Breastfeeding and Reduced Risk of Sudden InfantDeath Syndrome: A Meta-analysis

abstractCONTEXT: Benets of breastfeeding include lower risk of postneonatalmortality. However, it is unclear whether breastfeeding specicallylowers sudden infant death syndrome (SIDS) risk, because study re-sults have been conicting.

OBJECTIVE: To perform a meta-analysis to measure the associationbetween breastfeeding and SIDS.

METHODS: We identied 288 studies with data on breastfeeding andSIDS through aMedline search (19662009), review articles, andmeta-analyses. Twenty-four original case-control studieswere identied thatprovided data on the relationship between breastfeeding and SIDS risk.Two teams of 2 reviewers evaluated study quality according to presetcriteria; 6 studies were excluded, which resulted in 18 studies foranalysis. Univariable and multivariable odds ratios were extracted. Asummary odds ratio (SOR) was calculated for the odds ratios by usingthe xed-effect and random-effect inverse-variance methods of meta-analysis. The Breslow-Day test for heterogeneity was performed.

RESULTS: For infants who received any amount of breast milk for anyduration, the univariable SOR was 0.40 (95% condence interval [CI]:0.350.44), and themultivariable SOR was 0.55 (95% CI: 0.440.69). Forany breastfeeding at 2 months of age or older, the univariable SOR was0.38 (95% CI: 0.270.54). The univariable SOR for exclusive breastfeed-ing of any duration was 0.27 (95% CI: 0.240.31).

CONCLUSIONS: Breastfeeding is protective against SIDS, and this ef-fect is stronger when breastfeeding is exclusive. The recommendationto breastfeed infants should be includedwith other SIDS risk-reductionmessages to both reduce the risk of SIDS and promote breastfeedingfor its many other infant andmaternal health benets. Pediatrics 2011;128:000

AUTHORS: Fern R. Hauck, MD, MS,a John M. D. Thompson,PhD,b Kawai O. Tanabe, MPH,a Rachel Y. Moon, MD,c andMechtild M. Vennemann, MD, PhDd

aDepartment of Family Medicine, University of Virginia,Charlottesville, Virginia; bDepartment of Paediatrics; Child andYouth Health, University of Auckland, Auckland, New Zealand;cDivision of General Pediatrics and Community Health, ChildrensNational Medical Center and George Washington UniversitySchool of Medicine and Health Sciences, Washington, DC; anddInstitute of Legal Medicine, University of Muenster, Muenster,Germany

KEY WORDSSIDS, sudden infant death syndrome, risk factors, breastfeeding

ABBREVIATIONSSIDSsudden infant death syndromeORodds ratioSORsummary odds ratioCIcondence interval

The authors had full access to all of the data in the study andtake responsibility for the integrity of the data and the accuracyof the data analysis. The results, conclusions, andrecommendations presented in this article are endorsed by theEpidemiology Working Group of the International Society of thePrevention of Infant Deaths.

www.pediatrics.org/cgi/doi/10.1542/peds.2010-3000

doi:10.1542/peds.2010-3000

Accepted for publication Mar 14, 2011

Address correspondence to Fern R. Hauck, MD, MS, Departmentof Family Medicine, University of Virginia School of Medicine, POBox 800729, Charlottesville, VA 22908. E-mail: frh8e@virginia.edu

PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275).

Copyright 2011 by the American Academy of Pediatrics

FINANCIAL DISCLOSURE: The authors have indicated they haveno nancial relationships relevant to this article to disclose.

REVIEW ARTICLES

PEDIATRICS Volume 128, Number 1, July 2011 1

There are many physical and emo-tional benets to breastfeeding,1 in-cluding a reduced risk of postneonatalmortality.2 However, it is unclearwhether breastfeeding specicallylowers the risk of sudden infant deathsyndrome (SIDS). Physiologic sleepstudies have shown that breastfed in-fants have lower arousal thresholdsthan formula-fed infants,3,4 which mayprovide a mechanism for protectionagainst SIDS. However, epidemiologicstudies have been inconsistent inshowing a protective effect of breast-feeding on the risk of SIDS; some studyresults have supported a protective ef-fect,511 and others have not.2,5,8,1219

The authors of a meta-analysis andqualitative literature review publishedin 2000 concluded that there was a sta-tistically signicant increase in SIDSrisk for bottle-fed infants.20 These au-thors, however, dened SIDS loosely(as any sudden and unexplained deathin an infant or young child) and in-cluded studies in which the denitionsof breastfeeding exposure differed,and there were other methodologicaws. A more recent meta-analysisconducted by the Agency for Health-care Research and Quality analyzed 6studies and found a statistically signif-icant decrease in SIDS in infants whowere ever breastfed compared with in-fants who were never breastfed (ad-justed summary odds ratio [SOR]: 0.64[95% condence interval (CI): 0.510.81]).21 We performed our meta-analysis to quantify and evaluate theprotective effect of breastfeedingagainst SIDS, including the inuence ofexclusive breastfeeding and longerbreastfeeding duration, and to make arecommendation on the potential util-ity of breastfeeding as a strategy forreducing the risk of SIDS. Our hypothe-ses were that (1) breastfeeding is as-sociated with a decreased risk of SIDSand (2) exclusive breastfeeding andbreastfeeding for longer duration areassociated with the greatest reduction

in risk. This report of our methods andndings follows the guidelines for re-porting meta-analyses of observa-tional studies proposed by Stroup etal22 and the PRISMA Group.23

METHODS

Data Sources and Study Selection

We searched the Ovid Medline data-base (January 1966 through December2009) to collect data on breastfeedingand its association with SIDS. Thesearch strategy included published ar-ticles limited to humans with the Med-ical Subject Headings terms suddeninfant death and breast feedingwithkey words sudden infant death syn-drome, SIDS, cot death, andbreastfeeding. Combining searchesresulted in 265 abstracts (Fig 1). Anadditional 23 studies were identiedthrough review articles and meta-analyses, for a total of 288 studies.These studies were reviewed by teamsof 2 independent reviewers who evalu-ated each abstract for relevance onthe basis of title and abstract. One hun-

dred eighty-four reports were ex-cluded on the basis of the abstractsalone, and 104 articles were pulled forfurther review. Two reviewers (Dr Hauckand Ms Tanabe) reviewed all pulled arti-cles for inclusion and for overlappingdata. Twenty-four original case-controlstudies that provided data on the rela-tionship between breastfeeding andSIDS risk2,511,1419,2433 were identied.Two teams of 2 reviewers indepen-dently evaluated 12 studies each ac-cording to preset criteria; in cases ofdisagreement, a third reviewer evalu-ated the study, and a consensus opin-ion was reached.

Because the search was not limited toarticles written in English, 12 of the ar-ticles were in other languages (3 Ger-man, 3 Italian, 1 Japanese, 1 Spanish, 1Polish, and 3 Norwegian). After review-ing the articles and abstracts, either inthe original language or the Englishtranslation, none of these studies wasdeemed relevant except 1 Norwegianstudy, for which an article published in

265 abstracts identified in Ovid Medline

23 studies identified through review articles and meta-analyses

288 abstracts reviewed

184 reports excludeda

104 articles pulled for full review

24 articles included in second-level review

80 articles excluded for not meeting first-level inclusion criteriab

6 articles excluded for not meeting second-level inclusion criteriac

18 articles included in meta-analysis

FIGURE 1Study inclusion and exclusion ow diagram. a Exclusion criteria: duplication, no apparent relevance.b First-level inclusion criteria: articles that reported an association between breastfeeding and SIDS.c Second-level inclusion criteria: see criteria listed in Table 1; an additional study was excluded for notproviding ORs that could be used in calculating an SOR.

2 HAUCK et al

English with the same content replacedthe one published in Norwegian.

Data Extraction

The teams evaluated the eligible stud-ies on the basis of the 6 criteria devel-oped by the American Academy of Pe-diatrics Task Force on Positioning andSIDS for its literature review on the re-lationship between sleeping positionand SIDS. These criteria are (1) an ap-propriate denition for SIDS, (2) autop-sies performed in 98% of cases, (3)an adequate description of SIDS ascer-tainment in the study population, (4)matched control subjects, (5) an ade-quate description of the process of con-trol selection, and (6) inclusion of suf-cient data to calculate ORs and 95% CIsor inclusion of the actual ORs and CIs. Inour review, 19 of 24 studies satised all 6criteria; the failed criteria of 5 studiesare listed in Table 1.2,9,16,19,32 Anotherstudy could not be included becausethe ORs were presented in a way thatwas not compatible with our analy-ses,33 which resulted in a total of 18studies for the meta-analysis.

Statistical Analysis

Several different denitions forbreastfeeding were examined: (1)breastfeeding of any amount (partialor exclusive) or duration, includingbreastfeeding at discharge from hos-pital (any breastfeeding); (2) breast-feeding of any amount at the age of 2months or older (breastfeeding 2months); and (3) exclusive breast-feeding (ie, no formula supplementa-tion) for any duration (exclusivebreastfeeding).

The univariable and multivariable ORswere extracted from each study for thedifferent associations between breast-feeding and SIDS. A separate SOR wascalculated for the univariable andmul-tivariable ORs by using the xed-effectand random-effect inverse-variancemethods of meta-analysis. The Breslow-

Day test for heterogeneity was per-formed. A P value of.05 was consid-ered to indicate that heterogeneitywas present. Analyses were conductedindependently by 2 authors (DrsThompson and Vennemann), one by us-ing RevMan 5.0 (Nordic Cochrane Cen-tre, Cochrane Collaboration, Copenha-gen, Denmark) and one by usingComprehensive Meta Analysis 2.2.048(Biostat, Englewood, NJ). Any discrep-ancies were investigated and resolved.

RESULTS

Eighteen case-control studies were in-cluded in the meta-analysis (Table 1),and data for any breastfeeding wereprovided for all of them.* The forestplot for the univariable ORs with therandom-effect model is shown in Fig 2;the SOR was 0.40 (95% CI: 0.350.44;I2 71%). Multivariable ORs were re-ported for only 7 of the 23 stud-ies5,7,8,10,17,18,34; a univariable pooledanalysis of the results from these 7studies resulted in an OR of 0.36 (95%CI: 0.310.42), which is consistent withthe results when all 18 studies wereincluded. The multivariable pooled es-timate revealed a movement of the ORtoward the null; however, it remainedstatistically signicant at 0.55 (95% CI:0.440.69) (Fig 3). There was no heter-ogeneity (I2 40%).

Three studies provided informationabout any breastfeeding at 2months ofage or older.5,11,15 The summary uni-variable estimate for the 3 studies was0.38 (95% CI: 0.270.54; I2 78%). Be-cause only 2 of the studies providedmultivariable ORs,5,11 meta-analysis toobtain a summary multivariable esti-mate was not performed.

Eight studies provided informationon exclusive breastfeeding of any du-ration.57,14,15,26,29,34 The univariable SORwas 0.27 (95% CI: 0.240.31; I2 87%)(Fig 4). None of these studies provid-

ed multivariable ORs for exclusivebreastfeeding.

As noted previously, 5 studies failed tomeet 1 or more quality criteria.2,9,16,19,32

A sensitivity analysis was conducted todetermine the SORs for any breast-feeding with these 5 studies included.The resulting univariable SOR was 0.49(95% CI: 0.450.53). The multivariableSOR was 0.68 (95% CI: 0.580.80).These results are slightly higher thanthe SORs that excluded the respectivestudies.

DISCUSSION ANDRECOMMENDATIONS

Our meta-analysis of 18 studies re-veals that breastfeeding to any extentand of any duration is protectiveagainst SIDS. The protective effect isstronger for exclusive breastfeeding.The summary multivariable OR sug-gests that breastfeeding itself is pro-tective and not merely a marker ofother potentially protective factorssuch as the absence of smoke expo-sure or sociodemographic factors.Therefore, we recommend that moth-ers breastfeed their infants as a poten-tial way to reduce their risk of SIDS.Ideally, breastfeeding should be exclu-sive (ie, formula should not be given)for at least 4 to 6months and should becontinued until the infant is at least 1year of age. Exceptions to this recom-mendation include conditions underwhich breastfeeding is contraindi-cated, such as for infants whose moth-ers use illegal drugs.35 This recommen-dation is consistent with the AmericanAcademy of Pediatrics policy state-ment on breastfeeding and the use ofhuman milk, which endorses exclusivebreastfeeding to 6 months and contin-uation for at least the rst year of life.35

Some breastfeeding advocates haveexpressed concern that promotion ofother factors shown in epidemiologicstudies to be protective against SIDS,such as pacier use and room-sharing*Refs 58, 10, 11, 14, 15, 17, 18, 2430, and 34.

REVIEW ARTICLES

PEDIATRICS Volume 128, Number 1, July 2011 3

TABLE1StudiesIncludedintheMeta-analysis

Study(Year)

Country

Yearsof

Study

Total

Cases,N

Total

Controls,N

Breastfeeding

Cases,n(%)

Breastfeeding

Controls,n(%)

CrudeOR

(95%CI)

Covariatesa

TimetoInterviewb

Failed

Criteria

Anybreastfeeding

BartholomewandMacArthur24(1988)

Scotland

Unknown

7979

15(19)

25(32)

0.51(0.231.13)

2wkafterdeath

None

Naeyeetal25(1976)

UnitedStates

19521966

125

375

16(13)

60(16)

0.77(0.411.44)

Nointerview

None

Biering-Srensenetal15(1978)

Denmark

19561971

123

520

98(75)

480(92)

0.33(0.180.59)

Nointerview

None

SteeleandLangworth26(1966)

Canada

19601961

80157

21(26)

75(48)

0.39(0.210.73)

NAc

None

Protestosetal27(1973)

UnitedKingdom

19601972

94135

22(23)

71(53)

0.28(0.150.52)

Nointerview

None

Murphyetal28(1982)

UnitedKingdom

19651973,

19751977

9947223

16(16)

14223(30)

0.54(0.291.01)

NA

None

GriceandMcGlashan29(1981)

Tasmania

19701976

121

153

53(44)

101(66)

0.40(0.240.67)

Nointerview

None

McGlashan30(1989)

Tasmania

19801986

167

334

115(69)

252(75)

0.72(0.471.11)

Upto5wkafterdeath

None

Fordetal6(1993)

NewZealand

19871990

356

1529

275(77)

1371(90)

0.39(0.290.53)

Within1moofdeath

None

Gilbertetal14(1995)

UnitedKingdom

19871989,

19901991

98196

56(57)

144(73)

0.48(0.270.84)

2interviews;immediately

andat23mo

None

Ponsonbyetal18(1995)

Tasmania

19881991

58120

22(38)

63(53)

0.50(0.260.98)1,4

6wkafterdeath

None

Klonoff-CohenandEdelstein31(1995)

UnitedStates

19891992

200

200

114(57)

151(76)

0.43(0.270.68)3,4,6,13,17

NANone

Mitchelletal5(1997)

NewZealand

19911993

120

918

98(82)

809(88)

0.60(0.351.03)16,7,9,11,12,19

NANone

Wennergrenetal10(1997)

Denmark,Norway,

Sweden

19921995

244

863

184(75)

729(84)

0.59(0.410.85)4,6,9

NANone

Schellscheidtetal11(1997)

Germany

19931994

58156

29(50)

129(83)

0.21(0.100.44)4,6,15,16,19

Within2wk

None

Flemingetal17(1996)

UnitedKingdom

19931995

195

780

88(45)

470(60)

0.50(0.350.71)17,9,10,14,16,

18,19

2interviews;within5d

and2wkofdeath

None

Haucketal8(2003)

UnitedStates

19931996

260

260

55(21)

130(50)

0.20(0.120.35)1,5,11,13

2wkafterdeath

None

Vennemannetal7(2009)

Germany

19982001

333

998

165(50)

827(83)

0.19(0.140.25)17,15,16,18

1moafterdeath

None

Breastfeeding2mo

Biering-Srensenetal15(1978)

Denmark

19561971

97503

22(23)

278(55)

0.24(0.140.41)

Nointerview

None

Mitchelletal5(1997)

NewZealand

19911993

64778

46(72)

600(77)

0.76(0.411.39)16,7,9,11,12,19

NANone

Schellscheidtetal11(1997)

Germany

19931994

58156

7(12)

50(32)

0.29(0.110.74)4,6,15,16,19

Within2wkofdeath

None

Studiesexcludedfornotmeeting

eligibilitycriteria

Fedrick31(1974)

UnitedKingdom

19661970

154

409

63(37)

157(31)

1.11(0.751.65)

Nointerview

13

Watsonetal16(1981)

UnitedKingdom

19751979

308

236

164(53)

161(68)

0.53(0.370.76)

2interviews;immediately

and3wklater

13

ChenandRogan2(2004)

UnitedStates

1988

591

7740

187(32)

3073(40)

0.70(0.590.84)1,3,4,5,9,12,19

NA14

Stray-Pedersenetal19(2005)

Norway

19892003

2372

16(73)

65(90)

0.29(0.090.97)4,7

1moafterdeath

4,5

Jonville-Braetal9(2001)

France

19951997

111

341

37(33)

160(47)

0.57(0.350.92)3,4,6,9,14,17,19

2interviews;dayofdeath

and3moafterdeath

2

Almetal33(2002)

Denmark,Norway,

Sweden

19921995

239

841

109(46)

626(74)

0.29(0.210.39)46,8

NANonec

Studieswerescoredonthefollowingcriteria:(1)anappropriatedenitionforSIDS;(2)autopsiesperformedin98%ofcases;(3)anadequatedescriptionofSIDSascertainmentinthestudypopulation;(4)matchedcontrolsubjects;(5)anadequate

descriptionoftheprocessofcontrolselection;and(6)inclusionofsufcientdatatocalculateORsand95%CIsortheactualORsandCIswereprovided.NAindicatesnotavailable,notprovided.

aCovariates:1,maternalage;2,parity;3,birthweight;4,infantexposuretotobaccosmoke(beforeorafterdelivery);5,factorsrelatedtosocioeconomicstatus;6,infantsleepposition;7,bed-sharing;8,infantage;9,infantgender;10,gestation;11,

maritalstatus;12,race/ethnicity;13,factorsrelatedtoprenatalcare;14,factorsrelatingtosurfaceonwhichinfantwasplaced;15,pillow/cushionuse;16,factorsrelatedtooverheating;17,postneonatalinfanthealthproblems;18,pacieruse;19,other.

bTimetointerviewwasdenedastimefrominfantsdeathoridenticationofcontrolstointerviewwithparents.

cORsinthearticlewereprovidedforintervalsofbreastfeedingduration;thus,wecouldnotidentifyanORtouseincalculatingtheSORforthismeta-analysis.

4 HAUCK et al

Study or SubgroupBartholomew and MacArthur24 (1988)Biering-Srensen et al15 (1978)Fleming et al17 (1996)Ford et al6 (1993)Gilbert et al14 (1995)Grice and McGlashan29 (1981)Hauck et al8 (2003)Klonoff-Cohen and Edelstein31 (1995)McGlashan30 (1989)Mitchell et al5 (1997)Murphy et al28 (1982)Naeye et al25 (1976)Ponsonby et al18 (1995)Protestos et al27 (1973)Schellscheidt et al11 (1997)Steele and Langworth26 (1966)Vennemann et al7 (2009)Wennergren et al10 (1997)

Total (95% CI)Heterogeneity: = 57.66, df = 17 (P < .00001); I = 71%Test for overall effect: z = 16.35 (P < .00001)

log[]-0.67334-1.10866-0.69315-0.94161-0.73397-0.91629-1.60944-0.84397

-0.3285-0.51083-0.61619-0.26136-0.69315-1.27297-1.56065-0.94161-1.66073

-0.530268

SE0.4038280.2984880.1804420.1538260.2864810.2618980.2802580.231851

0.219230.2753520.319582

0.320470.3410770.3121880.3720960.3178410.147913

0.18724

Weight2.0%3.6%9.8%

13.5%3.9%4.7%4.1%6.0%6.7%4.2%3.1%3.1%2.8%3.3%2.3%3.2%

14.6%9.1%

100.0%

IV, Fixed, 95% CI0.51 [0.231.13]0.33 [0.180.59]0.50 [0.350.71]0.39 [0.290.53]0.48 [0.270.84]0.40 [0.240.67]0.20 [0.120.35]0.43 [0.270.68]0.72 [0.471.11]0.60 [0.351.03]0.54 [0.291.01]0.77 [0.411.44]0.50 [0.260.98]0.28 [0.150.52]0.21 [0.100.44]0.39 [0.210.73]0.19 [0.140.25]0.59 [0.410.85]

0.40 [0.350.44]

IV, Fixed, 95% CI

0.01 0.1 1 10 100Favors breastfeeding Favors not breastfeeding

FIGURE 2Univariable analysis of any breastfeeding versus no breastfeeding.

Study or SubgroupFleming et al17 (1996)Hauck et al8 (2003)Klonoff-Cohen and Edelstein31 (1995)Mitchell et al5 (1997)Ponsonby et al18 (1995)Vennemann et al7 (2009)Wennergren et al10 (1997)

Total (95% CI)Heterogeneity: = 10.08, df = 6 (P = .12); I = 40%Test for overall effect: z = 5.28 (P < .00001)

log[]0.058269-0.91629

-0.89159812-0.07257-0.15082-0.84397

-0.693147

SE0.3176570.319582

0.33463050.4203370.4012450.2393540.21979

Weight12.6%12.4%11.4%7.2%7.9%

22.2%26.3%

100.0%

IV, Fixed, 95% CI1.06 [0.571.98]0.40 [0.210.75]0.41 [0.210.79]0.93 [0.412.12]0.86 [0.391.89]0.43 [0.270.69]0.50 [0.330.77]

0.55 [0.440.69]

IV, Fixed, 95% CI

0.01 0.1 1 10 100Favors breastfeeding Favors not breastfeeding

FIGURE 3Multivariable analysis of any breastfeeding versus no breastfeeding.

Study or SubgroupBiering-Srensen et al15 (1978)Ford et al6 (1993)Gilbert et al14 (1995)Grice and McGlashan29 (1981)Klonoff-Cohen and Edelstein31 (1995)Mitchell et al5 (1997)Steele and Langworth26 (1966)Vennemann et al7 (2009)

Total (95% CI)Heterogeneity: = 54.32, df = 7 (P < .00001); I = 87%Test for overall effect: z = 18.11 (P < .00001)

log[]-0.91629-1.89712-0.74149-0.91879-0.77436-0.43078-0.94161-1.60944

SE0.3247360.130313

0.308950.249950.20420.2325

0.2952690.149946

Weight4.9%

30.5%5.4%8.3%

12.4%9.6%5.9%

23.0%

100.0%

IV, Fixed, 95% CI0.40 [0.210.76]0.15 [0.120.19]0.48 [0.260.87]0.40 [0.240.65]0.46 [0.310.69]0.65 [0.411.03]0.39 [0.220.70]0.20 [0.150.27]

0.27 [0.240.31]

IV, Fixed, 95% CI

0.01 0.1 1 10 100Favors breastfeeding Favors not breastfeeding

FIGURE 4Univariable analysis of exclusive breastfeeding of any duration.

REVIEW ARTICLES

PEDIATRICS Volume 128, Number 1, July 2011 5

without bed-sharing, is inconsistentwith promotion of breastfeeding. Al-though some observational studieshave revealed an association betweenpacier use and decreased breast-feeding duration,3638 this associationwas not borne out by several random-ized clinical trials3941 and 1 systematicreview.42 The American Academy of Pe-diatrics policy statements on breast-feeding and the use of human milk35

and SIDS43 both indicate that pacierscan be used by breastfed infants oncebreastfeeding has been well estab-lished. Mother-infant bed-sharing (orsleeping in the same bed) is often pro-moted as a way to increase breast-feeding rates4446; however, althoughbed-sharing is associated with in-creased breastfeeding duration, it isunclear whether the practice of bed-sharing increases the practice ofbreastfeeding or if parents whochoose to breastfeed subsequently de-cide to bed-share.47 Room-sharingwithout bed-sharing (sleeping in thesame room with the infants crib orbassinet close to the parents bed) isrecommended for all infants as a wayto reduce the risk of SIDS and acciden-tal suffocation while facilitating feed-ing and monitoring of the infant.43 Onestudy from the Netherlands revealedthat the benets of breastfeeding do notoutweigh the increased risk of SIDS as-sociated with bed-sharing.48 Additionalstudies to analyze the contribution ofmultiple simultaneous factors (such asbed-sharing and breastfeeding or paci-er use and breastfeeding) to SIDS riskare needed.

Although causation cannot be provenin case-control studies, on which theseresults are based, the factors thathave been proposed to support causal-ity in observational studies are foundin this meta-analysis: (1) consistentndings; (2) strong association; (3)dose-response effect; (4) causal factorpreceding the outcome; and (5) biolog-

ical plausibility.49 Although the studieswere from many different countries,and heterogeneous populations wererepresented, the individual ORs forbreastfeeding in relation to SIDS weresimilar. The association betweenbreastfeeding and SIDS risk reductionis strong, there is a dose response,and the causal factor (ie, breastfeed-ing) precedes the outcome. The pro-tective effect of breastfeeding againstSIDS also has biological plausibility.Breastfed infants are more easilyaroused from active sleep thanformula-fed infants at 2 to 3 months ofage, which is within the 2- to 4-monthpeak age during which SIDS occurs.4

Breastfeeding also confers immuno-logic advantages over formula feedingby providing immunoglobulins and cy-tokines that may protect infants dur-ing the vulnerable period for SIDS,when their own production of immuno-globulin G is low and their maternallyacquired levels are decreasing. Infantswho die from SIDS often have had aminor infection in the days precedingdeath that was not sufcient alone tohave caused death. These infectionsmay induce proinammatory cyto-kines that may cause respiratory orcardiac dysfunction, fever, shock, hy-poglycemia, and arousal decits.7,50 Al-though the possibility of reverse cau-sality cannot be ruled out entirely (ie,certain infants may be difcult tobreastfeed because of underlyinghealth conditions that may make themmore susceptible to SIDS), most SIDSdeaths occur in previously healthy in-fants; therefore, it would not likely ac-count for many of the SIDS deaths. An-other potential concern is thatinadequate recall of breastfeeding dura-tion may bias results. However, the timeto interview after the infant death in theincluded studies was generally short.

The 2005 American Academy of Pediat-rics policy statement on SIDS did notendorse breastfeeding as a means to

reduce the risk of SIDS because of theinsufcient strength of evidence avail-able at that time.43 Although therewere several studies that had found aprotective effect of breastfeeding, af-ter controlling for possible confound-ing factors, the protective effect hadbeen eliminated for some, so clearconclusions could not be drawn. Stud-ies published since that statement,which are included in our currentmeta-analysis, notably the more de-tailed analysis of Vennemann et al,7

showed a strongly protective effect ofbreastfeeding even after controllingfor confounders. The meta-analysis byIp et al21 consisted of many but not allof the studies included in our currentanalysis, and our ndings were similarto theirs. In the Ip et al analysis, everbreastfeeding was associated withcrude and adjusted SORs of 0.41 (95%CI: 0.280.58) and 0.64 (95% CI: 0.510.81), respectively. The authors did notreport results for exclusive breast-feeding or specic durations.

A potential limitation of our meta-analysis is that studies fromwhich sig-nicant associations are reportedmaybe preferentially published, whichcould result in an overestimate of thetrue effect and could bias the results.22

It is unlikely that this was the case forseveral reasons. The breastfeeding re-sults were all part of larger studiesthat examined potential risk and pro-tective factors for SIDS; thus, resultswere published along with other nd-ings. There was some heterogeneity ofresults, which indicates that resultswere not selectively reported. Studiespublished in languages other than Eng-lish were included. Finally, we attendinternational SIDS meetings regularly,participate in SIDS organizations andListservs, and have frequent contactwith SIDS researchers around theworld; we are not aware of other un-published studies that would contra-dict these ndings. A limitation identi-

6 HAUCK et al

ed by this meta-analysis was thesmall number of studies that pre-sented data on breastfeeding dura-tion, and when presented, there weredifferent ways in which duration wasdened, which made it difcult to poolthe results. This is an area that needsfurther investigation.

CONCLUSIONS

There are many known benets tobreastfeeding, and breastfeedingshould be recommended for all new-born infants to enhance maternal and

infant well-being. The best time to be-gin the dialogue with mothers aboutbreastfeeding plans is the prenatal pe-riod, and it should be included withother SIDS risk-reduction messagesand materials that are traditionallygiven to expectant mothers duringpregnancy. The same benets ofbreastfeeding in protecting againstSIDS are found for black infants as forthose in other groups.8 However,breastfeeding initiation and continua-tion occur less frequently among blackmothers and those of other racial/eth-

nic minorities and among socially dis-advantaged mothers.51 In addition,these same groups have a higher inci-dence of SIDS.52 Thus, it is essentialthat breastfeeding interventions tar-get these higher-risk populations,53,54

and future research should focus ondeveloping and evaluating innovativeintervention methods. All health pro-fessionals should speak in 1 voiceabout the importance of breastfeed-ing, which now adds SIDS risk reduc-tion to its long list of maternal and in-fant health benets.

REFERENCES

1. Lawrence RA. A Review of the Medical Ben-ets and Contraindications to Breastfeed-ing in the United States (Maternal and ChildHealth Technical Information Bulletin). Ar-lington, VA: National Center for Education inMaternal and Child Health; 1997

2. Chen A, Rogan WJ. Breastfeeding and therisk of postneonatal death in the UnitedStates. Pediatrics. 2004;113(5). Available at:www.pediatrics.org/cgi/content/full/113/5/e435

3. Franco P, Scaillet S, Wermenbol V, Valente F,Groswasser J, Kahn A. The inuence of apacier on infants arousals from sleep. JPediatr. 2000;136(6):775779

4. Horne RS, Parslow PM, Ferens D, Watts AM,Adamson TM. Comparison of evoked arous-ability in breast and formula fed infants.Arch Dis Child. 2004;89(1):2225

5. Mitchell EA, Tuohy PG, Brunt JM, et al. Riskfactors for sudden infant death syndromefollowing the prevention campaign in NewZealand: a prospective study. Pediatrics.1997;100(5):835840

6. Ford RP, Taylor BJ, Mitchell EA, et al. Breast-feeding and the risk of sudden infant deathsyndrome. Int J Epidemiol. 1993;22(5):885890

7. Vennemann MM, Bajanowski T, BrinkmannB, et al; GeSID Study Group. Does breast-feeding reduce the risk of sudden infantdeath syndrome? Pediatrics. 2009;123(3).Available at: www.pediatrics.org/cgi/content/full/123/3/e406

8. Hauck FR, Herman SM, Donovan M, et al.Sleep environment and the risk of suddeninfant death syndrome in an urbanpopulation: the Chicago Infant MortalityStudy. Pediatrics . 2003;111(5 pt 2):12071214

9. Jonville-Bra AP, Autret-Leca E, Barbeillon F,

Paris-Llado J; French Reference Centers forSIDS. Sudden unexpected death in infantsunder 3 months of age and vaccinationstatus: a case-control study. Br J Clin Phar-macol. 2001;51(3):271276

10. Wennergren G, Alm B, yen N, et al. The de-cline in the incidence of SIDS in Scandinaviaand its relation to risk-intervention cam-paigns. Acta Paediatr. 1997;86(9):963968

11. Schellscheidt J, Ott A, Jorch G. Epidemiolog-ical features of sudden infant death after aGerman intervention campaign in 1992. EurJ Pediatr. 1997;156(8):655660

12. Tappin D, Brooke H, Ecob R, Gibson A. Usedinfant mattresses and sudden infant deathsyndrome in Scotland: case-control study.BMJ. 2002;325(7371):10071009

13. Kraus JF, Greenland S, Bulterys M. Risk fac-tors for sudden infant death syndrome inthe US Collaborative Perinatal Project. Int JEpidemiol. 1989;18(1):113120

14. Gilbert RE, Wigeld RE, Fleming PJ, Berry PJ,Rudd PT. Bottle feeding and the sudden in-fant death syndrome. BMJ. 1995;310(6972):8890

15. Biering-Srensen F, Jrgensen T, Hilden J.Sudden infant death in Copenhagen19561971. I. Infant feeding. Acta PaediatrScand. 1978;67(2):129137

16. Watson E, Gardner A, Carpenter RG. An epi-demiological and sociological study of unex-pected death in infancy in nine areas ofsouthern England. I: Epidemiology. Med SciLaw. 1981;21(2):7888

17. Fleming PJ, Blair PS, Bacon C, et al. Environ-ment of infants during sleep and risk of thesudden infant death syndrome: results of19935 case-control study for condentialinquiry into stillbirths and deaths in in-fancy. Condential Enquiry Into Stillbirths

and Deaths Regional Coordinators and Re-searchers. BMJ. 1996;313(7051):191195

18. Ponsonby AL, Dwyer T, Kasl SV, Cochrane JA.The Tasmanian SIDS Case-Control Study:univariable and multivariable risk factoranalysis. Paediatr Perinat Epidemiol. 1995;9(3):256272

19. Stray-Pedersen A, Arnestad M, Vege A,Sveum L, Rognum TO. Bed sharing and sud-den infant death [in Norwegian]. TidsskrNor Laegeforen. 2005;125(21):29192921

20. McVea KL, Turner PD, Peppler DK. The role ofbreastfeeding in sudden infant death syn-drome. J Hum Lact. 2000;16(1):1320

21. Ip S, Chung M, Raman G, et al. Breastfeedingandmaternal and infant health outcomes indeveloped countries. Evid Rep Technol As-sess (Full Rep). 2007;(153):1186

22. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies inepidemiology: a proposal for reporting.Meta-analysis of Observational Studies inEpidemiology (MOOSE) group. JAMA. 2000;283(15):20082012

23. Moher D, Liberati A, Tetzlaff J, Altman DG;PRISMA Group. Preferred reporting itemsfor systematic reviews and meta-analyses:the PRISMA statement. Ann Intern Med.2009;151(4):264269, W64

24. Bartholomew S, MacArthur BA. Comparisonof infants dying from the sudden infantdeath syndromewithmatched live controls.Soc Sci Med. 1988;27(4):393397

25. Naeye RL, Ladis B, Drage JS. Sudden infantdeath syndrome: a prospective study. Am JDis Child. 1976;130(11):12071210

26. Steele R, Langworth JT. The relationship ofantenatal and postnatal factors to suddenunexpected death in infancy. CanMed AssocJ. 1966;94(22):11651171

27. Protestos CD, Carpenter RG, McWeeny PM,

REVIEW ARTICLES

PEDIATRICS Volume 128, Number 1, July 2011 7

Emery JL. Obstetric and perinatal historiesof children who died unexpectedly (cotdeath). Arch Dis Child. 1973;48(11):835841

28. Murphy JF, Newcombe RG, Sibert JR. Theepidemiology of sudden infant death syn-drome. J Epidemiol Community Health.1982;36(1):1721

29. Grice AC, McGlashan ND. Obstetric factors in171 sudden infant deaths in Tasmania,19701976. Med J Aust. 1981;1(1):2631

30. McGlashan ND. Sudden infant deaths in Tas-mania, 19801986: a seven year prospec-tive study. Soc Sci Med . 1989;29(8):10151026

31. Klonoff-Cohen HS, Edelstein SL. A case-control study of routine and death scenesleep position and sudden infant death syn-drome in southern California. JAMA. 1995;273(10):790794

32. Fedrick J. Sudden unexpected death in in-fants in the Oxford record linkage area: de-tails of pregnancy, delivery, and abnormal-ity in the infant. Br J Prev Soc Med. 1974;28(3):164171

33. Alm B, Wennergren G, Norvenius SG, et al.Breast feeding and the sudden infant deathsyndrome in Scandinavia, 199295. Arch DisChild. 2002;86(6):400402

34. Klonoff-Cohen HS, Edelstein SL, Lefkowitz ES,et al. The effect of passive smoking and to-bacco exposure through breast milk onsudden infant death syndrome. JAMA. 1995;273(10):795798

35. Gartner LM, Morton J, Lawrence RA, et al;American Academy of Pediatrics, Section onBreastfeeding. Breastfeeding and the use ofhuman milk. Pediatrics . 2005;115(2):496506

36. Aarts C, Hrnell A, Kylberg E, Hofvander Y,Gebre-Medhin M. Breastfeeding patterns inrelation to thumb sucking and pacier use.

Pediatrics. 1999;104(4). Available at: www.pediatrics.org/cgi/content/full/104/4/e50

37. Benis MM. Are paciers associated withearly weaning frombreastfeeding? Adv Neo-natal Care. 2002;2(5):259266

38. Scott JA, Binns CW, OddyWH, GrahamKI. Pre-dictors of breastfeeding duration: evidencefrom a cohort study. Pediatrics. 2006;117(4). Available at: www.pediatrics.org/cgi/content/full/117/4/e646

39. Kramer MS, Barr RG, Dagenais S, et al. Pac-ier use, early weaning, and cry/fussbehavior: a randomized controlled trial.JAMA. 2001;286(3):322326

40. Collins CT, Ryan P, Crowther CA, McPhee AJ,Paterson S, Hiller JE. Effect of bottles, cups,and dummies on breast feeding in preterminfants: a randomised controlled trial. BMJ.2004;329(7459):193198

41. Jenik AG, Vain NE, Gorestein AN, Jacobi NE;Pacier and Breastfeeding Trial Group.Does the recommendation to use a pacierinuence the prevalence of breastfeeding?J Pediatr. 2009;155(3):350.e1354.e1

42. OConnor NR, Tanabe KO, Siadaty MS, HauckFR. Paciers and breastfeeding: a system-atic review. Arch Pediatr AdolescMed. 2009;163(4):378382

43. American Academy of Pediatrics, TaskForce on Sudden Infant Death Syndrome.The changing concept of sudden infantdeath syndrome: diagnostic coding shifts,controversies regarding the sleeping envi-ronment, and new variables to consider inreducing risk. Pediatrics. 2005;116(5):12451255

44. Academy of Breastfeeding Medicine Proto-col Committee. ABM clinical protocol #6:guideline on co-sleeping and breastfeeding.Revision, March 2008. Breastfeed Med.2008;3(1):3843

45. McKenna J. Sleeping with your baby. NewBeginnings. 2009;26(1):49

46. McKenna J. Sleeping With Your Baby: A Par-ents Guide to Cosleeping. Washington, DC:Platypus Media, LLC; 2007

47. Horsley T, Clifford T, Barrowman N, et al. Ben-ets and harms associated with the practiceof bed sharing: a systematic review. Arch Pe-diatr Adolesc Med. 2007;161(3):237245

48. Ruys JH, de Jonge GA, Brand R, EngelbertsAC, Semmekrot BA. Bed-sharing in the rstfour months of life: a risk factor for suddeninfant death. Acta Paediatr. 2007;96(10):13991403

49. Hill AB. Statistical evidence and inference.InPrinciples of Medical Statistics. Vol 9. NewYork, NY: Oxford University Press; 1971:309323

50. Blackwell CC, Weir DM. The role of infectionin sudden infant death syndrome. FEMS Im-munol Med Microbiol. 1999;25(12):16

51. Grummer-Strawn L, Scanlon KS, Darling N,Conrey EJ. Racial and socioeconomic dis-parities in breastfeeding: United States,2004. MMWR Morb Mortal Wkly Rep. 2006;55(12):335339

52. Mathews TJ, MacDorman MF. Infant mortal-ity statistics from the 2006 period linkedbirth/infant death data set. Natl Vital StatRep. 2010;58(17):131

53. MacGregor E, Hughes M. Breastfeeding ex-periences of mothers from disadvantagedgroups: a review. Community Pract. 2010;83(7):3033

54. Centers for Disease Control and Prevention.Racial and ethnic differences in breastfeed-ing initiation and duration, by State: Na-tional Immunization Survey, United States,20042008. MMWR Morb Mortal Wkly Rep.2010;59(11):327334

8 HAUCK et al

Breastfeeding and Reduced Risk of Sudden Infant Death Syndrome: A Meta-analysisMETHODSData Sources and Study SelectionData ExtractionStatistical Analysis

RESULTSDISCUSSION AND RECOMMENDATIONSCONCLUSIONSREFERENCES