Evidence on the long-term effects of breastfeeding

<strong>Evidence</strong> on the long-termeffects of breastfeedingSYSTEMATIC REVIEWS AND META-ANALYSESBernardo L. Horta, MD, PhDUniversidade Federal de Pelotas, Pelotas, BrazilRajiv Bahl, MD, PhDDepartment of Child and Adolescent Health and Development,World Health Organization, Geneva, SwitzerlandJosé C. Martines, MD, PhDDepartment of Child and Adolescent Health and Development,World Health Organization, Geneva, SwitzerlandCesar G. Victora, MD, PhDUniversidade Federal de Pelotas, Pelotas, Brazili

ivEVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

Executive summaryBackground: Breastfeeding presents clear short-term benefits for child health, mainly protectionagainst morbidity and mortality from infectious diseases. On the other hand, there is somecontroversy on the long-term consequences of breastfeeding. Whereas some studies reported thatbreastfed subjects present a higher level of school achievement and performance in intelligencetests, as well as lower blood pressure, lower total cholesterol and a lower prevalence of overweightand obesity, others have failed to detect such associations.Objectives: The primary objective of this series of systematic reviews was to assess the effects ofbreastfeeding on blood pressure, diabetes and related indicators, serum cholesterol, overweight andobesity, and intellectual performance.Search strategy: Two independent literature searches were conducted at the World HealthOrganization in Geneva, Switzerland, and at the University of Pelotas in Brazil, comprising theMEDLINE (1966 to March 2006) and Scientific Citation Index databases.Selection criteria: We selected observational and randomized studies, published in English, French,Portuguese and Spanish, assessing the effects of breastfeeding on blood pressure, obesity/overweight,total cholesterol, type-2 diabetes, and intellectual performance. Studies that restricted the measurementof outcomes to infancy were excluded from the meta-analyses. The type of comparison group used(e.g. never breastfed or breastfed for less than x months) did not constitute a selection criterion.Data extraction and analysis: Two reviewers independently evaluated study quality, using astandardized protocol, and disagreement was resolved by consensus rating. Fixed and random-effectsmodels were used to pool the effect estimates, and a random-effects regression was used to assessseveral potential sources of heterogeneity.Effect on blood pressure: We included 30 and 25 estimates for systolic and diastolic blood pressure,respectively. In a random-effects model, systolic (mean difference: -1.21 mmHg; 95% confidenceinterval (CI): -1.72 to -0.70) and diastolic blood pressures (mean difference: -0.49 mm Hg; 95% CI:-0.87 to -0.11) were lower among breastfed subjects. Publication bias was evident, with smallerstudies reporting a greater protective effect of breastfeeding. However, even among studies with>1000 participants a statistically significant effect of breastfeeding was observed (mean differencein systolic blood pressure: -0.59 mmHg; 95% CI: -1.00 to -0.19). Adjustment for confounding wasalso a source of heterogeneity between study results, but even among those studies controlling forseveral socioeconomic and demographic variables, systolic (mean difference: -1.19; 95% CI: -1.70 to-0.69) and diastolic (mean difference: -0.61; 95% CI: -1.12 to -0.10) blood pressures were loweramong breastfed subjects. Publication bias and residual confounding may be responsible for part(but not all) of the observed effect of breastfeeding on blood pressure.Effect on serum cholesterol: Breastfed subjects presented lower mean total cholesterol in adulthood(mean difference: -0.18; 95% CI: -0.30 to -0.06 mmol/L), whereas for children and adolescents theassociation was not statistically significant. Age at assessment of cholesterol explained about 60% ofthe heterogeneity between studies, whereas study size, control for confounding, year of birth and1

categorization of breastfeeding duration did not play a significant role. The evidence suggests thatbreastfeeding is related to lower cholesterol levels and this association is not due to publication biasor residual confounding.Effect on overweight and obesity: We obtained 39 estimates of the effect of breastfeeding onprevalence of overweight/obesity. In a random-effects model, breastfed individuals were less likely tobe considered as overweight and/or obese, with a pooled odds ratio of 0.78 (95% CI: 0.72–0.84).Control for confounding, age at assessment, year of birth, and study design did not modify the effectof breastfeeding. Because a statistically significant protective effect was observed among thosestudies that controlled for socioeconomic status and parental anthropometry, as well as with >1500participants, the effect of breastfeeding was not likely to be due to publication bias or confounding.Effect on type-2 diabetes: We identified five papers that evaluated the relationship betweenbreastfeeding duration and type-2 diabetes. Breastfed subjects were less likely to present type-2diabetes (pooled odds ratio: 0.63; 95% CI: 0.45–0.89).Effect on intelligence and schooling: For the assessment of performance in intelligence tests, weobtained data from eight studies that controlled for intellectual stimulation at home and collectedinformation on infant feeding in infancy, in which the duration of breastfeeding was of at least onemonth among breastfed subjects. Performance in intelligence tests was higher among those subjectswho had been breastfed (mean difference: 4.9; 95% CI: 2.97–6.92). Positive studies included arandomized trial. Regarding school performance in late adolescence or young adulthood, three studiesshowed a positive effect of breastfeeding.Limitations: Because nearly all studies included in the analyses are observational, it is not possibleto completely rule out the possibility that these results may be partly explained by self-selection ofbreastfeeding mothers or by residual confounding. Publication bias was assessed by examining theeffect of study size on the estimates and was found not to be important for most outcomes. Veryfew studies were available from low/middle-income countries, where the effect of breastfeeding maybe modified by social and cultural conditions.Reviewers’ conclusions: The available evidence suggests that breastfeeding may have long-termbenefits. Subjects who were breastfed experienced lower mean blood pressure and total cholesterol,as well as higher performance in intelligence tests. Furthermore, the prevalence of overweight/obesityand type-2 diabetes was lower among breastfed subjects. All effects were statistically significant, butfor some outcomes their magnitude was relatively modest.2EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

I. IntroductionBreastfeeding brings clear short-term benefitsfor child health by reducing mortality and morbidityfrom infectious diseases. A collaborativereanalysis of studies conducted in middle/lowincomecountries reported a reduced risk ofmortality from infectious diseases amongbreastfed infants, up to the second birthday (1).Kramer et al (2) reviewed the evidence on theeffects on child health and growth of exclusivebreastfeeding for 6 months. Infants who wereexclusively breastfed for 6 months presentedlower morbidity from gastrointestinal and allergicdiseases, while showing similar growthrates to non-breastfed children.Based on such evidence, WHO (3) andUNICEF (4) now recommend that every infantshould be exclusively breastfed for the first sixmonths of life, with continued breastfeeding forup to two years or longer. In this review we addressthe long-term consequences of breastfeedingon adult health and intellectual development.Current interest in the long-term consequencesof early life exposures has been fuelledby the original finding of Barker et al (5) thatsize at birth and in infancy was related to thedevelopment of adult diseases – including diabetes,hypertension and cardiovascular conditions.These findings led to the fetal origin hypothesis,which postulates that adverse intrauterineconditions would be responsible for fetalmalnutrition and low birthweight, a process thatwould also increase the susceptibility to chronicdiseases in adulthood. Indeed, epidemiologicalstudies in several countries have reported increasedrisks of chronic diseases (6-8) amongadults who were small at birth.Because many studies on the long-term consequencesof intrauterine growth may be affectedby confounding variables – particularly socioeconomicstatus – and by inappropriate statisticalanalyses, some authors challenge whether or notthese associations are causal (9). On the otherhand, Lucas et al (10) pointed out that methodologicalflaws in studies of intrauterine growthmay have deflected attention from important postnatalexposures – such as infant growth and feedingpatterns – that could also be related to thedevelopment of chronic diseases.The notion that nutrition during early phasesof human development can alter organ function,and thereby predispose – or programme – individualsto a later onset of adult disease, is anarea of considerable interest to researchers andof great concern to public health. This idea originatesfrom the more general concept in developmentalbiology which was defined by Lucasas “programming” (11). This is defined as theprocess whereby a stimulus or insult applied ata critical or sensitive period of development resultsin a long-term or permanent effect on thestructure or function of the organism. This hypothesisis currently described as the “developmentalorigins of health and disease” (12).Over 400 scientific publications are availableon the association between breastfeedingand health outcomes beyond infancy. Some researchersclaim that the benefits of breastfeedinginclude increased school achievement or performancein intelligence tests, reduced meanblood pressure, lower total cholesterol, and alower prevalence of overweight and obesity. Onthe other hand, other studies have failed to detectsuch associations. The evidence on longtermeffects of breastfeeding may be importantfor further promotion of this healthy practicethroughout the world.The Department of Child and AdolescentHealth and Development in the World HealthOrganization commissioned the present systematicreview of the available evidence on longtermconsequences of breastfeeding. The followinglong-term outcomes of public health importancewere examined: blood pressure, diabetesand related indicators, serum cholesterol, overweightand obesity, and intellectual performance.These outcomes are of great interest to researchers,as evidenced by the number of publicationsidentified. This report describes the methods,results and conclusions of this review.3

II.General methodological issues in studiesof the long-term effects of breastfeedingStudy designThe strength of scientific inference depends onthe internal validity of the study. Randomizedcontrolled trials, if properly designed and conducted,are considered as the gold standard ofdesign validity, being less susceptible than otherdesigns to selection and information bias, as wellas to confounding (13). Furthermore, there areclearly defined standards for conducting andreporting on randomized clinical trials, all intendedto increase the validity of their resultsand interpretation (14).As mentioned previously, breastfeeding hasclear short-term benefits – i.e. it reduces morbidityand mortality from infectious diseases (1)– and it is now unethical to randomly allocateinfants to breastmilk or formula. However, about20 years ago the evidence supporting breastfeedingwas not so clear-cut and randomized trialscould be carried out. In a British study from1982, pre-term infants were randomly assignedto formula or banked breastmilk; recent followupof these subjects has provided an opportunityto assess whether breastfeeding can programmethe later occurrence of risk factors forcardiovascular diseases (15).Most recently, in the Promotion ofBreastfeeding Intervention Trial (16) in Belarus,maternal hospitals and their corresponding polyclinicswere randomly assigned to implement ornot to implement the Baby-Friendly HospitalInitiative. Duration and exclusivity ofbreastfeeding were higher in the interventiongroup (16). Because breastfeeding promotionwas randomized, rather than breastfeeding perse, the trial was ethically sound. Follow-up ofthese children will provide an excellent opportunityfor studying the long-term effects ofbreastfeeding. It should be noted, however, thatthe Belarus study has relatively low statisticalpower because compliance with breastfeedingpromotion was far from perfect.Search for evidence on the long-term consequencesof breastfeeding should not be restrictedto randomized trials, because of their smallnumber. It must involve tracking down the bestavailable studies with rigorous design; prospectivebirth cohort studies should be consideredas the next best design in terms of strength ofevidence. It has been shown that the results ofmeta-analysis of well-designed observationalstudies, with either cohort or case-control design,can be remarkably similar to that ofrandomized controlled trials on the same topic(17). Nevertheless, birth cohort studies are susceptibleto self-selection and confounding, issuesthat will be discussed below.In the next section, the main methodologicalissues affecting these studies will be described,as well as the strategies used to minimizethe impact of these limitations on the findingsof the present review.Factors affecting the internalvalidity of individual studiesIn the context of studies on long-term consequencesof breastfeeding, the following issuesshould be considered.Losses to follow-upA major source of bias in cohort studies as wellas randomized controlled trials with long-termoutcomes relates to the need for follow-up ofindividuals for a period of time after exposurein order to assess the occurrence of the outcomesof interest. If a large proportion of subjectsare lost during follow-up, the study’s validityis reduced. Baseline data, such as breastfeedingstatus, should be examined to determinewhether there are systematic differences betweensubjects who were followed up and thosewho were not; if the losses are similar accordingto the baseline characteristics, selection biasis unlikely (18).4EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

MisclassificationWhen the methods used for obtaining informationon either infant feeding or the outcomesare inaccurate, misclassification may occur. Thismay take two forms: differential and non-differentialmisclassification.Misclassification of breastfeeding durationis more likely in retrospective than in prospectivedesigns. Huttly et al (19), in a prospectivestudy, compared the actual breastfeeding durationwith the duration reported retrospectivelyby the mothers. They observed a systematic biastowards reporting longer durations ofbreastfeeding for wealthier and more educatedmothers, while those from low socioeconomicstatus families did not tend to err more in onedirection than in the other. Because high socioeconomicstatus is related to a lower prevalenceof cardiovascular diseases, such differentialmisclassification would exaggerate the long-termbenefits of breastfeeding.On the other hand, in nondifferentialmisclassification, measurement error is independentof exposure or outcome status. Thisleads to a dilution of the actual effect, becausesome breastfed subjects are classified as nonbreastfedand vice-versa. Consequently, one isless likely to detect an association, even if onereally exists.Confounding by socioeconomicstatusSocioeconomic status is one of the most importantconfounders in studies on the long-termeffects of breastfeeding. In most societies,breastfeeding rates differ among social groups(20). The direction of confounding by socioeconomicstatus may vary between high-incomeand low/middle-income populations. In high-incomecountries, breastfeeding mothers tend tobe of higher educational and socioeconomic status(21); other things being equal, their offspringwill have a lower prevalence of cardiovascularrisk factors and higher educational attainmentbecause they belong to the upper social classes.Consequently, confounding by socioeconomicstatus may overestimate the beneficial effectsof breastfeeding.On the other hand, in low/middle-income settingsbreastfeeding is often more commonamong the poor (22). Thus, confounding by socioeconomicstatus may underestimate the beneficialeffects of breastfeeding – for example, oneducational attainment – because breastfed subjectswill tend to be poorer. Depending on theassociation between cardiovascular risk factorsand wealth in these societies, confounding canact in either direction. For example, if high cholesterollevels are more frequent among the rich,the protective effect of breastfeeding will beoverestimated.Even if confounding factors are controlledthrough multivariable analyses, there is a possibilityof residual confounding. Inaccurate measurementof confounders, as well as incorrectspecification of statistical models, may precludefull adjustment for confounding and lead to estimatesof the impact of breastfeeding that arebiased. For example, if information on familyincome is not precise, control for this imperfectvariable will not fully account for the confoundingeffect of true income.As discussed above, the direction of residualconfounding may vary between high-income andless developed settings. These differences willbe explored when analysing the results of thisreview for each outcome.Self-selectionEven within the same social group, mothers whobreastfeed are likely to be more health-consciousthan those who do not breastfeed. This may alsolead them to promote other healthy habitsamong their children, including prevention ofoverweight, promotion of physical exercise andintellectual stimulation. This may be particularlytrue in high-income populations. Becausethese maternal attributes are difficult to measure,it is not possible to include them in theanalyses as confounding factors. Nevertheless,this possibility should be taken into accountwhen interpreting the study’s results.Adjustment for potential mediatingfactorsSeveral studies on the long-term consequencesof breastfeeding have adjusted their estimatesfor variables that may represent mediating factors,or links, in the causal chain leading frombreastmilk to the outcomes. Adjustment forGENERAL METHODOLOGICAL ISSUES IN STUDIES OF THE LONG-TERM EFFECTS OF BREASTFEEDING5

mediating factors will tend to underestimate theoverall effect of breastfeeding, e.g. adjusting forweight at the time of blood pressure measurementwhen evaluating the association betweenbreastfeeding and blood pressure in later life.The “adjusted” estimate will reflect the residualeffect of breastfeeding which is not mediatedby current weight (23).Main sources of heterogeneityamong studiesCurrent epidemiological practice places limitedvalue on the findings of a single study. <strong>Evidence</strong> isbuilt by pooling the results from several studies,if possible from different populations, eitherthrough systematic reviews or meta-analyses.A major concern regarding systematic reviewsand meta-analyses is the extent to whichthe results of different studies can be pooled.Heterogeneity of studies is unavoidable, and mayeven be positive as it enhances generalizability.The question is not whether heterogeneity ispresent, but if it seriously undermines the conclusionsbeing drawn. Rigorous meta-analysesshould incorporate a detailed investigation ofpotential sources of heterogeneity (24). In thepresent meta-analyses, the following possiblesources of heterogeneity were considered for allreviews.Year of birthStudies on the long-term effect of breastfeedinghave included subjects born during several decadesin the last century. During this period, thediets of non-breastfed infants in now high-incomecountries have changed markedly. In thefirst decades of the 20 th century, most nonbreastfedinfants received formulations based onwhole cow’s milk or top milk (25), with a highsodium concentration and levels of cholesteroland fatty acids that are similar to those inmature breastmilk. By the 1950s, commerciallyprepared formulas became increasingly popular.At this time, formulas tended to have a highsodium concentration and low levels of iron andessential fatty acids. Only after 1980, the sodiumcontent was reduced and nowadays themajority of formulas have levels that are similarto those in breastmilk (26). Therefore, the periodof the study cohorts’ births may affect thelong-term effects of breastfeeding, being a sourceof heterogeneity among the studies.Length of recall of breastfeedingMisclassification of breastfeeding duration hasbeen discussed above. Feeding histories wereoften assessed retrospectively and the length ofrecall has varied widely among studies. Accordingto the above-mentioned study by Huttly etal (19) in southern Brazil, as many as 24% ofthe mothers misclassified the duration ofbreastfeeding, and misclassification increasedwith the time elapsed since weaning. Other studieshave also reported poor maternal recall ofbreastfeeding duration (27-29). As in Brazil,Promislow et al (27) reported from the UnitedStates that mothers who breastfed for a shortperiod were more likely to exaggeratebreastfeeding duration, while the opposite wasobserved for women who breastfed for long periods.Length of recall is therefore a potentialsource of heterogeneity among studies.Source of information onbreastfeeding durationThe vast majority of the studies reviewed assessedinfant feeding by maternal recall, whileothers relied on information collected by healthworkers or on the subjects’ own reports. Marmotet al (30), in England, observed that about65% of subjects correctly recalled whether theyhad been breastfed or formula-fed, and bottlefedsubjects were more likely to report wronglythat they had been breastfed. If misclassificationwere independent of other factors related tomorbidity in adulthood, such as socioeconomicstatus, this misclassification would be non-differentialand would tend to underestimate thelong-term effects of breastfeeding.Categories of breastfeeding durationAmong the reviewed studies, most comparedever-breastfed subjects to those who were neverbreastfed. Other studies compared subjectsbreastfed for less than a given number of months,often 2-3 months (including those who werenever breastfed), to those breastfed for longerperiods. Few studies treated breastfeeding durationas a continuous or ordinal variable with6EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

several categories, thus allowing dose-responseanalyses. Furthermore, breastfeeding patterns(exclusive, predominant, or partial) have rarelybeen assessed.Studies comparing ever versus never breastfedsubjects may be subject to misclassification.The study by Huttly et al (19) showed thatmothers who had actually breastfed for up to 4weeks often reported, at a later time, that theynever breastfed.The comparison of ever versus neverbreastfed makes sense if the early weeks of lifeare regarded as a critical period for the programmingeffect of breastfeeding on adult diseases(31). On the other hand, if there is nocritical window and breastfeeding has a cumulativeeffect, comparisons of ever versus neverbreastfed infants will lead to substantial underestimationof the effect of breastfeeding.Study settingNearly all studies on the long-term consequencesof breastfeeding have been conducted in highincomecountries and in predominantly Caucasianpopulations. The findings from these studiesmay not hold for other populations exposedto different environmental and nutritional conditions,such as ethnic minorities in high-incomecountries (32) or populations from lessdeveloped countries.Among studies carried out in the last fewdecades, the type of milk fed to non-breastfedinfants would have varied substantially betweenhigh-income countries (where most babies receiveindustrialized formulas) and those fromlow- and middle-income countries (where wholeor diluted animal milk is often used).In this sense, lack of breastfeeding is an unusualvariable in epidemiological studies. For exposuresto, for example, smoking, alcohol or environmentalrisks, the reference category ismade up of those who are unexposed. In thecase of breastfeeding, however, those “unexposed”to it are themselves exposed to a numberof other foodstuffs, including animal milk, industrializedor home-made formulas, or traditionalweaning foods. Because alternative foodsvary markedly from one setting to another, theeffects of breastfeeding may be particularly affectedby where the study was carried out. Thelocation (area, country) of the study is thereforea potential modifier of the effect ofbreastfeeding.Adjustment for potential mediatingfactorsThis issue has been discussed above. Inappropriateadjustment was investigated as a potentialsource of heterogeneity among studies.GENERAL METHODOLOGICAL ISSUES IN STUDIES OF THE LONG-TERM EFFECTS OF BREASTFEEDING7

III. Search methodsSelection criteria for studiesIn the present meta-analyses, we selected observationaland randomized studies, published inEnglish, French, Spanish or Portuguese, examiningthe long-term effects of breastfeeding on thefollowing outcomes: blood pressure, overweightor obesity, cholesterol, type-2 diabetes, and intellectualperformance. Studies restricted tooutcome measurement in infants were excludedfrom the meta-analyses.Only those studies with internal comparisongroups were included. The type of comparisongroup used (never breastfed, breastfed forless than x months, etc.) did not constitute aneligibility criterion, but, as discussed above, theway in which breastfeeding was categorized wasinvestigated as a potential source of heterogeneityamong the studies.Types of outcome measuresAccording to the objectives of the present review,we looked for studies with the followingoutcomes:blood pressure: mean difference (in mmHg)in systolic and diastolic blood pressure;cholesterol: mean difference (in mg/dl) intotal cholesterol;overweight and obesity: odds ratio comparingbreastfed and non-breastfed subjects;type-2 diabetes: odds ratio comparingbreastfed and non-breastfed subjects (oralternatively, mean difference in bloodglucose levels);intellectual performance: mean attainedschooling and performance scores in developmentaltests.Search strategyIn order to prevent selection bias (33) by capturingas many relevant studies as possible, twoindependent literature searches were conducted:one at the Department of Child and AdolescentHealth and Development in the World HealthOrganization (R.B.) and another at the FederalUniversity of Pelotas in Brazil (B.L.H.).Medline (1966 to March 2006) was searchedusing the following terms for breastfeeding duration:breastfeeding; breast feeding; breastfed;breastfeed; bottle feeding; bottle fed; bottle feed;infant feeding; human milk; formula milk; formulafeed; formula fed; weaning.We combined the breastfeeding terms, withthe following terms for each of the studied outcomes:Cholesterol: cholesterol; LDL; HDL;triglycerides; or blood lipids.Type-2 diabetes: diabetes; glucose; or glycemia.Intellectual performance: schooling; development;or intelligence.Blood pressure: blood pressure; hypertension;systolic blood pressure; or diastolicblood pressure.Overweight or obesity: overweight; obesity;body mass index; growth; weight; height;child growth.In addition to the electronic search, the referencelists of the articles initially identified weresearched, and we also perused the ScientificCitation Index for papers citing the articles identified.Attempts were made to contact the authorsof all studies that did not provide sufficientdata to estimate the pooled mean effects.We also contacted the authors to clarify anyqueries on the study’s methodology.8EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

IV. Review methodsAssessment of study qualityEligible studies were evaluated for methodologicalquality prior to consideration of their results.The following a priori criteria for qualityassessment were used:a. Losses to follow-up (%)b. Type of study(0) Observational(1) Randomizedc. Birth cohort(0) No(1) Yesd. Length of recall of breastfeeding duration(0) >3 years(1)

papers. Therefore, such articles are more likelyto be identified and included in a meta-analysis.Publication bias is more likely to affect smallstudies, since the greater amount of time andmoney spent in larger studies makes them morelikely to be published, even when the results arenot statistically significant (34).In the present meta-analyses, funnel plot andEgger’s test were employed to assess whetherthere was any evidence of publication bias (37).We did not use the Begg test (38) because theregression method (Egger’s test) showed a betterperformance for detection of funnel plotasymmetry (39). Furthermore, we stratified theanalyses according to study size, in order to assessthe potential impact of publication bias onthe pooled estimate.Assessing heterogeneityThe next step after obtaining pooled results isto assess whether certain study characteristicsmay explain the variability between results. Inthe present meta-analyses we used a randomeffectsregression model developed by Berkey etal (40) for evaluation of sources of heterogeneity.In this approach, if the data are homogeneousor if heterogeneity is fully explained by thecovariates, the random-effects model is reducedto a fixed-effect. This analysis was performedusing the METAREG command within STATA.In these random-effects regression (meta-regression)models, each of the items used to assessstudy quality was considered as a covariate, insteadof using an overall quality score. This allowsthe identification of aspects of the studydesign, if any, which may be responsible for theheterogeneity between studies (41). Furthermore,the following study characteristics werealso included as covariates in random-effectsregression:a. Definition of breastfeedingb Birth yearc Age at outcome assessment:(0) 1–9 years(1) 10–19 years(2) >19 yearsd. Study size (n)e. Provenance (high-income country /middle/low-income country).As discussed in the Introduction, the presentreview was aimed at assessing the long-termconsequences of breastfeeding on five differentoutcomes. This resulted in five separate metaanalyseswhich are described below.10EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

V. Results and discussionReview 1 - Breastfeeding and blood pressure in later lifeHigh blood pressure in adulthood is associatedwith increased risk of ischaemic heart diseaseand stroke (42,43). It has been suggested thatadult blood pressure is influenced by early lifeexposures, such as intrauterine growth, catchupgrowth, and infant feeding (44).Biological plausibilityThree possible biological mechanisms for a possibleprogramming effect of breastfeeding onblood pressure have been proposed.Differences in sodium content betweenbreastmilk and formulaAs previously discussed (section on generalmethodological issues), until the 1980s the sodiumcontent of breastmilk in Western countrieswas much lower than that of formulas (26).Because low sodium intake is related to lowerblood pressure (45), it has been suggested thatdifferences in sodium content between breastmilkand formula would be one of the mechanismsfor the programming of later blood pressure.However, evidence on the existence of aneffect of early salt intake on later blood pressureis controversial. Whitten & Stewart (46)reported that blood pressure at eight years ofage was not correlated with sodium intake atthe age of eight months. In another study,Singhal et al (47) followed pre-term infants whohad been randomly assigned to receive two differenttypes of infant formulas which differedgreatly in salt content. Blood pressure at age13-16 years was independent of the type of formulathe subject had received, but was loweramong the breastfed subjects. On the otherhand, Geleijnse et al (48) reported that adjustedsystolic blood pressure at age of 15 years was3.6 mmHg (95% CI: -6.6 to -0.5) lower in childrenassigned to a low sodium diet in the first 6months of life. Therefore, there is no consensuson whether the sodium content of infant dietsmay lead to higher blood pressure in the future.Fatty acid content of breastmilkLong-chain polyunsaturated fatty acids arepresent in breastmilk, but not in most brandsof formula (49); these substances are importantstructural components of tissue membrane systems,including the vascular endothelium (50).<strong>Evidence</strong> suggests that dietary supplementationwith long-chain polyunsaturated fatty acids lowersthe blood pressure in hypertensive subjects(51). Furthermore, Forsyth et al (52) reportedthat blood pressure at 6 years was lower amongformula-fed children who had been assigned toa formula supplemented with long-chain polyunsaturatedfatty acids than among thoserandomized to a standard formula. This is, therefore,a potential mechanism for a possible effectof breastfeeding.ObesityThe effect of early infant feeding on blood pressuremight also be mediated by overweight orobesity in adulthood, as this is a risk factor forhypertension (53). On the other hand, as discussedbelow, the evidence suggests thatbreastfeeding has only a small protective effectagainst excess weight. Whether or not this smalleffect may influence blood pressure levels remainsto be proven.In conclusion, of the three postulated mechanisms,only the fatty acid content of breastmilkappears to be supported by the literature, butthere may well be other mechanisms that arecurrently unknown.11

Specific methodological issuesThe meta-analysis by Martin et al in 2005(55) included 15 studies that related breastfeedingto blood pressure measured after the age ofGeneral methodological issues affecting studiesof the long-term consequences of breastfeeding12 months. Meta-regression analysis was usedwere addressed above. An additional issue is thatto evaluate whether the mean effect ofseveral studies included adjustment for currentbreastfeeding varied according to the subject’sweight, body mass index or ponderal index inage when the blood pressure was measured, yeartheir multivariate analyses. Had breastfeedingof birth, study size, length of recall ofbeen associated with adult weight, adjustmentbreastfeeding duration, follow-up rates, andfor the latter would lead to an underestimationwhether or not confounding variables were adjustedfor.of its true effect on blood pressure. However, asdiscussed in section 1.3 above, there is no strongevidence of a breastfeeding effect on adultweight, and therefore one would not expect adjustmentBreastfeeding and systolic bloodpressurefor weight to change the associationFigure 1.1 shows that, in spite of the differencewith blood pressure.in the admission criteria, the effect size in bothmeta-analyses was similar, systolic blood pressurebeing significantly lower among breastfedOverview of the existing metaanalysesinfants.We reviewed two existing meta-analyses on theinfluence of breastfeeding on blood pressure inlater life (54,55).The meta-analysis by Owen et al in 2003 (54)obtained information from 25 studies, includingHowever, there was evidence of publicationbias; the effect size decreased with increasingstudy size. Owen et al (54) noted that studieswith fewer than 300 participants reported amean difference in systolic blood pressure ofFigure 1.1. Mean difference in systolic blood pressure betweenbreastfed and non breastfed subjects-2.05 mmHg (95% CI: -3.30 to-0.80) when comparing breastfedand non breastfed infants,whereas among studies with0-0.5-1-1.5-2more than 1000 subjects themean difference was -0.16mmHg (95% CI: -0.60 to 0.28).Martin et al (55) also reportedsimilar differences in mean effectaccording to study size.In terms of heterogeneityamong studies, the magnitude of-2.5the effect was independent of theOw en et al (54) Martin et al (55)subjects’ age. On the other hand,Martin et al (55) observed thatMeta-analysisthe protective effect ofbreastfeeding was higher amongthose born in or before 1980those in which blood pressure was meas-ured in infancy. Meta-regression analysis wasused to evaluate whether there were differencesin the mean effect of breastfeeding accordingto the subject’s age, year of birth, study size,length of recall for information on breastfeedingduration (for retrospective studies), and adjustmentfor current body size.(mean difference: -2.7 mmHg) compared withthose born after (mean difference: -0.8 mmHg),whereas Owen et al (54) failed to observe suchan association.Martin et al (55) noticed that only 6 of the15 studies included in their meta-analysis hadcontrolled for confounding by socioeconomicstatus and maternal variables (body mass index,Mean difference (95% CI) in systolicblood pressure (mm Hg)12EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

Unlike for systolic blood pressure, there wasno evidence of publication bias. The mean effectof breastfeeding on diastolic blood pressurewas similar among studies with 1000 ormore participants (mean difference: -0.4 mmHg;95% CI: -0.9 to 0.1) and smaller studies (

Studies not included in the previousmeta-analysesWe have identified four recently published studiesthat have not been included in the publishedmeta-analyses. Below are summarized the findingsfrom these studies.Martin et al (57) studied a cohort of 1580men living in Caerphilly, South Wales, who wereaged 45-59 years when examined between 1979and 1983. Information on breastfeeding durationwas obtained from the subjects’ mothersor a close female relative. Difference in systolicblood pressure between breastfed and bottle-fedsubjects was -0.11 mmHg (95% CI: -2.28 to 2.06)and for diastolic blood pressure -0.21 mmHg(95% CI: -1.67 to 1.25).Martin et al also studied (58) a historic cohortbased on a follow-up of the subjects whoparticipated in a 1-week survey of diet and healthwhen aged 0 to 19 years, between 1937 and1939. This study was conducted in 16 centresin England and Scotland. Information onbreastfeeding duration was obtained from thesubject’s mother at the time of the survey ondiet and health. Differences between breastfedand non-breastfed subjects were -1.62 mmHg(95% CI: -6.66 to 3.41) and -0.74 mmHg (95%CI: -3.06 to 1.57) for systolic and diastolic bloodpressure, respectively.Lawlor et al (60) evaluated random samplesof schoolchildren aged 9 years and similar samplesof 15-year-olds from Estonia (n=1174) andDenmark (n=1018). Even after controlling forpossible confounding variables, the systolic bloodpressure was lower among those children whohad ever been exclusively breastfed (difference:-1.7 mmHg; 95% CI: -3.0 to -0.5).In a recently published paper, Horta et al(61) observed in a cohort of over 1000 15-yearoldsin Pelotas (Brazil) that breastfeeding wasnot related to systolic (difference -1.31 mmHg;95% CI: -3.92 to 1.30) or diastolic blood pressure(difference -0.64 mmHg; 95% CI: -2.91 to1.63).Update of existing meta-analysisA new meta-analysis was carried out which includedthe four recently published studies describedabove, all the papers in previously publishedmeta-analyses, and those identified by thetwo independent literature searches at WHOand at the University of Pelotas. It was possibleto include 30 estimates on the effect ofbreastfeeding on systolic blood pressure, and 25on diastolic blood pressure (Table 1.2). Fig. 1.3and 1.4 show the forest plot for systolic anddiastolic blood pressures, respectively. Systolic(mean difference: -1.21 mmHg; 95% CI: -1.72to -0.70) and diastolic blood pressures (meandifference: -0.49 mmHg; 95% CI: -0.87 to -0.11)were both lower among those subjects who hadbeen breastfed. Random-effects models wereused in both analyses because heterogeneityamong studies was statistically significant.Similar to the previously published metaanalyses,publication bias was clearly present.Table 1.2 shows that the mean difference wasinversely related to the study size, with largerstudies reporting a smaller protective effect ofbreastfeed-ing. This was more marked forsystolic than for diastolic blood pressure. Thisis confirmed by examination of the funnel plotswhich are clearly asymmetrical, with small studiesreporting a higher protective effect ofbreastfeeding. (Fig. 1.5 and 1.6)ConclusionAccording to Owen et al (54), the associationbetween breastfeeding and lower blood pressurewas mainly due to publication bias, and any effectof breastfeeding was modest and of limitedclinical or public health relevance. In spite ofnot being able to exclude residual confoundingand publication bias, Martin et al (55) concludedthat breastfeeding was negatively associated withblood pressure. They argued that even a smallprotective effect of breastfeeding would be importantfrom a public health perspective. Forexample, a reduction in mean population bloodpressure of 2 mmHg could lower the prevalenceof hypertension by up to 17%, the number ofcoronary heart disease events by 6%, and strokeby 15%. Three large studies were published sincethe last review, two of which found no associationand one found a protective effect ofbreastfeeding.Both meta-analyses may have been affectedby publication bias. Small studies with negativeresults are less likely to be published, and this14EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

Table 1.2. Breastfeeding and blood pressure in later life: studies included in the meta-analysis in ascending order of subjects’ age at which outcome was measured.Author, year(reference)Study designYear of birthof subjectsAge at bloodpressuremeasurementGenderComparison groupsMean difference in mm Hg (Standarderror of mean difference)Systolic bloodpressureDiastolic bloodpressureBoulton 1981(62)Cross-sectional1976-91 yearAllBreastfed for ≥3 months (n=14) vs. Formula fedfrom 3 months (n=60) vs.Bottle fed (n=185)-2.17 (SE 1.61)0.48 (SE 1.42)Lawlor 2004 (59)Cohort1981-45 yearsAllBreastfed for ≥6 months (n=1191) vs. Breastfedfor 15 wk (n=74) vs.Bottle fed (n=105)-2.84 (SE 1.40)-1.79 (SE 1.23)Martin 2004 (56)Cohort1991-26-9 yearsAllEver breastfed (n=5586) vs. Never breastfed(n=1141)-0.8 (0.4)-0.6 (0.3)Lucas 1994 (69)Randomized controlledtrial1982-57-8 yearsAllAllocated to banked breastmilk (n=66) vs.Allocated to preterm formula (n=64)0.1 (SE 1.55)-0.7 (SE 1.32)Rona 1996 (70)Cross-sectional(England 1993 sample)1983-58-9 yearsAllExclusively breastfed for ≥3 mo (n=157) vs.Bottle fed (n=308)1.46 (SE 0.84)2.54 (SE 0.76)Rona 1996 (70)Cross-sectional(England 1994 sample)1984-68-9 yearsAllExclusively breastfed for ≥3 mo (n=213) vs.Bottle fed (n=318)-2.75 (SE 0.86)-2.34 (SE 0.75)Rona 1996 (70)Cross-sectional(Scotland 1994 sample)1983-68-9 yearsAllExclusively breastfed for ≥3 mo (n=124) vs.Bottle fed (n=273)-1.29 (SE 1.01)-0.97 (SE 0.84)Esposito-Del Puente1994 (71)Cross-sectional1980-29-11 yearsAllBreastfed (n=43) vs. Bottle fed (n=17)-4.13 (SE 2.86)0.11 (SE 0.86)RESULTS AND DISCUSSION - REVIEW 115

Table 1.2. (continued)Author, year (reference)Lawlor 2005 (60)British Cohort Study (72)Singhal 2001 (47)Owen 2002 (54)Taittonen 1996 (73)Horta 2006 (61)Williams 1992 (67)Kolacek 1993 (74)Leeson 2001 (75)Martin 2003 (76)Ravelli 2000 (77)Wadsworth 1987 (78)Fall 1995 (79)Martin 2005 (57, 58)Martin 2005 (58)Study designCross-sectionalCohortRandomizedcontrolled trialCross-sectionalCohortCohortCohortCohortCross-sectionalCohortCohortCohortCohortCohortCohortYear of birthof subjectsNot stated19701982-51982-61962-7419821972-31968-91969-751972-41943-719461920-301920-381918-39Age at bloodpressuremeasurement9-15 years10 years13-16 years13-16 years12-24 years15 years18 years18-23 years20-28 years23-27 years48-53 years53 years60-71 years45-59 years63-82 yearsGenderAllAllAllAllAllAllAllAllAllAllAllAllFemaleMaleAllComparison groupsExclusively breastfed vs. Not exclusivelybreastfedBreastfed for ≥3 months (n=951) vs. Neverbreastfed (n=5133)Allocated to banked breastmilk (n=66) vs.Allocated to preterm formula (n=64)Exclusively breastfed ≥3 months (n=980) vs.Exclusively bottle fed (n=951)Breastfed >3 months (n=760) vs. Bottle fed(n=150)Breastfed (n=966) vs. Bottle fed (n=85)Exclusively breastfed {median 28 wk} (n=197)or bottle fed (n=319)Exclusively breastfed ≥3 months (n=199) vs.Never breastfed (n=78)Breastfed (n=149) vs. Bottle fed (n=182)Lowest (n=193) vs. Highest quartile (n=79) ofdried formula milk consumption at age 3 monthsExclusively breastfed (n=520) vs. Bottle fed(n=105)Exclusively breastfed for ≥3 months (n=1507)vs. Never breastfed (n=628)Exclusively breastfed (n=862) vs. Bottle fed(n=68)Breastfed (n=1159) vs. Bottle fed (n=417)Breastfed (n=272) vs. Bottle fed (n=90)Mean difference in mm Hg (Standarderror of mean difference)Systolic bloodpressureDiastolic bloodpressure-1.7 (SE 0.66)Not stated-0.16 (SE 0.38)-0.21 (SE 0.33)-2.7 (SE 1.52)-3.2 (SE 1.31)0.32 (SE 0.64)-0.17 (SE 0.35)-5.48 (SE 1.14)Not stated-1.31 (SE 1.33)-0.64 (SE 1.16)-2.63 (SE 1.34)-2.66 (SE 1.36)0.00 (SE 1.96)0.75 (SE 1.40)0.0 (SE 1.55)-1.0 (SE 0.94)-3.7 (SE 1.38)-1.4 (SE 0.97)0.2 (SE 1.65)0.9 (SE 1.08)-0.94 (SE 0.97)-0.32 (SE 0.37)-2.64 (SE 2.88)-0.96 (SE 1.38)-0.11 (SE 1.11)-0.21 (SE 0.74)-1.62 (SE 2.57)-0.74 (SE 1.18)16EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

11111212111111121212121112121212121212121212121211123456712345671212121212121212121211111212111212Figure 1.3.Mean difference in systolic blood pressure in mm Hg (and its 95% confidence interval) betweenbreastfed and non-breastfed subjects in different studies. Whether the estimate was for males(M), females (F) and all (A) is indicated in parenthesisBoulton 1981 (M)Zeman 1981 (A)Baronowski 1992 (A)Whincup 1989 (A)Forsyth 2003 (A)Williams 1992 (A)Wilson 1998 (A)Martin 2004 (A)Lucas 1994 (A)Rona 1996 (A)Rona 1996 (A)Rona 1996 (A)Esposito-Del Puente 1994 (A)British Cohort Study (A)Singhal 2001 (A)Owen 2002 (A)Taittonen 1996 (A)Williams 1992 (A)Kolacek 1993 (A)Leeson 2001 (A)Martin 2003 (A)Ravelli 2000 (A)Wadsworth 1987 (A)Fall 1995 (F)Martin 2005 (A)Martin 2005 (M)Lawlor 2005 (A)Lawlor 2004 (A)Horta 2006 (A)Smith 1995 (A)1211211212121Combined-10 -5 -1 0 1 5Mean difference (mm Hg) Mean lower in breastfed Mean higher in breastfedsubjectssubjectsleads to an overestimate of the pooled meandifference due to selective inclusion of smallpositive studies.Lack of control for confounding is anothermethodological issue, as pointed out by Martinet al (55). Most studies did not provide estimatesadjusted for confounding by socioeconomicstatus and maternal characteristics; inthe studies that showed both adjusted and cruderesults, the latter tended to overestimate theprotective effect of breastfeeding. The majorityof these studies are from developed countries,and as previously discussed, the direction of confoundingmay vary according to the level of economicdevelopment of the population.In summary, the present updated meta-analysesshow that there are small but significant protectiveeffects of breastfeeding on systolic anddiastolic blood pressure. Publication bias is unlikelyto explain this finding because a significantprotective effect was observed even amongthe larger studies. However, residual confoundingcannot be excluded because of the markedreduction in effect size after adjustment forknown confounders.RESULTS AND DISCUSSION - REVIEW 117

Figure 1.4.Mean difference in systolic blood pressure in mm Hg (and its 95% confidence interval) betweenbreastfed and non-breastfed subjects in different studies. Whether the estimate was for males(M), females (F) and all (A) is indicated in parenthesisZeman 1981 (A)Baronowski 1992 (A)Whincup 1989 (A)Forsyth 2003 (A)Williams 1992 (A)Wilson 1998 (A)Martin 2004 (A)Lucas 1994 (A)Rona 1996 (A)Rona 1996 (A)Rona 1996 (A)Esposito-Del Puente 1994 (A)British Cohort Study (A)Singhal 2001 (A)Owen 2002 (A)Williams 1992 (A)Kolacek 1993 (A)Leeson 2001 (A)Martin 2003 (A)Ravelli 2000 (A)Wadsworth 1987 (A)Fall 1995 (F)Martin 2005 (A)Martin 2005 (M)Horta 2006 (A)Combined-5 -1 0 1 5Mean difference (mm Hg)Mean lower in breastfedsubjectsMean higher in breastfedsubjectsFigure 1.5.Funnel plot showing mean differencein systolic blood pressure (mm Hg) bystandard error of mean differenceFigure 1.6.Funnel plot showing mean differencein diastolic blood pressure (mm Hg) bystandard error of mean difference00Standard error120.5131.5-10 -5 0 5 -4 -2 0 2 4Mean difference (mm Hg)Mean difference (mm Hg)18EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

Table 1.3.Breastfeeding and blood pressure in later life: Random-effects meta-analyses by subgroupsSubgroup analysisSystolic blood pressureDiastolic blood pressureNumber ofestimatesMean difference (95%confidence interval)P valueNumber ofestimatesMean difference (95%confidence interval)P valueBy age group1 to 9 years9 to 19 years>19 years1488-1.06 (-1.70 to -0.42)-1.81 (-3.13 to -0.50)-0.95 (-1.97 to 0.07)0.040.0070.071168-0.54 (-1.27 to 0.19)-0.53 (-1.22 to 0.17)-0.38 (-0.90 to 0.15)0.140.140.16By study size

Review 2 - Breastfeeding and blood cholesterol in later lifeConcentrations of total cholesterol and LDLcholesterol (low-density lipoprotein) are importantrisk factors for coronary heart disease (80).It has been suggested that total cholesterol andLDL cholesterol may be programmed by earlylife exposures, such as rapid early growth (81)and infant feeding (82).Biological plausibilityThe cholesterol content is markedly higher inbreastmilk than in most commercially availableformulas. High cholesterol intake in infancy mayhave a long-term programming effect on synthesisof cholesterol by down-regulation of hepatichydroxymethylglutaryl coenzyme A (HMG-CoA) (83). This hypothesis is supported by studieswith animals, which showed that early exposureto increased levels of cholesterol is associatedwith decreased cholesterol levels at a laterage. Indeed, Devlin et al (84) reported thatHMG-CoA reductase was higher (P < .05) informula-fed than in milk-fed piglets, whereas LDLreceptor mRNA was not independent of earlydiet. HMG-CoA is the rate-limiting enzyme insynthesis of cholesterol from acetate, and HMG-CoA reductase inhibitors, the so-called statins,have an important cholesterol-lowering effect(85).Therefore, nutritional programming by thehigh cholesterol content of breastmilk has beenproposed as a potential mechanism for the associationbetween breastfeeding duration duringinfancy and lower cholesterol levels in laterages.Specific methodological issuesGeneral methodological issues affecting studiesof the long-term consequences of breastfeedingwere addressed in the Introduction. An additionalissue affecting studies of cholesterol levelsis that four studies included in theirmultivariate analyses an adjustment for weight,body mass index or ponderal index measured atthe same time as cholesterol levels. Hadbreastfeeding been associated with adult weight,adjustment for the latter would lead to an underestimationof its true effect on cholesterol.However, as discussed in the review on the longtermeffects of breastfeeding on blood pressure,the evidence suggests that the effect ofbreastfeeding on adult weight is weak, and thereforeone would not expect an adjustment for weightto change the association with cholesterol.Overview of the evidenceA previous meta-analysis (82) on this associationshowed that mean total cholesterol in infancywas higher among those breastfed (meandifference: 0.64; 95% CI: 0.50 to 0.79 mmol/L),whereas among adults the total cholesterol waslower among those who had been breastfed(mean difference -0.18; 95% CI: -0.30 to -0.06mmol/L).The electronic search carried out at WHOyielded 37 potentially relevant publications, 23of which provided data on the mean differencein total cholesterol between breastfed and nonbreastfedsubjects. From these 23 studies, 28estimates of total cholesterol were derived, ofwhich 18 included both genders and two weregender specific. No additional studies were identifiedby the independent search at the Universityof Pelotas. Table 2.1 presents a descriptionof the studies included in the present meta-analysis,and Figure 2.1 shows the distribution of thestudies’ results. There was strong evidence ofheterogeneity between studies (P = 0.003), andthe mean difference using a random-effectsmodel was -0.03 (95% CI: -0.10 to 0.03), suggestingno overall association betweenbreastfeeding and cholesterol levels.However, Table 2.2 shows marked effectmodification by age group. In adults (>19 years),breastfed subjects had mean total cholesterollevels 0.18 mmol/L (95% CI: 0.06 to 0.30 mmol/L) lower than those who were bottle-fed, andthere was no heterogeneity between studies (P= 0.86). For children and adolescents, the associationwas not significant. Figure 2.2 shows theforest plot for studies in adults.Other subgroup analyses were carried out.Studies with a length of recall of breastfeedingduration >3 years resulted in lower mean total20EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

Table 2.1. Breastfeeding and blood cholesterol in later life: studies included in the meta-analysis in ascending order of subjects' age at which outcome was measuredAuthor, year(reference)Study designYear of birthof subjectsAge atcholesterolmeasurementGenderComparison groupsMean difference intotal cholesterolmmol/L (SE)Jooste 1991 (86)Cohort1981-61 yearAllExclusively breastfed (n=110) vs. Formula fed (n=201) for the first 3months-0.07 (SE 0.09)Mize 1995 (87) CohortNot stated 1 yearAll Ever breastfed (n=23) vs. Formula fed from birth (n=33)0.35 (SE 0.21)Freedman 1992 (88) Cross-sectional1972-90 1-4 years All Ever breastfed (n=2232) vs. Bottle fed (n=1796)0.03 (SE 0.02)Friedman 1975 (89) Cross-sectionalNot stated 1.5-2 years All Breastfed (n=31) vs. Bottle fed (n=55)0.02 (SE 0.19)Ward 1980 (90) Cross-sectional19742.5-3 years All Breastfed for ≥1 month (n=28) vs. Bottle fed (n=46)0.25 (SE 0.13)Routi 1997 (91)Cohort1990-23 yearsAllBreastmilk only source of milk at 7 months (n=295) vs. Formula onlysource of milk at 7 months (n=324)0.12 (SE 0.06)Huttunen 1983 (92)Cohort19755 yearsAllExclusively breastfed for 6 months (n=35) vs. Formula feedingstarted between 1 to 6 months (n=32)0.2 (SE 0.11)Elaraby 1985 (93) Cross-sectionalNot stated 5-10 years All Breastfed for >6 months (n=50) vs. Artificially fed (n=50)-0.57 (SE 0.21)Plancoulaine 2000 (94)Cross-sectional1981-75-11 yearsMaleFemaleBreastfed (n=111) vs. Formula fed (n=131)Breastfed (n=101) vs. Formula fed (n=118)-0.3 (SE 0.13)0 (SE 0.16)Crawford 1981 (95)Cohort1969-706 yearsAllExclusively breastfed (n=32) vs. Exclusively formula-fed (n=16) at 3months-0.04 (SE 0.20)Hodgson 1976 (96)Cohort1962-57-12 yearsMaleFemaleExclusively breastfed (n= 12) vs. Formula fed (n=10) for the first 3monthsExclusively breastfed (n= 18) vs. Formula fed (n=10) for the first 3months0.54 (SE 0.34)0.23 (SE 0.27)Fomon 1984 (97)Cohort1966-718 yearsMaleFemaleBreastfed (n=70) vs. Formula fed (148)Breastfed (n=76) vs. Formula fed (105)-0.13 (SE 0.09)0.03 (SE 0.12)Hromodova 1997 (98) CohortNot stated 12-13 years All Breastfed (n=44) vs. Formula fed (n=13)-0.09 (SE 0.27)Owen 2002 (82)Cross-sectional1982-613-16 yearsAllExclusively breastfed (n= 620) vs. Exclusively formula fed (n=542)for the first 3 months-0.02 (SE 0.04)Singhal 2004 (99)Randomized controlledtrial1982-513-16 yearsAllAllocated to banked breastmilk (n=66) vs. Allocated to pretermformula (n=64)-0.4 (SE 0.20)Friedman 1975 (89) Cross-sectionalNot stated 15-19 years All Breastfed (n=86) vs. bottle fed (n=94)0.03 (SE 0.25)RESULTS AND DISCUSSION - REVIEW 221

Mean difference intotal cholesterolmmol/L (SE)Comparison groups-0.2 (SE 0.32)-0.2 (SE 0.31)Exclusively breastfed (n=35) vs. Formula fed (n=17)Exclusively breastfed (n=52) vs. Formula fed (n=14)Any breastfeeding (n=149) vs. Exclusively bottle fed (n=182)-0.18 (SE 0.11)-0.1 (SE 0.25)-0.5 (SE 0.23)Exclusively breastfed for 5 months (n=57) vs. Exclusively formulafed (n=20)Exclusively breastfed for 5 months (n=68) vs. Exclusively formulafed (n=27)-0.16 (SE 0.11)Exclusively breastfed (n=520) vs. Partly or exclusively bottle fed(n=105) in the first 10 daysBreastfed (n=272) vs. Bottle fed (n=90)0.12 (SE 0.28)Exclusively breastfed (n=344) vs. Formula fed (n=25)-0.32 (SE 0.25)Exclusively breastfed (n=208) vs. Formula fed (n=11)0.12 (SE 0.43)cholesterol for subjects who were breastfed, althoughthe difference was not statistically significant.Adjustment for body size at the timeof cholesterol assessment was a source of heterogeneitybetween studies; the protective effectof breastfeeding was restricted to studiesthat adjusted for body size (mean difference:-0.20; 95% CI: -0.33 to -0.06). Both types ofheterogeneity, however, could be explained bythe age ranges of the study subjects becauselonger recall and adjustment for current size weremore frequent in studies of adults. Age at assessmentof serum cholesterol explained 59.7%of the overall heterogeneity; further adjustmentfor body size and length of recall did not providefurther explanation for heterogeneity in therandom-effects model.Concerning publication bias, the funnel plotis quite symmetrical, with small studies tendingto report either positive or negative effectsof breastfeeding, with no evidence of bias. Indeed,Egger’s test was not statistically significant(P=0.16). Furthermore, Table 2.2 showsthat the mean difference in total cholesterol wasindependent of study size.Table 2.1. (continued)GenderAge atcholesterolmeasurementYear of birthof subjectsStudy designAuthor, year(reference)MaleFemale18-23 years1968-9CohortKolacek 1993 (100)Leeson 2001 (75) Cross-sectional1969-75 20-28 years AllMale31-32 years1946CohortMarmot 1980 (30)FemaleAll48-53 years1943-7CohortRavelli 2000 (77)Martin 2005 (58) Cohort1918-39 63-82 years MaleFall 1992 (101) Cohort1920-30 59-70 years MaleFall 1995 (79)Cohort1920-30 60-71 years FemaleConclusionThis meta-analysis suggests that the associationbetween breastfeeding and total cholesterol variesaccording to age. Whereas no significanteffect was observed in children or adolescents,mean cholesterol levels among adults who werebreastfed were 0.18 mmol/L (6.9 mg/dl) lowerthan among non-breastfed subjects. This associationdid not seem to be due to either publicationbias or confounding. The median level ofcholesterol in the adult studies included in thereview was about 5.7 mmol/L; the observed reductionassociated with breastfeeding correspondsto about 3.2% of this median.22EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

123123123123123123123123123451234512345123451234512312345612345612345612345612312345612312345678123456781212Figure 2.1.Mean difference in total cholesterol in mmol/L (and its 95% confidence interval) betweenbreastfed and non-breastfed subjects in different studies. Whether the estimate was for males(M), females (F) and all (A) is indicated in parenthesisJooste 1991 (A)Mize 1995 (A)Freedman 1992 (A)Friedman 1975 (A)Ward 1980 (A)Routi 1997 (A)Huttunen 1983 (A)Plancoulaine 2000 (M)Plancoulaine 2000 (F)Crawford 1981 (A)Hodgson 1976 (M)Hodgson 1976 (F)Hromodova 1997 (A)Owen 2002 (A)Friedman 1975 (A)Kolacek 1993 (M)Kolacek 1993 (F)Leeson 2001 (A)Marmot 1980 (M)Marmot 1980 (F)Ravelli 2000 (A)Martin 2005 (A)Fall 1992 (M)Fall 1995 (F)Singhal 2004 (A)Elaraby 1985 (A)Fomon 1984 (M)Fomon 1984 (F)Combined-5 -1 0 1 5Mean difference (mmol/L) Mean lower in breastfed Mean higher in breastfedsubjectssubjectsFigure 2.2.Mean difference in total cholesterol in mmol/L (and its 95% confidence interval) betweenbreastfed and non-breastfed subjects during adult life. Whether the estimate was for males(M), females (F) and all (A) is indicated in parenthesisKolacek 1993 (M)Kolacek 1993 (F)Leeson 2001 (A)Marmot 1980 (M)Marmot 1980 (F)Ravelli 2000 (A)Martin 2005 (A)Fall 1992 (M)Fall 1995 (F)Combined123123123123123-1.5 -1 -0.5 0 0.5 1 1.5Mean difference (mmol/L)Mean lower in breastfedsubjectsMean higher in breastfedsubjectsRESULTS AND DISCUSSION - REVIEW 223

Figure 2.3.Funnel plot showing mean difference in total cholesterol(mmol/L) by standard error of mean difference01Standard error234-1 -0.5 0 0.5 1Mean difference (mmol/L)Table 2.2.Breastfeeding and blood cholesterol in later life: Random-effects meta-analyses of cholesterollevels by subgroupSubgroup analysisNumber of estimatesof total cholesterolMean difference (95%confidence interval)P valueBy age group1 to 9 years9 to 19 years>19 years15490.02 (-0.06 to 0.11)-0.07 (-0.21 to 0.08)-0.18 (-0.30 to -0.06)0.630.370.004By study size

Review 3 - Breastfeeding and the risk of overweight and obesityin later lifeIt has been proposed that breastfeeding promotionmight be an effective way to preventthe development of obesity (102).Biological plausibilitySeveral possible biological mechanisms for aprotective effect of breastfeeding against overweightand obesity have been proposed.Differences in protein intake and energy metabolismmay be one of the biological mechanismslinking breastfeeding to later obesity.Lower protein intake and reduced energy metabolismwere reported among breastfed infants(103). Rolland-Cachera et al (104) observed thathigher protein intakes in early life, regardless ofthe type of feeding, was associated with an increasedrisk of later obesity.Another possibility is that breastfed and formula-fedinfants have different hormonal responsesto feeding, with formula feeding leadingto a greater insulin response resulting in fatdeposition and increased number of adipocytes(105).Finally, limited evidence suggests thatbreastfed infants adapt more readily to new foodssuch as vegetables, thus reducing the caloricdensity of their subsequent diets (106).Specific methodological issuesGeneral methodological issues affecting studiesof the long-term consequences of breastfeedingwere addressed in the Introduction. In addition,the following methodological issues should betaken into account when studying overweight/obesity as the outcome.Definition of overweight/obesityAlthough different criteria and percentiles havebeen used in the definition of obesity, the resultsof the studies have been similar. Arenz etal (107) reported no difference in mean effectamong studies using the 90 th , 95 th or 97 th percentileto define obesity. Therefore, differencesin the definition of overweight/obesity shouldnot be considered as a major methodologicalflaw in this meta-analysis. Existing cut-offs foroverweight/obesity will have to be reassessed inthe light of the new WHO Growth Standards.Mean body mass index or prevalenceof overweight/obesityGrummer-Strawn suggested that breastfeedingmay be associated both with a lower prevalenceof overweight/obesity and with underweight inlater life, due to a smaller variance of weightrelatedindices in subjects who were breastfed(32). Therefore, the mean body mass indexwould remain unchanged but breastfeedingwould still have an effect on the upper tail ofthe body mass index distribution - that is, onthe prevalence of overweight/obesity.Overview of existing metaanalysesThe protective effect of breastfeeding againstchildhood obesity was initially proposed byKramer in 1981 (108). More recently, severalstudies were published on this topic. We identifiedfour systematic reviews on the relationshipbetween breastfeeding and overweight.In 2004, Arenz et al (107) were the first topublish a systematic review of the evidence concerningthe protective effect of breastfeedingduration against childhood obesity. To be includedin their meta-analysis, the studies hadto fulfill the following criteria:Analyses had to be adjusted for at leastthree of the following possible confounders:birth weight, parental overweight,parental smoking, dietary factors, physicalactivity, and socioeconomic status.Odds ratios or relative risks had to bereported.Age at the last follow-up had to be between5 and 18 years.Obesity had to be defined by body massindex percentiles >90, 95 or 97.25

Figure 3.1.Odds of obesity10.90.80.70.60.50.40.30.2Only nine studies were included in this metaanalysis;19 were not eligible. The main reasonsfor exclusion were failure to report adequatelyadjusted estimates and a definition of obesitythat did not match the study criteria. Otherauthors (see below) were more flexible in acceptingdifferent definitions of overweight, andwere able to include a larger number of studiesin their meta-analyses.Analyses were stratified according to the followingstudy characteristics: type of design, agegroup, definition of breastfeeding, number ofvariables included in the multivariable analysis,and definition of obesity.Odds ratio of obesity according to study size,modified from Owen et al.< 500 500 - 2500 > 2500Number of participantsTheir pooled odds ratio was 0.78 (95% CI:0.71–0.85), and there was no sign of heterogeneityamong the nine studies. The protectiveeffect of breastfeeding was independent of thefollowing study characteristics: study design (cohortor cross-sectional); age at obesity assessment(6 years); definition ofbreastfeeding (never vs. ever or other definition);and definition of obesity (95 th or 97 th percentile).On the other hand, the protective effectof breastfeeding was slightly more pronouncedin studies that adjusted their estimates for lessthan 7 variables (pooled odds ratio: 0.69; 95%CI: 0.59–0.81), compared to those that adjustedfor 7 or more variables (pooled odds ratio: 0.78;95% CI: 0.70–0.87).The funnel plot was clearly asymmetric, withsmall studies tending to report a higher protectiveeffect of breastfeeding duration. This is astrong suggestion of publication bias.Owen et al published two meta-analyses. Inthe first (109), the authors managed to obtainodds ratio estimates from 28 of 61 studies reportingon the relationship betweenbreastfeeding and obesity. Unlike the Arenzmeta-analysis (107), Owen et al (109) includedstudies that provided only crude odds ratios andwere flexible in terms of the definition of obesity.Meta-regression analysis was used to investigatedifferences in the pooled odds ratio accordingto study size, age group at outcomemeasurement, year of birth and attrition rate,definition of obesity, and length of recallfor information on breastfeeding duration.Figure 3.1 shows that small studiestended to report stronger protective effectsof breastfeeding (pooled odds ratio: 0.43;95% CI: 0.33–0.55) than larger studies(pooled odds ratio: 0.88; 95% CI: 0.86–0.90).This protective effect may be due toconfounding by socioeconomic status andparental body composition. In developedcountries, women who breastfeed tend tohave higher socioeconomic status and maytherefore be more “nutrition-conscious”(21). Owen et al (109) identified 6 studieswhose estimates were adjusted for the followingconfounders: socioeconomic status,parental BMI, and maternal smoking. Inthese studies, adjustment for confounding reducedthe pooled odds ratio from 0.86 (95%CI: 0.81–0.91) to 0.93 (95% CI: 0.88–0.99). Thissuggests that at least part of the effect ofbreastfeeding on body composition is due toconfounding by socioeconomic status and parentalbody composition, and that crude estimatesshould not be included in the meta-analysis.Even after adjustment for socioeconomicstatus, the possibility of residual confoundingcannot be ruled out.Other methodological characteristics of thestudies, such as obesity definition and maternalrecall of breastfeeding duration, were not relatedto differences in the studies’ results.As in the meta-analysis by Arenz et al (107),publication bias was evident with small studiesreporting a stronger protective effect of26EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

eastfeeding (Fig. 3.1). Selective reporting wasalso evident; studies that failed to report oddsratios were much less likely to conclude thatbreastfeeding was associated with a reduced riskof obesity (1 of 35 studies), compared with studiesthat did provide odds ratios (18 of 29 studies);this difference was statistically significant(P 9each month of increase in breastfeedingDuration of breast feeding (months)duration was associated with a 4% decreasein the odds of overweight (OR:0.96; 95% CI: 0.94–0.98).Update of existing meta-analysesBecause previously published meta-analysesA new meta-analysis was carried out including(107,109,110) treated overweight/obesity asthe recently published studies described above,dichotomous variables, Owen et al (111) conducteda fourth meta-analysis to assess the ef-all the papers included in previously publishedmeta-analyses, and those identified by the twofect of breastfeeding on mean BMI. The authorsindependent literature searches at WHO andof the 70 identified studies were asked to provideinformation on mean differences in BMI,at the University of Pelotas. It was possible toinclude 33 studies with 39 estimates on the effectof breastfeeding on prevalence of over-according to breastfeeding duration; these estimateswere adjusted for age, socioeconomic status,maternal BMI, and maternal smoking inweight/obesity (Table 3.1). The forest plot showsthat results were clearly heterogeneous (Fig. 3.3).pregnancy.In a random-effects model, including all studies,breastfed individuals were less likely to beIn the fixed-effects model including 36 studies,breastfed subjects had lower mean BMIconsidered as overweight/obese, and the pooled(mean difference: -0.04; 95% CI: -0.05, -0.02).ratio was 0.78 (95% CI: 0.72–0.84).In spite of efforts to prevent publication bias,Odds ratio of overweightRESULTS AND DISCUSSION - REVIEW 327

Table 3.1. Breastfeeding and overweight/obesity in later life: studies included in the meta-analysis in ascending order of subjects' age at which outcome was measuredAuthor, year (reference)Study designYear of birthof subjectsAge measuredGenderComparison groupsOutcomeOdds ratio (95%confidence interval)Strbak 1991 (114)CohortNot stated1-7 yearsAllBreastfed for >1 months (n=741) vs. Breastfed 6 months vs. Never breastfed (n=677)*Overweight orobesity0.86 (0.44–1.65)Armstrong 2002 (118)Cohort1995-63 yearsAllExclusively breastfed (n=8751) vs. Bottle fed at 6-8weeks (n=23449)Obesity0.72 (0.65–0.79)Hediger 2001 (119)Cross-sectional1983-913-5 yearsAllEver breastfed (n=1158) vs. Never breastfed (n=1498)Overweight onlyObesity0.63 (0.41–0.96)0.84 (0.62–1.13)Grummer-Strawn 2004(32)Cohort1988-924 yearsAllBreastfed for ≥12 months (n=293) vs. Never breastfed(n=7084)Overweight orobesity0.72 (0.65–0.80)Dubois 2006 (120)Araujo 2006 (121)CohortCohort199819934 years4 yearsAllAllBreastfed for ≥3 months (n=710) vs. Breastfed for 9 months (n=194) vs. Breastfed

Table 3.1. (continued)Author, year (reference)Study designWadsworth 1999 (129)CohortThorsdottir 2003 (130)CohortBergmann 2003 (131)CohortReilly 2005 (132)CohortEid 1970 (133)CohortLiese 2001 (134)Cross-sectionalSung 2003 (A) (135)Cross-sectionalToschke 2002 (136)Cross-sectionalPoulton 2001 (117)CohortLi 2003 (122)Cross-sectionalGillman 2001 (137)Cross-sectionalLi 2005 (115)CohortElliott 1997 (138)CohortKramer 1981 (108)Case-controlKramer 1981 (108)Case-controlTulldahl 1999 (139)Cohort* total sampleYear of birthof subjects1946Not stated19901991-219611985-719901977-851972-31973-82Not stated1982-51977-80Not statedNot stated1979Age measured6 years6 years6 years7-8 years8 years9-10 years9-12 years6-14 years11 years9-18 years9-14 years1-14 years12-17 years12-18 years12-18 years15-16 yearsGenderAllMaleAllAllAllAllAllAllAllAllAllAllAllAllAllAllComparison groupsEver breastfed (n=2873) vs. Never breastfed (858)Breastfed for ≥6 months vs. Breastfed for 9 months (n=67) vs. Breastfed 2 months (n=73) vs. Breastfed for ≤2months (n=63)Ever breastfed (n=45) vs. Never breastfed (284)Ever breastfed (n=60) vs. Never breastfed (332)Breastfed for >2 months (n=390) vs. Breastfed for ≤2months (n=391)OutcomeOverweight onlyObesityOverweight orobesityOverweight onlyObesityObesityObesityOverweight orobesityOverweight orobesityOverweight orobesityObesityOverweightObesityOverweight onlyObesityObesityOverweightObesityObesityOverweight orobesityOdds ratio (95%confidence interval)0.94 (0.73–1.20)0.88 (0.59–1.32)0.33 (0.13–0.83)0.53 (0.31–0.89)0.46 (0.23–0.92)1.22 (0.87–1.71)0.42 (0.11–1.61)0.66 (0.52–0.87)0.45 (0.27–0.78)0.80 (0.71–0.90)0.80 (0.66–0.96)0.36 (0.1–1.28)0.73 (0.23–2.27)0.95 (0.84–1.07)0.78 (0.66–0.91)0.6 (0.3–1.6)0.51 (0.22–1.16)0.44 (0.21–0.93)0.29 (0.11–0.73)0.66 (0.44–0.98)RESULTS AND DISCUSSION - REVIEW 329

Odds ratio (95%confidence interval)OutcomeComparison groups0.73 (0.50–1.07)ObesityBreastfed for ≥1 month (n=489) vs. Breastfed for 6 months vs. Never breastfed (n=687)* Overweight 0.79 (0.46–1.34)0.64 (0.33–1.26)Breastfed for ≥4 months (n=121) vs. Never breastfed(n=78)0.34 (0.12–1.01)Overweight orobesityObesity0.93 (0.74–1.17)0.84 (0.67–1.05)ObesityObesityBreastfed for >1 month vs. Never breastfed (n=4662) *Breastfed for >1 month vs. Never breastfed (n=4625) *Ever breastfed (n=1999) vs. Never breastfed (n=386) Obesity0.73 (0.51–1.07)1.10 (0.88–1.37)ObesityBreastfed for >8 months (n=986) vs. Breastfed for 25) and obesity (e.g. BMI >30). Six of theseeight studies reported a more marked protectiveeffect against obesity than against overweightonly or overweight plus obesity. This supportsa causal effect of breastfeeding.Relevant studies not included inthe meta-analysesTable 3.1. (continued)GenderAge measuredYear of birthof subjectsStudy designAuthor, year (reference)Males18 years1982CohortVictora 2003 (140)Poulton 2001 (117) Cohort1972-3 21 years AllAll31-35 years1966CohortKvaavik 2005 (141)MalesFemales33 years1958CohortParsons 2003 (142)Richter 1981 (143) Cohort1957-9 Adult life AllAll56-66 years1934-44CohortEriksson 2003 (144)* total sampleA frequent limitation of observational studiesis inadequate adjustment for confounding bysocioeconomic and maternal variables. In theabsence of randomized studies, within-familyanalyses allow controlling for confounding bysocioeconomic, maternal variables, as well as selfselectionbias. Gillman et al (112) analysed datafrom 5614 sibling sets from the Growing UpToday Study to assess the association ofbreastfeeding with adolescent obesity withinsibling sets. The overall odds of overweight inthat study were smaller among subjects breastfedfor at least 7 months, compared with thosebreastfed for 3 months or less (OR: 0.85; 95%CI: 0.71–1.00). When the analyses were limitedto the 172 families in which one sibling wasbreastfed for at least 7 months and another for3 months or less, the resulting odds ratio wassimilar to that of the overall analyses OR: 0.89;95% CI: 0.50–1.59), but the upper 95% confidencelimit was well above the unity (1). Thissimilarity between the results obtained in the30EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

Figure 3.3.Odds ratio and its 95% confidence interval of being considered as overweight/obese, comparingbreastfed vs. non-breastfed subjects in different studies. Whether the estimate was formales (M), females (F) and all (A) is indicated in parenthesisStrbak 1991 (A)He 2000 (A)Poulton 2001 (A)Armstrong 2002 (A)Hediger 2001 (A)Grummer-Strawn 2004 (A)Li 2003 (A)Li 2003 (A)Scaglioni 2000 (A)O'Callaghan 1997 (A)Von Kries 1999 (A)Frye 2003 (A)Wadsworth 1999 (A)Bergmann 2003 (A)Li 2005 (A)Li 2005 (A)Dubois 2006 (A)Araujo 2006 (A)Maffeis 1994 (A)Burdette 2006 (A)Reilly 2005 (A)Eid 1970 (A)Liese 2001 (A)Sung 2003 (A)Toschke 2002 (A)Poulton 2001 (A)Gillman 2001 (A)Elliott 1997 (A)Kramer 1981 (A)Kramer 1981 (A)Tulldahl 1999 (A)Victora 2003 (A)Poulton 2001 (A)Kvaavik 2005 (A)Parsons 2003 (M)Parsons 2003 (F)Richter 1981 (A)Eriksson 2003 (A)Thorsdottir 2003 (M)Combined0.1 0.5 1 2 5 10Favours breastfeedingFavours not breastfeedingOdds ratio of overweight/obesittotal sample and those from the siblings studysuggests that confounding by socioeconomicstatus was not an important issue in this study.However, because heterogeneity in breastfeedingdurations among siblings is much smaller thanfor unrelated individuals, the effective samplesize for the within-family analysis was quite smallleading to less precise estimates. Another withinfamilyanalysis from the United States (113)found no association between breastfeedingduration and prevalence of obesity.RESULTS AND DISCUSSION - REVIEW 331

Table 3.2.Breastfeeding and the risk of overweight and obesity in later life: Random-effects metaanalysesof risk of overweight/obesity by subgroupSubgroup analysisNumber of estimatesPooled odds ratio and95% confidence intervalP valueBy age group1 to 9 years9 to 19 years>19 years221160.79 (0.71 to 0.87)0.69 (0.60 to 0.80)0.88 (0.74 to 1.04)0.0010.0010.13By study size

ConclusionThe evidence suggests that breastfeedingmay have a small protective effecton the prevalence of obesity. In spiteof the evidence of publication bias, aprotective effect of breastfeeding wasstill observed among the larger studies(>1500 participants), suggesting thatthis association was not due to publicationbias. With respect to confounding,studies that controlled for socioeconomicstatus and parental anthropometryalso reported that breastfeedingwas associated with a lower prevalenceof obesity. This effect seems tobe more important against obesitythan against overweight.Because the great majority of thepublished studies were conducted inWestern Europe and North America, we are notable to assess whether this association is presentin low and middle-income settings.Figure 3.4.00.2Standard error0.40.60.8Funnel plot showing odds ratio for overweight/obesityby standard error of odds ratioOdds ratio (log scale)1 2 3RESULTS AND DISCUSSION - REVIEW 333

subjects’ examination. After controlling for possibleconfounding variables, there was no associationbetween breastfeeding and insulin resistance.Difference in HOMA index was -4%(95% CI: -13 to 5) between ever and neverbreastfed subjects. Also, there was no evidenceof a dose-response trend.Martin et al (57) also studied a cohort ofmen from Caerphilly, South Wales, aged 45-59years when examined between 1979 and 1983.Information on breastfeeding duration was obtainedfrom the subjects’ mothers or a close femalerelative. Difference in the HOMA indexbetween breastfed and bottle-fed subjects was0.00 (95% CI: -0.10 to 0.10).Plancoulaine et al (94) evaluated a sampleof children aged 5 to 11 years in two small townsin northern France. Information on breastfeedingduration was obtained from the mother atthe moment of the children’s examination. Fastingblood glucose levels were similar amongbreastfed and non-breastfed children.Singhal et al (152) assessed 32-33 splitproinsulin concentration – a marker of insulinresistance – among subjects aged 13-16 yearswho were born pre-term and randomized toreceive a nutrient-enriched or lower-nutrientdiet. The level of 32-33 split proinsulin was loweramong those subjects who were randomized toreceive banked breastmilk, compared to thosereceiving pre-term formula; the difference wasnot statistically significant (P = 0.07).Conclusion<strong>Evidence</strong> on a possible programming effect ofbreastfeeding on glucose metabolism is sparse.Studies assessing the risk of type-2 diabetes reporteda protective effect of breastfeeding, witha pooled odds ratio of 0.63 (95% CI: 0.45–0.89)in breastfed compared to non-breastfed subjects.On the other hand, two other studies failed toreport an association between HOMA index, ameasure of insulin resistance, and breastfeedingduration, and a study on fasting blood glucoselevels was also negative. At this stage, it is notpossible to draw firm conclusions about the longtermeffect of breastfeeding on the risk of type-2 diabetes and related outcomes. Further studiesare badly needed on this topic.Figure 4.1.Odds ratio and 95% confidence interval of having type-2 diabetes in different studies, comparingbreastfed vs. non-breastfed subjects. Whether the estimate was for males (M), females (F)and all (A) is indicated in parenthesis.Young 2002 (A)Petit 1997 (A)Ravelli 2000 (A)Martin 2005 (A)Rich-Edwards 2004 (A)Combined0.1 0.5 1 2Odds ratio of type-2 diabetesFavours breastfeedingFavours not breastfeedingRESULTS AND DISCUSSION - REVIEW 435

Review 5 - Breastfeeding and school achievement/intelligencelevelsBiological plausibilityLong-chain polyunsaturated fatty acids arepresent in breastmilk, but not in most brandsof formula (49). These fatty acids are preferentiallyincorporated into neural cell membranes;structural lipids constitute about 60% of thehuman brain. The major lipid components includedocosahexaenoic (DHA) and arachidonic(AA) acids (153), which are important for retinaland cortical brain development (154). Bjerveet al (155) reported that the results of the Bayleymental and psychomotor development indexescorrelated positively with serum DHA concentrations.AA and DHA accumulate in the brain andretina most rapidly during the last trimester ofpregnancy and the first months after birth (156).Their reserves are limited at birth, especially inpre-term infants, and decline rapidly when lackingin the diet (157). Bottle-fed infants havebeen shown to have lower long-chain polyunsaturatedfatty acids in the phospholipids of thecerebral cortex than infants who are fedbreastmilk (158). This is, therefore, a potentialmechanism for an effect of breastfeeding on intellectualdevelopment.In addition to the chemical properties ofbreastmilk, breastfeeding enhances the bondingbetween mother and child (159, 160), which maycontribute to the child’s intellectual development.Specific methodological issuesGeneral methodological issues were addressedin the Introduction. Two points deserve specialattention. Because cognition and performancein intelligence tests are positively related to thestimulation received by the child (161) and becausebreastfeeding mothers may be more proneto stimulating their children (162), studies assessingthe long-term consequences ofbreastfeeding on intellectual performance shouldattempt to control for the quantity and qualityof stimulation. In addition, in societies wherebreastfeeding is more common among uppersocial groups, the possibility of confounding byparental education level has to be addressed.Overview of the evidenceWe identified three meta-analyses that assessedthe relationship between breastfeeding durationand performance in intelligence tests.In 1999, Anderson et al (163) were the firstto publish a meta-analysis on this topic, employingthe following selection criteria:the study included a comparison betweensubjects who were mostly breastfed withthose who were mostly bottle-fed;the outcome was measured with a widelyapplied test of cognitive development orability, yielding a single score;subjects were examined between infancyand adolescence.The studies were considered as including controlfor confounding if the estimates were adjustedfor a minimum of five variables, from alist of 15 potential confounders.Eleven studies were included in their analyses.In a random-effects model, the adjustedmean difference in cognitive function was of3.16 (95% CI: 2.35 to 3.98) points in favour ofbreastfed subjects. The effect of breastfeedingwas not modified by age at measurement. Thebenefit of breastfeeding was higher among lowbirthweight infants (mean difference: 5.18points; 95% CI: 3.59 to 6.77) although a significanteffect was also observed among normalbirthweight subjects (mean difference: 2.66points; 95% CI: 2.15 to 3.17).In 2000, Drane et al (164) carried out a secondsystematic review. Articles had to fulfil thefollowing criteria: subjects had to be born between 1960and 1998;cognition had to be measured by usingstandard tests;36EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

eastfeeding had to be measured as acategorical variable (exclusively breastfed,partially breastfed, exclusively formulafed)or as a “continuous variable” (durationof exclusive breastfeeding or proportionof the diet as breastmilk);analyses had to be adjusted for socioeconomicstatus and birth weight, at least.Twenty-four studies including subjects bornbetween 1960 and 1998 were identified, butonly five (165-169) fulfilled the three methodologicalpre-requisites. Another study (170),which met two standards (confounding and cognitionassessment) and partially met the standardfor breastfeeding measurement, was alsoincluded. Four of these studies reported a positiveeffect of breastfeeding on cognitive development,particularly for low-birthweight subjects.In 2002, Jain et al (171) published a thirdreview. Inclusion was restricted to studies publishedin English, and each study was assessedfor the following methodological aspects: studydesign, sample size, target population, qualityof feeding data, control of susceptibility bias,blinding and outcome measures. Of the 40 studiesidentified, only nine met all criteria for qualityof the feeding data; the breastfed group wasspecified as those who mostly breastfed, informationon infant feeding was collected in infancyand by interviewing the mother or fromhealth records, and duration of breastfeedingwas of at least 1 month among the breastfedsubjects. Only two of these nine studies presentedestimates adjusted for socioeconomicstatus and stimulation at home, and their resultswere conflicting. Wigg et al (172) reportedthat, after controlling for socioeconomic statusand quality of the child’s environment, breastfedsubjects presented a small advantage in theWechsler Intelligence Scale for Children at11-13 years of age (mean difference: 0.8; 95%CI: -1.9 to 3.5 points). Johnson et al (173) reportedthat the adjusted difference betweenbreastfed and non-breastfed 3-year-olds in theStanford-Binet Composite IQ Scale was 5.0points (95% CI: 0.3 to 9.5), while in the PeabodyPicture Vocabulary Test the difference was 4.6points (95% CI: 0.7 to 8.5). Seven other studiescontrolled for both socioeconomic status andstimulation/interaction of the child but did notfulfil the remaining methodological criteria.Three of these studies reported that performancein the intelligence test was higher amongbreastfed subjects (165,174,175), whereas fourfailed to show an association (162,176-178).Given the conflicting results from high-qualitystudies, the authors stated that the evidence onthe effect of breastfeeding on cognition was notconvincing.A very important study that was includedin the meta-analyses carried out by Andersonet al (163) and by Drane et al (164) was arandomized trial by Lucas et al (167), in whichnewborn pre-term babies received breastmilkor formula, and a significant improvement inWISC-R was found in the former. Because ofthe randomized design, these differences are notlikely to be due to confounding.Update of existing meta-analysisJain and colleagues (171) restricted their interpretationto whether or not the study resultshad been significant. They did not discuss thedirection of the effect in non-significant studies,nor did they try to pool the results in ameta-analysis. We carried out a meta-analysisincluding these papers, as well as three morerecent articles identified in our own systematicreview. According to the criteria proposed byJain and colleagues, all estimates were adjustedfor stimulation at home and fulfilled the otherquality criteria. Studies restricted to very lowbirthweight infants were not included. The threeadditional studies are described below.Quinn et al (179) studied a cohort of 7357singleton children whose mothers had been enrolledin the Mater Hospital-University ofQueensland Study of Pregnancy. Information onbreastfeeding duration was obtained from themother when the child was six months old. Atfive years, 4049 children were assessed with thePeabody Picture Vocabulary Test Revised(PPVT-R). Breastfeeding duration was positivelyassociated with the PPVT-R score, and after controllingfor confounding by socioeconomic status,birthweight, and stimulation in the home,the mean score for children breastfed for sixRESULTS AND DISCUSSION - REVIEW 537

months or more was 8.2 (95% CI: 6.5 to 9.9)points higher for females and 5.8 (95% CI: 4.1to 7.5) points higher for males, when comparedto those never breastfed.Clark et al (180) followed a cohort of childrenborn between 1991 and 1996 in urban communitiesnear Santiago, Chile, who were enrolledin a study on prevention of iron deficiency inhealthy full-term infants. At six months, the infantswere randomly assigned to receive ironsupplementation. At five years, information from784 (62.6%) subjects was gathered. Childrenbreastfed for 8 months had lowerscores for language, motor and cognition teststhan those breastfed for 2-8 months, after adjustmentfor socioeconomic factors and homestimulation.Angelsen et al (181) reported from Norwaythat the odds of having a low total intelligencequotient at five years of age was higher amongthose children breastfed for 8 months (14%) ascompared to those breastfed for 4–7 months(16.4%),

lescence or young adulthood suggests thatbreastfeeding is also positively associated witheducational attainment (168,182,183).The issue remains of whether the associationis related to the properties of breastmilkitself, or whether breastfeeding enhances thebonding between mother and child, and thuscontributes to intellectual development. Althoughin observational studies it is not possibleto disentangle these two effects, the positiveresults from the randomized trial carriedout by Lucas et al (167) suggest that the nutritionalproperties of breastmilk alone seem tohave an effect.Figure 5.1.Mean difference in cognitive development scores and its 95% confidence interval betweenbreastfed and non-breastfed subjects in different studies. Whether the estimate was for males(M), females (F) and all (A) is indicated in parenthesisJohnson (A) 1996Morrow-Tlucak (A) 1988Rogers (A) 1978Jacobson (A) 1992Lucas (A) 1992Wigg (A) 1998Quinn (M) 2001Quinn (F) 2001Combined-2 -1 0 1 2 3 4 5 6 7 8 9 10 15Mean lower in breastfedsubjectsMean higher in breastfedsubjectsMean differenceRESULTS AND DISCUSSION - REVIEW 539

VI. ConclusionsThe available evidence suggests that breastfeedingmay have long-term benefits. Subjects whohad been breastfed were found to have a lowermean blood pressure and lower total cholesterol,and showed higher performance in intelligencetests. Furthermore, the prevalence of overweight/obesity and type-2 diabetes was lower amongbreastfed subjects. All effects were statisticallysignificant, but for some outcomes their magnitudewas relatively modest.The Table 6.1 summarizes the magnitude ofthe effects of breastfeeding based on the fivemeta-analyses described above. For all outcomes,except performance in intelligence tests, we providea comparison of these effects with thoseobserved for other public health interventions.For blood pressure, the effect of breastfeedingwas smaller than those derived from other publichealth interventions targeted at adults, suchas dietary advice, physical activity, salt restriction,and multiple risk factor interventions. Onthe other hand, for total cholesterol amongadults, the magnitude of the breastfeeding ef-fect was similar to that of dietary advice inadulthood. Similarly, for the prevention of type-2 diabetes, the magnitude was similar to thatof diet and physical activity. Concerning obesity,whereas Summerbell et al (184) reportedthat combined dietary education and physicalactivity interventions were not effective in reducingchildhood obesity and overweight, wenoticed that breastfeeding was associated witha 22% reduction in the prevalence of overweight/obesity.This Table is intended for illustrative purposesonly. It should be interpreted with cautionbecause it includes a comparison of theeffect of actual interventions – none of themwith perfect compliance levels – with the grossdifference of the effect between breastfed andnon-breastfed subjects, which corresponds to anintervention with 100% compliance. Only thelong-term follow-up of subjects involved inbreastfeeding trials will provide a more accurateestimate of the impact of breastfeedingpromotion.40EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

Table 6.1. Summary of effect of breastfeeding and comparison with other interventionsOutcomeRange of effect of other public health interventions in later life Breastfeeding(current review)Diet / Dietary adviceExerciseMultiple risk factorinterventionModest salt restrictionPooled effect size (95%confidence interval)ConclusionBlood pressure(mean difference in mmHg, 95% CI)SystolicDiastolic-2.1 (-2.8 to -1.4)-1.6 (-2.7 to -0.6)-3.84 (-4.97 to -2.72)-2.58 (-3.35 to -1.81)-4.2 (-4.6 to -3.8)-2.7 (-2.9 to -2.5)-2.03 (-2.56 to -1.50)-0.99 (-1.4 to -0.57)-1.21 (-1.72 to -0.70)-0.49 (-0.87 to -0.11)The effect of breastfeeding issignificant, but smaller than the effectof other interventions.ref. (45)ref. (185)ref. (186)ref. (187)Total serum cholesterol(mean difference inmmol/L, 95% CI)-0.13 (-0.23 to -0.03)ref. (45)-0.14 (-0.16 to -0.12)ref. (186)-0.18 (-0.30 to -0.06)The effect of breastfeeding issignificant and larger than the effectof other interventions.Overweight or obesityDiet/dietary advice and exercise5/6 studies showed no effect on childhoodobesity. Meta-analysis not done because ofheterogeneity between studies.Odds ratio0.78 (0.72 to 0.84)The effect of breastfeeding issignificant (22% reduction), whileother interventions showed no effect.ref. (184)Type II diabetesDiet/dietary advice and exercise31-46% reduction in risk in persons withimpaired glucose tolerance.ref. (188)Odds ratio0.63 (0.45 to 0.89)The effect of breastfeeding issignificant (37% reduction) and ofsimilar magnitude to the effect of otherinterventions.Intelligence test scoresMean difference4.9 points (2.97 to 6.92)The effect of breastfeeding issignificant, with a substantial effectsize.CONCLUSIONS41

References1. WHO Collaborative Study Team on theRole of Breastfeeding on the Preventionof Infant Mortality. Effect of breastfeedingon infant and child mortality due toinfectious diseases in less developedcountries: a pooled analysis. Lancet, 2000,355(9202):451-5.2. Kramer MS, Kakuma R. The optimal durationof exclusive breastfeeding: a systematicreview. Advances in ExperimentalMedicine and Biology, 2004, 554:63-77.3. World Health Organization. Global Strategyfor Infant and Young Child Feeding, TheOptimal Duration of Exclusive Breastfeeding.Geneva, World Health Organization,2001.4. World Health Organization and UNICEF.Protecting, Promoting and Supporting Breastfeeding:The Special Role of Maternity Services.Geneva, World Health Organization,1989.5. Barker DJP. Fetal and infant origins of adultdisease. London, BMJ Publishing, 1992.6. Rich-Edwards JW et al. Birth weight andrisk of cardiovascular disease in a cohortof women followed up since 1976. BritishMedical Journal, 1997, 315(7105):396-400.7. Newsome CA et al. Is birth weight relatedto later glucose and insulin metabolism?A systematic review. Diabetic Medicine,2003, 20(5):339-48.8. Horta BL et al. Early and late growth andblood pressure in adolescence. Journal ofEpidemiology and Community Health, 2003,57(3):226-30.9. Joseph KS, Kramer MS. Review of theevidence on fetal and early childhoodantecedents of adult chronic disease. EpidemiologicReviews, 1996, 18(2):158-74.10. Lucas A, Fewtrell MS, Cole TJ. Fetal originsof adult disease - the hypothesis revisited.British Medical Journal, 1999,319(7204):245-9.11. Lucas A. Programming by early nutritionin man. Ciba Foundation Symposium, 1991,156:38-55.12. Gluckman PD, Hanson MA. Living withthe past: evolution, development, andpatterns of disease. Science, 2004,305(5691):1733-6.13. Chalmers I. Unbiased, relevant, and reliableassessments in health care: importantprogress during the past century, butplenty of scope for doing better. BritishMedical Journal, 1998, 317(7167):1167-8.14. Altman DG et al. The revised CONSORTstatement for reporting randomized trials:explanation and elaboration. Annalsof Internal Medicine, 2001, 134(8):663-94.15. Lucas A et al. Multicentre trial on feedinglow birthweight infants: effects of dieton early growth. Archives of Disease in Childhood,1984, 59(8):722-30.16. Kramer MS et al. Promotion ofBreastfeeding Intervention Trial(PROBIT): a randomized trial in theRepublic of Belarus. The Journal of theAmerican Medical Association, 2001,285(4):413-20.17. Concato J, Shah N, Horwitz RI.Randomized, controlled trials, observationalstudies, and the hierarchy of researchdesigns. New England Journal ofMedicine, 2000, 342(25):1887-92.18. Victora CG et al. The Pelotas birth cohortstudy, Rio Grande do Sul, Brazil,1982-2001. Cadernos de Saude Publica,2003, 19(5):1241-56.42EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

19. Huttly SR et al. Do mothers overestimatebreast feeding duration? An exampleof recall bias from a study in southernBrazil. American Journal of Epidemiology,1990, 132(3):572-5.20. World Health Organization. Contemporarypatterns of breastfeeding. Report on the WHOCollaborative Study on Breastfeeding. Geneva,World Health Organization, 1981.21. Bauchner H, Leventhal JM, Shapiro ED.Studies of breastfeeding and infections.How good is the evidence? The Journal ofthe American Medical Association,1986,256(7):887-92.22. Horta BL et al. [Breastfeeding and feedingpatterns in two cohorts of childrenin southern Brazil: trends and differences].Cadernos de Saude Publica, 1996, 12(Supplement 1):43-48.23. Victora CG et al. The role of conceptualframeworks in epidemiological analysis:a hierarchical approach. International Journalof Epidemiology, 1997, 26(1):224-7.24. Thompson SG. Why sources of heterogeneityin meta-analysis should be investigated.British Medical Journal, 1994,309(6965):1351-5.25. Barr SI et al. Effects of increased consumptionof fluid milk on energy andnutrient intake, body weight, and cardiovascularrisk factors in healthy olderadults. Journal of the American Dietetic Association,2000, 100(7):810-7.26. Fomon S. Infant feeding in the 20th century:formula and beikost. The Journal ofNutrition, 2001, 131(2):409S-20S.27. Promislow JH, Gladen BC, Sandler DP.Maternal recall of breastfeeding durationby elderly women. American Journal of Epidemiology,2005, 161(3):289-96.28. Eaton-Evans J, Dugdale AE. Recall bymothers of the birth weights and feedingof their children. Human NutritionApplied Nutrition, 1986, 40(3):171-5.29. Kark JD et al. Validity of maternal reportingof breast feeding history and the associationwith blood lipids in 17 year olds inJerusalem. Journal of Epidemiology and CommunityHealth, 1984, 38(3):218-25.30. Marmot MG et al. Effect of breast-feedingon plasma cholesterol and weight inyoung adults. Journal of Epidemiology andCommunity Health, 1980, 34(3):164-7.31. Kuh D et al. Life course epidemiology.Journal of Epidemiology and CommunityHealth, 2003, 57(10):778-83.32. Grummer-Strawn LM, Mei Z. Doesbreastfeeding protect against pediatricoverweight? Analysis of longitudinal datafrom the Centers for Disease Control andPrevention Pediatric Nutrition SurveillanceSystem. Pediatrics, 2004, 113(2):e81-6.33. Egger M, Smith GD. Bias in location andselection of studies. British Medical Journal,1998, 316(7124):61-6.34. Greenland S. Quantitative methods in thereview of epidemiologic literature. EpidemiologicReviews, 1987, 9:1-30.35. DerSimonian R, Laird N. Meta-analysisin clinical trials. Controlled Clinical Trials,1986, 7(3):177-88.36. Normand SL. Meta-analysis: formulating,evaluating, combining, and reporting.Statistics in Medicine, 1999, 18(3):321-59.37. Egger M et al. Bias in meta-analysis detectedby a simple, graphical test. BritishMedical Journal, 1997, 315(7109):629-34.38. Begg CB, Mazumdar M. Operating characteristicsof a rank correlation test forpublication bias. Biometrics, 1994,50(4):1088-101.39. Sterne JA, Gavaghan D, Egger M. Publicationand related bias in meta-analysis:power of statistical tests and prevalencein the literature. Journal of Clinical Epidemiology,2000, 53(11):1119-29.40. Berkey CS et al. A random-effects regressionmodel for meta-analysis. Statistics inMedicine, 1995, 14(4):395-411.REFERENCES43

41. Greenland S. Invited commentary: a criticallook at some popular meta-analyticmethods. American Journal of Epidemiology,1994, 140(3):290-6.42. Lawes CM et al. Blood pressure and cardiovasculardisease in the Asia Pacificregion. Journal of Hypertension, 2003,21(4):707-16.43. Boulanger JM, Hill MD. Hypertension andstroke: 2005 Canadian HypertensionEducational Program recommendations.The Canadian Journal of Neurological Sciences,2005, 32(4):403-8.44. Forsen T et al. Growth in utero and duringchildhood among women who developcoronary heart disease: longitudinal study.British Medical Journal, 1999,319(7222):1403-7.45. Brunner EJ et al. Dietary advice for reducingcardiovascular risk. Cochrane Databaseof Systematic Reviews, 2005(4):CD002128.46. Whitten CF, Stewart RA. The effect ofdietary sodium in infancy on blood pressureand related factors. Studies of infantsfed salted and unsalted diets for fivemonths at eight months and eight yearsof age. Acta Paediatrica Scandinavica Supplement,1980, 279:1-17.47. Singhal A, Cole TJ, Lucas A. Early nutritionin preterm infants and later bloodpressure: two cohorts after randomisedtrials. Lancet, 2001, 357(9254):413-9.48. Geleijnse JM et al. Long-term effects ofneonatal sodium restriction on bloodpressure. Hypertension, 1997, 29(4):913-7.49. Koletzko B et al. Long chain polyunsaturatedfatty acids (LC-PUFA) and perinataldevelopment. Acta Paediatrica, 2001,90(4):460-4.50. Engler MM et al. The effects of a dietrich in docosahexaenoic acid on organand vascular fatty acid composition inspontaneously hypertensive rats.Prostaglandins, Leukotrienes, and Essential FattyAcids, 1999, 61(5):289-95.51. Morris MC, Sacks F, Rosner B. Does fishoil lower blood pressure? A meta-analysisof controlled trials. Circulation, 1993,88(2):523-33.52. Forsyth JS et al. Long chain polyunsaturatedfatty acid supplementation in infantformula and blood pressure in laterchildhood: follow up of a randomisedcontrolled trial. British Medical Journal,2003, 326(7396):953.53. Perry IJ, Whincup PH, Shaper AG. Environmentalfactors in the development ofessential hypertension. British MedicalBulletin, 1994, 50(2):246-59.54. Owen CG et al. Effect of breast feeding ininfancy on blood pressure in later life: systematicreview and meta-analysis. BritishMedical Journal, 2003, 327(7425):1189-95.55. Martin RM, Gunnell D, Smith GD.Breastfeeding in infancy and blood pressurein later life: systematic review andmeta-analysis. American Journal of Epidemiology,2005, 161(1):15-26.56. Martin RM et al. Does breastfeeding ininfancy lower blood pressure in childhood?The Avon Longitudinal Study ofParents and Children (ALSPAC). Circulation,2004, 109(10):1259-66.57. Martin RM et al. Breastfeeding and cardiovasculardisease risk factors, incidence,and mortality: the Caerphilly study. Journalof Epidemiology and Community Health,2005, 59(2):121-9.58. Martin RM et al. Breastfeeding andatherosclerosis: intima-media thicknessand plaques at 65-year follow-up of theBoyd Orr cohort. Arteriosclerosis, Thrombosis,and Vascular Biology, 2005,25(7):1482-8.59. Lawlor DA et al. Associations of parental,birth, and early life characteristicswith systolic blood pressure at 5 years ofage: findings from the Mater-Universitystudy of pregnancy and its outcomes.Circulation, 2004, 110(16):2417-23.44EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

60. Lawlor DA et al. Infant feeding and componentsof the metabolic syndrome: findingsfrom the European Youth HeartStudy. Archives of Disease in Childhood, 2005,90(6):582-8.61. Horta BL et al. Breastfeeding duration andblood pressure among Brazilian adolescents.Acta Paediatrica, 2006, 95(3):325-31.62. Boulton J. Nutrition in childhood and itsrelationships to early somatic growth,body fat, blood pressure, and physical fitness.Acta Paediatrica Scandinavica Supplement,1981, 284:1-85.63. Smith RE et al. Determinants of bloodpressure in Sowetan infants. South AfricanMedical Journal, 1995, 85(12 Pt 2):1339-42.64. Zeman J, Simkova M. [Blood pressurevalues in infants and young children inrelation to the duration of breast feeding].Ceskoslovenská pediatrie, 1981,36(10):593-4.65. Baranowski T et al. Height, infant-feedingpractices and cardiovascular functioningamong 3 or 4 year old children inthree ethnic groups. Journal of Clinical Epidemiology,1992, 45(5):513-8.66. Whincup PH, Cook DG, Shaper AG.Early influences on blood pressure: astudy of children aged 5-7 years. BritishMedical Journal, 1989, 299(6699):587-91.67. Williams S, St George IM, Silva PA. Intrauterinegrowth retardation and bloodpressure at age seven and eighteen. Journalof Clinical Epidemiology, 1992,45(11):1257-63.68. Wilson AC et al. Relation of infant dietto childhood health: seven year follow upof cohort of children in Dundee infantfeeding study. British Medical Journal, 1998,316(7124):21-5.69. Lucas A, Morley R. Does early nutritionin infants born before term programmelater blood pressure? British Medical Journal,1994, 309(6950):304-8.70. Rona RJ, Qureshi S, Chinn S. Factors relatedto total cholesterol and blood pressurein British 9 year olds. Journal of Epidemiologyand Community Health, 1996,50(5):512-18.71. Esposito-Del Puente A et al. Familial andenvironmental influences on body compositionand body fat distribution inchildhood in southern Italy. InternationalJournal of Obesity and Related Metabolic Disorders,1994, 18(9):596-601.72. Butler NR et al. Recent findings from the1970 child health and education study: preliminarycommunication. Journal of the RoyalSociety of Medicine, 1982, 75(10):781-4.73. Taittonen L et al. Prenatal and postnatalfactors in predicting later blood pressureamong children: cardiovascular risk inyoung Finns. Pediatric Research, 1996,40(4):627-32.74. Kolacek S, Kapetanovic T, Luzar V. Earlydeterminants of cardiovascular risk factorsin adults. B. Blood pressure. ActaPaediatrica, 1993, 82(4):377-82.75. Leeson CP et al. Duration of breast feedingand arterial distensibility in earlyadult life: population based study. BritishMedical Journal, 2001, 322(7287):643-7.76. Martin RM et al. Infant nutrition andblood pressure in early adulthood: theBarry Caerphilly Growth study. The AmericanJournal of Clinical Nutrition, 2003,77(6):1489-97.77. Ravelli AC et al. Infant feeding and adultglucose tolerance, lipid profile, blood pressure,and obesity. Archives of Disease in Childhood,2000, 82(3):248-52.78. Wadsworth ME. Follow-up of the first nationalbirth cohort: findings from the MedicalResearch Council National Survey ofHealth and Development. Paediatric and PerinatalEpidemiology, 1987, 1(1):95-117.79. Fall CH et al. Fetal and infant growth andcardiovascular risk factors in women.British Medical Journal, 1995, 310(6977):428-32.REFERENCES45

80. Law MR, Wald NJ, Thompson SG. By howmuch and how quickly does reduction inserum cholesterol concentration lowerrisk of ischaemic heart disease? BritishMedical Journal, 1994, 308(6925):367-72.81. Lauren L et al. Relationship betweenbirthweight and blood lipid concentrationsin later life: evidence from the existingliterature. International Journal ofEpidemiology, 2003, 32(5):862-76.82. Owen CG et al. Infant feeding and bloodcholesterol: a study in adolescents and asystematic review. Pediatrics, 2002,110(3):597-608.83. Wong WW et al. Effect of dietary cholesterolon cholesterol synthesis in breastfedand formula-fed infants. Journal of LipidResearch, 1993, 34(8):1403-11.84. Devlin AM et al. Early diet influenceshepatic hydroxymethyl glutaryl coenzymeA reductase and 7alpha-hydroxylasemRNA but not low-density lipoproteinreceptor mRNA during development.Metabolism, 1998, 47(1):20-6.85. LaRosa JC, He J, Vupputuri S. Effect ofstatins on risk of coronary disease: ameta-analysis of randomized controlledtrials. The Journal of the American MedicalAssociation, 1999, 282(24):2340-6.86. Jooste PL et al. Effect of breastfeedingon the plasma cholesterol and growth ofinfants. Journal of Pediatric Gastroenterologyand Nutrition, 1991, 13(2):139-42.87. Mize CE et al. Lipoprotein-cholesterolresponses in healthy infants fed defineddiets from ages 1 to 12 months: comparisonof diets predominant in oleic acidversus linoleic acid, with parallel observationsin infants fed a human milk-baseddiet. Journal of Lipid Research, 1995,36(6):1178-87.88. Freedman DS, Lee SL, Byers T, Kuester S,Sell KI. Serum cholesterol levels in amultiracial sample of 7,439 preschoolchildren from Arizona. Preventive Medicine,1992, 21(2):162-76.89. Friedman G, Goldberg SJ. Concurrent andsubsequent serum cholesterol of breastandformula-fed infants. The American Journalof Clinical Nutrition, 1975, 28(1):42-5.90. Ward SD et al. Determinants of plasmacholesterol in children – a family study.The American Journal of Clinical Nutrition,1980, 33(1):63-70.91. Routi T et al. Tracking of serum lipoprotein(a) concentration and its contributionto serum cholesterol values in childrenfrom 7 to 36 months of age in theSTRIP Baby Study. Special Turku CoronaryRisk Factor Intervention Project forBabies. Annals of Medicine, 1997,29(6):541-7.92. Huttunen JK et al. Fat composition ofthe infant diet does not influence subsequentserum lipid levels in man. Atherosclerosis,1983, 46(1):87-94.93. Elaraby II EDMK, Aref GH. Effect ofbreast vs. humanized milk on concurrentand subsequent serum cholesterol levels.Saudi Medical Journal, 1985, 6:480-3.94. Plancoulaine S et al. Infant feeding patternsare related to blood cholesterolconcentration in prepubertal childrenaged 5-11 y: the Fleurbaix-Laventie VilleSante study. European Journal of ClinicalNutrition, 2000, 54(2):114-9.95. Crawford PB et al. Serum cholesterol of6-year-olds in relation to environmentalfactors. Journal of the American Dietetic Association,1981, 78(1):41-6.96. Hodgson PA et al. Comparison of serumcholesterol in children fed high, moderate,or low cholesterol milk diets duringneonatal period. Metabolism, 1976,25(7):739-46.97. Fomon SJ et al. Indices of fatness andserum cholesterol at age eight years inrelation to feeding and growth duringearly infancy. Pediatric Research, 1984,18(12):1233-8.46EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

98. Hromadova M et al. Relationship betweenthe duration of the breastfeedingperiod and the lipoprotein profile of childrenat the age of 13 years. PhysiologicalResearch, 1997, 46(1):21-5.99. Singhal A et al. Breastmilk feeding andlipoprotein profile in adolescents bornpreterm: follow-up of a prospectiverandomised study. Lancet, 2004,363(9421):1571-8.100. Kolacek S et al. Early determinants ofcardiovascular risk factors in adults. A.Plasma lipids. Acta Paediatrica, 1993,82(8):699-704.101. Fall CH et al. Relation of infant feedingto adult serum cholesterol concentrationand death from ischaemic heart disease.British Medical Journal, 1992, 304(6830):801-5.102. Gillman MW. Breast-feeding and obesity. TheJournal of Pediatrics, 2002, 141(6):749-50.103. Whitehead RG. For how long is exclusivebreastfeeding adequate to satisfy the dietaryenergy needs of the average young baby?Pediatric Research, 1995, 37(2):239-43.104. Rolland-Cachera MF et al. Influence ofmacronutrients on adiposity development:a follow up study of nutrition andgrowth from 10 months to 8 years of age.International Journal of Obesity and RelatedMetabolic Disorders, 1995, 19(8):573-8.105. Lucas A et al. Breast vs. bottle: endocrineresponses are different with formula feeding.Lancet, 1980, 1(8181):1267-9.106. Birch LL, Fisher JO. Development of eatingbehaviors among children and adolescents.Pediatrics, 1998, 101(3 Part2):539-49.107. Arenz S et al. Breastfeeding and childhoodobesity – a systematic review. InternationalJournal of Obesity and Related MetabolicDisorders, 2004, 28(10):1247-56.108. Kramer MS. Do breastfeeding and delayedintroduction of solid foods protectagainst subsequent obesity? The Journalof Pediatrics, 1981, 98(6):883-7.109. Owen CG et al. Effect of infant feedingon the risk of obesity across the lifecourse: a quantitative review of publishedevidence. Pediatrics, 2005, 115(5):1367-77.110. Harder T et al. Duration of breastfeedingand risk of overweight: a meta-analysis.American Journal of Epidemiology, 2005,162(5):397-403.111. Owen CG et al. The effect of breastfeedingon mean body mass index throughoutlife: a quantitative review of publishedand unpublished observational evidence.The American Journal of Clinical Nutrition,2005, 82(6):1298-307.112. Gillman MW et al. Breastfeeding andoverweight in adolescence. Epidemiology,2006, 17(1):112-4.113. Nelson MC, Gordon-Larsen P, Adair LS.Are adolescents who were breastfed lesslikely to be overweight? Analyses of siblingpairs to reduce confounding. Epidemiology,2005, 16(2):247-53.114. Strbak V et al. Late effects of breastfeedingand early weaning: seven-year prospectivestudy in children. Endocrine Regulations,1991, 25(1-2):53-7.115. Li C, Kaur H et al. Additive interactionsof maternal prepregnancy BMI andbreastfeeding on childhood overweight.Obesity Research, 2005, 13(2):362-71.116. He Q et al. Risk factors of obesity inpreschool children in China: a populationbasedcase-control study. International Journalof Obesity and Related Metabolic Disorders,2000, 24(11):1528-36.117. Poulton R, Williams S. Breastfeeding andrisk of overweight. The Journal of the AmericanMedical Association, 2001,286(12):1449-50.118. Armstrong J, Reilly JJ. Breastfeeding andlowering the risk of childhood obesity.Lancet, 2002, 359(9322):2003-4.119. Hediger ML et al. Association betweeninfant breastfeeding and overweight inyoung children. The Journal of the AmericanMedical Association, 2001, 285(19):2453-60.REFERENCES47

120. Dubois L, Girard M. Early determinantsof overweight at 4.5 years in a population-basedlongitudinal study. InternationalJournal of Obesity (London), 2006,30(4):610-7.121. Araujo CL et al. Breastfeeding and overweightin childhood: evidence from thePelotas 1993 birth cohort study. InternationalJournal of Obesity (London), 2006,30(3):500-6.122. Li L, Parsons TJ, Power C. Breastfeedingand obesity in childhood: cross sectionalstudy. British Medical Journal, 2003,327(7420):904-5.123. Maffeis C et al. Parental and perinatalfactors associated with childhood obesityin north-east Italy. International Journalof Obesity and Related Metabolic Disorders,1994, 18(5):301-5.124. Scaglioni S et al. Early macronutrientintake and overweight at five years of age.International Journal of Obesity and RelatedMetabolic Disorders, 2000, 24(6):777-81.125. O’Callaghan MJ et al. Prediction of obesityin children at 5 years: a cohort study.Journal of Paediatrics and Child Health, 1997,33(4):311-6.126. Burdette HL et al. Breastfeeding, introductionof complementary foods, and adiposityat 5 y of age. The American Journal of ClinicalNutrition, 2006, 83(3):550-8.127. von Kries R et al. Breastfeeding and obesity:cross sectional study. British MedicalJournal, 1999, 319(7203):147-50.128. Frye C, Heinrich J. Trends and predictorsof overweight and obesity in East Germanchildren. International Journal of Obesityand Related Metabolic Disorders, 2003,27(8):963-9.129. Wadsworth M et al. Breastfeeding andobesity. Relation may be accounted forby social factors. British Medical Journal,1999, 319(7224):1576.130. Thorsdottir I, Gunnarsdottir I, PalssonGI. Association of birth weight andbreast-feeding with coronary heart diseaserisk factors at the age of 6 years.Nutrition, Metabolism, and CardiovascularDiseases, 2003, 13(5):267-72.131. Bergmann KE et al. Early determinants ofchildhood overweight and adiposity in abirth cohort study: role of breast-feeding.International Journal of Obesity and RelatedMetabolic Disorders, 2003, 27(2):162-72.132. Reilly JJ et al. Early life risk factors forobesity in childhood: cohort study. BritishMedical Journal, 2005, 330(7504):1357.133. Eid EE. Follow-up study of physicalgrowth of children who had excessiveweight gain in first six months of life.British Medical Journal, 1970, 2(701):74-6.134. Liese AD et al. Inverse association ofoverweight and breast feeding in 9 to 10-y-old children in Germany. InternationalJournal of Obesity and Related Metabolic Disorders,2001, 25(11):1644-50.135. Sung RY et al. High prevalence of insulinresistance and metabolic syndrome inoverweight/obese preadolescent HongKong Chinese children aged 9-12 years.Diabetes Care, 2003, 26(1):250-1.136. Toschke AM et al. Overweight and obesityin 6- to 14-year-old Czech children in 1991:protective effect of breast-feeding. The Journalof Pediatrics, 2002, 141(6):764-9.137. Gillman MW et al. Risk of overweightamong adolescents who were breastfedas infants. The Journal of the American MedicalAssociation, 2001, 285(19):2461-7.138. Elliott KG et al. Duration of breastfeedingassociated with obesity during adolescence.Obesity Research, 1997, 5(6):538-41.139. Tulldahl J et al. Mode of infant feedingand achieved growth in adolescence: earlyfeeding patterns in relation to growth andbody composition in adolescence. ObesityResearch, 1999, 7(5):431-7.48EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

140. Victora CG et al. Anthropometry andbody composition of 18 year old menaccording to duration of breast feeding:birth cohort study from Brazil. BritishMedical Journal, 2003, 327(7420):901.141. Kvaavik E, Tell GS, Klepp KI. Surveys ofNorwegian youth indicated that breastfeeding reduced subsequent risk of obesity.Journal of Clinical Epidemiology, 2005,58(8):849-55.142. Parsons TJ, Power C, Manor O. Infantfeeding and obesity through thelifecourse. Archives of Disease in Childhood,2003, 88(9):793-4.143. Richter J. [Influence of duration of breastfeedingon body-weight-development].Ärztliche Jugendkunde, 1981, 72(3):166-9.144. Eriksson J et al. Obesity from cradle tograve. International Journal of Obesity andRelated Metabolic Disorders, 2003,27(6):722-7.145. Baur LA et al. The fatty acid compositionof skeletal muscle membrane phospholipid:its relationship with the typeof feeding and plasma glucose levels inyoung children. Metabolism, 1998;47(1):106-12.146. Arslanian S. Type-2 diabetes in children:clinical aspects and risk factors. HormoneResearch, 2002, 57 (Supplement 1):19-28.147. Aynsley-Green A. The endocrinology offeeding in the newborn. Baillière’s ClinicalEndocrinology and Metabolism, 1989,3(3):837-68.148. Salmenpera L et al. Effects of feeding regimenon blood glucose levels and plasmaconcentrations of pancreatic hormonesand gut regulatory peptides at 9 monthsof age: comparison between infants fedwith milk formula and infants exclusivelybreast-fed from birth. Journal of PediatricGastroenterology and Nutrition, 1988,7(5):651-6.149. Pettitt DJ et al. Breastfeeding and incidenceof non-insulin-dependent diabetesmellitus in Pima Indians. Lancet, 1997,350(9072):166-8.150. Young TK et al. Type-2 diabetes mellitusin children: prenatal and early infancy riskfactors among native Canadians. Archivesof Pediatrics and Adolescent Medicine, 2002,156(7):651-5.151. Rich-Edwards JW et al. BreastfeedingDuring Infancy and the Risk of CardiovascularDisease in Adulthood. Epidemiology,2004, 15(5):550-556.152. Singhal A et al. Low nutrient intake andearly growth for later insulin resistancein adolescents born preterm. Lancet, 2003,361(9363):1089-97.153. Crawford MA. The role of essential fattyacids in neural development: implicationsfor perinatal nutrition. The American Journalof Clinical Nutrition, 1993, 57(5Supplement):703S-709S; discussion709S-710S.154. Birch EE et al. Dietary essential fatty acidsupply and visual acuity development.Investigative Ophthalmology and Visual Science,1992, 33(11):3242-53.155. Bjerve KS et al. Omega-3 fatty acids: essentialfatty acids with important biologicaleffects, and serum phospholipidfatty acids as markers of dietary omega3-fatty acid intake. The American Journalof Clinical Nutrition, 1993, 57(5Supplement):801S-805S; discussion805S-806.156. Clandinin MT et al. Intrauterine fatty acidaccretion rates in human brain: implicationsfor fatty acid requirements. EarlyHuman Development, 1980, 4(2):121-9.157. van Beek RH, Carnielli VP, Sauer PJ.Nutrition in the neonate. Current Opinionin Pediatrics, 1995, 7(2):146-51.158. Makrides M et al. Fatty acid compositionof brain, retina, and erythrocytes inbreast- and formula-fed infants. The AmericanJournal of Clinical Nutrition, 1994,60(2):189-94.159. Renfrew MJ, Lang S, Woolridge MW.Early versus delayed initiation of breastfeeding.Cochrane Database of SystematicReviews, 2000(2):CD000043.REFERENCES49

160. Klaus M. Mother and infant: early emotionalties. Pediatrics, 1998, 102(5 SupplementE):1244-6.161. Johnson DL et al. Does HOME add tothe prediction of child intelligence overand above SES? The Journal of Genetic Psychology,1993, 154(1):33-40.162. Fergusson DM, Beautrais AL, Silva PA.Breastfeeding and cognitive developmentin the first seven years of life. Social Scienceand Medicine, 1982, 16(19):1705-8.163. Anderson JW, Johnstone BM, Remley DT.Breastfeeding and cognitive development:a meta-analysis. The American Journal ofClinical Nutrition, 1999, 70(4):525-35.164. Drane DL, Logemann JA. A critical evaluationof the evidence on the associationbetween type of infant feeding and cognitivedevelopment. Paediatric and PerinatalEpidemiology, 2000, 14(4):349-56.165. Lucas A et al. A randomised multicentrestudy of human milk versus formula andlater development in preterm infants.Archives of Disease in Childhood. Fetal andNeonatal Edition, 1994, 70(2):F141-6.166. Pollock JI. Long-term associations with infantfeeding in a clinically advantaged populationof babies. Developmental Medicine andChild Neurology, 1994, 36(5):429-40.167. Lucas A et al. Breast milk and subsequentintelligence quotient in children bornpreterm. Lancet, 1992, 339(8788):261-4.168. Horwood LJ, Fergusson DM. Breastfeedingand later cognitive and academic outcomes.Pediatrics, 1998, 101(1):E9.169. Greene LC et al. Relationship betweenearly diet and subsequent cognitive performanceduring adolescence. BiochemicalSociety Transactions, 1995, 23(2):376S.170. Rogan WJ, Gladen BC. Breastfeeding andcognitive development. Early Human Development,1993, 31(3):181-93.171. Jain A, Concato J, Leventhal JM. Howgood is the evidence linking breastfeedingand intelligence? Pediatrics, 2002, 109(6):1044-53.172. Wigg NR et al. Does breastfeeding at sixmonths predict cognitive development?Australian and New Zealand Journal of PublicHealth, 1998, 22(2):232-6.173. Johnson DL et al. Breastfeeding and children’sintelligence. Psychological Reports,1996, 79(3 Pt 2):1179-85.174. Morrow-Tlucak M, Haude RH, ErnhartCB. Breastfeeding and cognitive developmentin the first 2 years of life. Social Scienceand Medicine, 1988, 26(6):635-9.175. Rodgers B. Feeding in infancy and laterability and attainment: a longitudinalstudy. Developmental Medicine and Child Neurology,1978, 20(4):421-6.176. Jacobson SW, Jacobson JL. Breastfeedingand intelligence. Lancet, 1992,339(8798):926.177. Lucas A et al. Early diet in preterm babiesand developmental status in infancy.Archives of Disease in Childhood, 1989,64(11):1570-8.178. Silva PA, Buckfield P, Spears GF. Somematernal and child developmental characteristicsassociated with breast feeding:a report from the DunedinMultidisciplinary Child DevelopmentStudy. Australian Paediatric Journal, 1978,14(4):265-8.179. Quinn PJ et al. The effect of breastfeedingon child development at 5 years: a cohortstudy. Journal of Paediatrics and ChildHealth, 2001, 37(5):465-9.180. Clark KM et al. Breastfeeding and mentaland motor development at 5 1/2 years.Ambulatory Pediatrics, 2006, 6(2):65-71.181. Angelsen NK et al. Breastfeeding andcognitive development at age 1 and 5years. Archives of Disease in Childhood, 2001,85(3):183-8.182. Richards M, Hardy R, Wadsworth ME.Long-term effects of breastfeeding in anational birth cohort: educational attainmentand midlife cognitive function.Public Health Nutrition, 2002, 5(5):631-5.50EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES

183. Victora CG et al. Breastfeeding and schoolachievement in Brazilian adolescents. ActaPaediatrica, 2005, 94(11):1656-60.184. Summerbell CD et al. Interventions forpreventing obesity in children. CochraneDatabase of Systematic Reviews, 2005(3):CD001871.185. Whelton SP et al. Effect of aerobic exerciseon blood pressure: a meta-analysisof randomized, controlled trials. Annalsof Internal Medicine, 2002, 136(7):493-503.187. He FJ, MacGregor GA. Effect of longertermmodest salt reduction on blood pressure.Cochrane Database of Systematic Reviews,2004(3):CD004937.188. Pan XR, Li GW, Hu YH, Wang JX, YangWY, An ZX, et al. Effects of diet and exercisein preventing NIDDM in peoplewith impaired glucose tolerance. The DaQing IGT and Diabetes Study. DiabetesCare, 1997, 20(4):537-44.186. Ebrahim S, Smith GD. Systematic reviewof randomised controlled trials of multiplerisk factor interventions for preventingcoronary heart disease. British MedicalJournal, 1997, 314(7095):1666-74.REFERENCES51

52EVIDENCE ON THE LONG-TERM EFFECTS OF BREASTFEEDING: SYSTEMATIC REVIEWS AND META-ANALYSES