breastfeeding lead to lowerblood cholesterol in adult life? A quanti-tative review of the evidence. Am J ClinNutr 2008;88:305–314

7. Stukel TA, Demidenko E, Dykes J,Karagas MR. Two-stage methods for theanalysis of pooled data. Stat Med 2001;20:2115–2130

8. Savilahti E, Saarinen KM. Early infantfeeding and type 1 diabetes. Eur J Nutr2009;48:243–249

9. Ponsonby AL, Pezic A, Cochrane J, et al.Infant anthropometry, early life infection,and subsequent risk of type 1 diabetesmellitus: a prospective birth cohort study.Pediatr Diabetes 2011;12:313–321

10. DIAMOND Project Group. Incidence andtrends of childhood Type 1 diabetesworldwide 1990-1999. Diabet Med 2006;23:857–866

11. Dahlquist GG, Patterson C, Soltesz G.Perinatal risk factors for childhood type 1diabetes in Europe. The EURODIABSubstudy 2 Study Group. Diabetes Care1999;22:1698–1702

12. Sadauskaite-Kuehne V, Ludvigsson J,Padaiga Z, Jasinskiene E, Samuelsson U.Longer breastfeeding is an independentprotective factor against development oftype 1 diabetes mellitus in childhood. Di-abetes Metab Res Rev 2004;20:150–157

13. Ip S, Chung M, Raman G, et al. Breast-feeding and Maternal and Infant HealthOutcomes in Developed Countries. EvidenceReport/Technology Assessment No. 153.Rockville, MD, Agency for HealthcareResearch and Quality, 2007

14. Bartick M, Reinhold A. The burden ofsuboptimal breastfeeding in the UnitedStates: a pediatric cost analysis. Pediatrics2010;125:e1048–e1056

15. KnipM,Virtanen SM, AkerblomHK. Infantfeeding and the risk of type 1 diabetes. Am JClin Nutr 2010;91:1506S–1513S

16. CatassiC, BonucciA,CoppaGV,Carlucci A,Giorgi PL. Intestinal permeability changesduring the first month: effect of naturalversus artificial feeding. J Pediatr Gastro-enterol Nutr 1995;21:383–386

17. Sadeharju K, Knip M, Virtanen SM, et al.;Finnish TRIGR Study Group. Maternalantibodies in breast milk protect the childfrom enterovirus infections. Pediatrics2007;119:941–946

18. Yeung WC, Rawlinson WD, Craig ME.Enterovirus infection and type 1 diabetesmellitus: systematic review and meta-analysis of observational molecular stud-ies. BMJ 2011;342:d35

19. Savilahti E, Akerblom HK, Tainio VM,Koskimies S. Children with newly di-agnosed insulin dependent diabetes mel-litus have increased levels of cow’s milkantibodies. Diabetes Res 1988;7:137–140

20. Hyppönen E, Kenward MG, Virtanen SM,et al. Infant feeding, early weight gain, andrisk of type 1 diabetes. Childhood Di-abetes in Finland (DiMe) Study Group.Diabetes Care 1999;22:1961–1965

21. Knip M, Virtanen SM, Becker D, Dupré J,Krischer JP, Åkerblom HK; TRIGR StudyGroup. Early feeding and risk of type 1 di-abetes: experiences from the Trial to Re-duce Insulin-dependent diabetes mellitusin the Genetically at Risk (TRIGR). Am JClin Nutr 2011;94(Suppl.):1814S–1820S



Breast-Feeding and Childhood-Onset Type 1 Diabetes A...

Documents

Transcript of Breast-Feeding and Childhood-Onset Type 1 Diabetes A...