1 Lancet Breastfeeding Series Paper 1

3 Breastfeeding in the 21st Century:

4

5 epidemiology, mechanisms and lifelong impact.

6
7
8
9 Prof. Cesar G Victora, MD1
10 Rajiv Bahl, MD2
11 Aluísio JD Barros, MD1
12 Giovanny V A França, PhD1
13 Prof. Susan Horton, PhD3
14 Julia Krasevec, MSc4
15 Prof. Simon Murch, PhD5
16 M. Jeeva Sankar, DM6
17 Neff Walker, PhD7
18 Nigel Rollins, MD2
19
20 on behalf of the Lancet Breastfeeding Series Group*
21
22
23 * Katie Allen, Nita Bhandari, Ranadip Chowdhury, Shyamali Dharmage, Elsa Giugliani, Nemat
24 Hajeebhoy, Bernardo Horta, Caroline Lodge, Chessa Lutter, Fatima Maia, Jose C. Martines,
25 Christian Loret de Mola, Karen G. Peres, Ellen Piwoz, Linda Richter, Bireshwar Sinha, Sunita Taneja

1
International Center for Equity in Health, Post‐Graduate Programme in Epidemiology, Federal University of
Pelotas, Brazil
2
Department of Maternal, Newborn, Child and Adolescent Health (MCA), World Health Organization,
Geneva, Switzerland
3
Department of Economics, University of Waterloo, Canada
4
Data and Analytics Section, Division of Data, Research, and Policy, UNICEF, New York, USA
5
University Hospital Coventry & Warwickshire, Coventry, UK
6
WHO Collaborating Centre for Training and Research in Newborn Care, All India Institute of Medical
Sciences (AIIMS), New Delhi, India
7
Institute for International Programs, Bloomberg School of Public Health, Baltimore, USA

© 2016. This manuscript version is made available under the Elsevier user license
http://www.elsevier.com/open-access/userlicense/1.0/
26 Abstract
27
28 The importance of breastfeeding (BF) in low and middle‐ income countries (LMICs) is well
29 recognized, but there is less consensus on its importance in high‐income countries (HICs). In
30 LMICs, only 37% of children under 6 months are exclusively breastfed. With few exceptions, BF
31 duration is very short in HICs. Our meta‐analyses indicate protection against child infections and
32 malocclusion, and increased intelligence, as well as likely reductions in overweight and diabetes.
33 We did not find associations with allergic conditions including asthma, blood pressure or
34 cholesterol, and there was an increase in dental caries. For nursing women, there was protection
35 against breast cancer and increased birth spacing, as well as likely protection against ovarian
36 cancer and type 2 diabetes. Scaling up BF to near universal level could prevent 823,000 underfive
37 deaths a year, as well as 20,000 deaths due to breast cancer. Recent epidemiological and
38 biological findings expand upon the known benefits of BF for women and children, rich and poor
39 alike.

40
41 Funding
42 Bill and Melinda Gates Foundation and the Wellcome Trust

43

2
44 Introduction
45
46 “In all mammalian species the reproductive cycle comprises both pregnancy and breast‐feeding: in
47 the absence of latter, none of these species, man included, could have survived”, wrote Bo
48 Vahlquist in 1981.1 Three years earlier, Derek and Patrice Jelliffe in their classical book “Breast milk
49 in the modern world”2 stated that “breast‐feeding is a matter of concern in both industrialised and
50 developing countries because it has such a wide range of often underappreciated consequences.”3
51 The Jelliffe’s anticipated that breastfeeding (BF) would be relevant to “present‐day interest in the
52 consequences of infant nutrition on subsequent adult health”. 3 These statements were challenged
53 by the American Academy of Pediatrics in 1984: “if there are benefits associated with breast‐
54 feeding in populations with good sanitation, nutrition and medical care, the benefits are
55 apparently modest”.4
56 In last three decades, BF recommendations have evolved markedly (webappendix 1). Results from
57 epidemiological studies and growing knowledge of the roles of epigenetics, stem cells and the
58 developmental origins of health and disease strongly support the ideas put forward by Vahlquist
59 and the Jelliffe’s. Never before in the history of science, has so much has been known about the
60 complex importance of BF for both mothers and children.
61 In the first of two Lancet articles, we describe current patterns and past trends in BF throughout
62 the world, review its short and long‐term health consequences for the child and mother, estimate
63 potential lives saved by scaling up BF, and summarize recent insights on how BF may permanently
64 shape the life course.
65 The second article5 covers the determinants of BF and the effectiveness of promotion
66 interventions. It discusses the role of BF in HIV transmission and how knowledge on this issue
67 evolved in recent years, and examines the lucrative market of breastmilk substitutes, the
68 environmental role of BF, and its economic implications.
69 In the context of the post‐2015 development agenda, the two articles document how essential BF
70 is for building a better world for future generations in all countries, rich and poor alike.
71

72 Methods
73
74 The following standard indicators were studied:6
75 ‐ Early initiation of BF: proportion of children born in the last 24 months who were put to the
76 breast within one hour of birth;
77 ‐ Exclusive BF under 6 months: proportion of infants 0–5 months of age who are fed exclusively
78 with breastmilk; this indicator is based on the diets of infants under 6 months during the 24
79 hours before the survey (in order to avoid recall bias), not on the proportion who are
80 exclusively breastfed for the full 6‐month period;
81 ‐ Continued BF at 1 year (12‐15 months): proportion of children 12–15 months of age who are
82 fed breastmilk;
83 ‐ Continued BF at 2 years (20‐23 months): proportion of children 20‐23 months of age who are
84 fed breastmilk.

3
 85 Because few high‐income countries (HICs) report on the above indicators, additional indicators
86 were calculated to allow global comparisons:
87 ‐ Ever BF: infants reported to have been breastfed, even if for a short period;
88 ‐ BF at 6 months: in HICs, the proportion of infants who were breastfed for 6 months or longer;
89 in LMICs with standardized surveys, the proportion of infants aged 4‐7 months (median age of
90 6 months) who are breastfed;
91 ‐ BF at 12 months: in HICs, the proportion breastfed for 12 months or longer; in LMICs, the
92 proportion of children aged 10‐13 months (median age of 12 months) who are breastfed.
93 The last three indicators were used exclusively for comparisons between HICs and LMICs.
94 Otherwise, we report on the standard international indicators (see webappendix 2)
95 For low and middle‐income countries (LMICs), we reanalysed national surveys carried out since
96 1993, including Demographic and Health Surveys (DHS)
97 (http://www.measuredhs.com/aboutsurveys/dhs/start.cfm), Multiple Indicator Cluster Surveys
98 (MICS) (http://www.childinfo.org/) and others (see webappendix 3). Virtually all had response
99 rates above 90% and adopted standardized questionnaires and indicators.
100 For all HICs with 50,000 or more annual births, we carried out systematic reviews of the published
101 and grey literature and contacted key informants (see webappendix 4). Many countries did not
102 have information on BF from national samples. Response rates were often in the 50‐70% range,
103 indicators were seldom standardized and recall periods tended to be long. Administrative or other
104 data were used when surveys were not available. If needed, we estimated the proportion of
105 infants breastfed at 12 months based on information available for BF at 6 months, and vice‐versa.
106 Time trends were obtained using multilevel linear regression models (hierarchical mixed models)
107 that take into account that countries contribute with two or more surveys for the analysis.
108 Departures from linearity were explored using fractional polynomial regression models.7 In all
109 analyses, country data were weighted by their populations of children under two years.
110 Webappendix 5 describes statistical methods.
111 Systematic searches of the literature and, whenever possible, meta‐analyses were carried out for
112 outcomes postulated to be associated with breastfeeding (see webappendix 6).
113 The Lives Saved Tool (LiST)8 was used to predict how many deaths of children under five years
114 would be prevented if 2013 BF patterns were scaled up in the 75 countries that are part of the
115 Countdown to 2015,9 which account for over 95% of all such deaths globally. We assumed that
116 95% of children under one month and 90% of those under six months would be exclusively
117 breastfed, and 90% of those aged 6‐23 months would be breastfed. The relative risks for the
118 protection against all infectious causes of death obtained from our new meta‐analyses were
119 applied to all infectious causes of death among children under two years of age, and also to the
120 15% of deaths due to complications of prematurity that occur after the first week (see
121 webappendix 7).
122 We also estimated the potential number of breast cancer deaths preventable by extending BF
123 duration (see webappendix 8).
124
125 Role of the Funding Source
126
127 This review was supported by a grant from the Bill and Melinda Gates Foundation to the World
128 Health Organization. CGV is recipient of a Wellcome Trust Senior Investigator Award. The sponsors

4
129 of the study had no role in study design, data collection, data analysis, data interpretation, or
130 writing of the paper.

131

132 Epidemiology: levels and trends

133

134 We obtained complete information for 127 out of all 139 LMICs (webappendix 3), accounting for
135 99% of the children from such countries. For HICs, we were able to obtain data on 37 out of 75
136 countries, but for several countries only a subset of the indicators were available (webappendix 4),
137 so that these data must be interpreted with caution.

138 Figure 1 shows the proportion of children aged around 12 months who are breastfed in 153
139 countries. The highest prevalences are in Sub‐Saharan Africa, South Asia and parts of Latin
140 America. In most HICs, BF prevalence at 12 months of age is below 20% (webappendix 4).
141 Important differences were observed, for example, between the UK (<1%) and the USA (27%), and
142 between Norway (35%) and Sweden (16%).

143 Figure 2 shows BF indicators according to country income groups. Information on early initiation,
144 exclusive BF or continued BF at 2 years was not available for most HICs. We found a strong inverse
145 correlation (Pearson’s r=‐0.84; P<0.001; see webappendix 9) with log gross domestic product per
146 capita; regression analyses showed that for each doubling in the latter, BF prevalence at 12
147 months decreased by 10 percent points.
148 The vast majority of mothers in all country groups started breastfeeding; only three countries
149 (France, Spain and USA) had rates below 80%. However, early initiation is low in all settings, and so
150 is exclusive breastfeeding. BF at 12 months is frequent in low and lower‐middle‐income settings,
151 but uncommon elsewhere.
152 Except for early initiation, prevalence of all indicators decreases with increasing national wealth.
153 Low‐income countries show high BF prevalence at all ages, but initiation and exclusive BF rates
154 are unsatisfactory even in these countries.
155 Surprisingly, most national level breastfeeding indicators are not strongly correlated
156 (webappendix 9). There is only a moderate correlation (Pearson’s r = 0.54) between exclusive and
157 of continued BF at 1 year in LMICs . Whereas continued BF prevalence is high throughout West
158 and Central Africa (top left of Figure 3), the exclusive breastfeeding rates vary widely. Countries
159 from Eastern and Southern Africa tend to have somewhat lower rates of continued BF but higher
160 rates of exclusive BF than their Western Africa neighbours. In Latin America and the Caribbean, as
161 well as in the CEE&CIS region, both indicators tend to be lower than in Africa. South Asia shows
162 relatively high rates of both indicators whereas the Middle East / North African countries present
163 lower rates. Finally, countries from East Asia and the Pacific region show moderate to high
164 prevalences of both indicators.

165 Among children aged under 6 months in LMICs, 36.3 million (63%) were not exclusively BF at the
166 time of the last national survey. The corresponding percentages for low, lower middle and upper
167 middle income countries were 53%, 61% and 63%, respectively. Among 6‐23 month old children,
168 64.8 million (37%) were not receiving any breastmilk, with corresponding rates of 18%, 34% and

5
169 55% in the three country groupings. In total, 101.1 million children in LMICs are not breastfed
170 according to international recommendations.
171 Most LMICs have data from several surveys over time, making it possible to explore time trends
172 both at national level and for children in the poorest and richest 20% of families. Our analyses
173 were restricted to surveys for which breakdown of BF indicators by wealth quintiles was possible
174 (214 surveys for exclusive and 217 for continued BF; see webappendix 5) representing 83% of the
175 total 2010 population of children under two years in LMICs. We report linear trends as there was
176 no evidence of departures from linearity. Exclusive BF rates increased slightly from 24.9% in 1993
177 to 35.7% in 2013 (Figure 4). Among the richest 20%, the increase was much steeper, from 15.8%
178 to 35.5%, respectively, whereas the poorest 20% followed the general trend. Continued BF at 1
179 year (12‐15 months) dropped slightly at global level (from 76.0% to 73.3%), particularly due to a
180 decline among the poor, as prevalence among the rich was stable.
181 In summary, early initiation and exclusive BF rates are below 50% in most countries. Continued BF
182 is one of the few nutrition indicators for which poor countries perform better than rich countries.
183 Unusually, data availability and quality is better in poor than in rich countries, which may reflect
184 how little importance BF receives in the latter. Exclusive and continued BF prevalence are only
185 moderately correlated, suggesting that countries require different promotion strategies. In
186 addition to national income levels, there are also important regional differences in BF patterns.
187 For LMICs as a whole, exclusive BF rates are increasing slightly, while prevalence of continued BF
188 at 1 year is declining.

189

190 Epidemiology: within‐country inequalities

191
192 We analysed 98 surveys from LMICs to investigate within‐country inequalities according to wealth
193 quintile (webappendix 10). We found that wealth related inequalities in exclusive BF are small but
194 disparities in continued BF rates are consistent: the poor tend to breastfeed for longer than the
195 rich in all country groupings, but particularly in middle‐income countries. Similar results based on
10
196 33 countries have been recently reported.
197 The literature review of studies from high‐income countries showed that high‐income, educated
198 women show higher BF rates. 11‐19 BF initiation in the USA was more common among mothers
199 with lower education up to the 1960’s but since then the social gradient had been reversed.4
200 BF is one of few positive health related behaviours in LMICs that is less frequent among the rich,
201 both between and within countries. The low rates of continued BF among the rich raise the
202 possibility that poor mothers will move towards breastmilk substitutes, a concern that is
203 reinforced by declining rates among the poor.

204

205 Short term effects on the child: mortality and morbidity

206
207 Table 1 shows the results of 28 meta‐analyses on the associations between BF and outcomes in
208 the child and mother, of which 22 were commissioned for this series.

6
209 Because studies varied regarding feeding classifications, for several outcomes we compared longer
210 versus shorter BF durations (e.g., never versus ever BF; BF for less or more than a given number of
211 months; and for a few outcomes longer versus shorter durations of exclusive BF). We tested for
212 heterogeneity due to the type of BF categorization, and in its absence we pooled the different
213 definitions. We describe the results of randomized trials on how BF promotion affects health,
214 nutrition or development outcomes, but not of trials where the endpoint was restricted to BF
215 indicators; these are reviewed in the second article in the series.5
216 Table 1 includes the results of the meta‐analyses, and presents their potential limitations,
217 including confounding, effect modification, reverse causation and publication bias.
218 Only three studies provide information on mortality according to exclusive, predominant, partial,
219 or no BF in the first six months of life. A strong protective effect is evident, with exclusively
220 breastfed infants showing only 7% of the risk of death as those who were not breastfed. Another
221 three studies showed that infants younger than six months who were not breastfed presented 3.5
222 (boys) and 4.1 (girls) fold increases in mortality compared to those who received any breastmilk,
223 and that protection declined with age.20 These results are supported by studies on children aged 6‐
224 23 months, among whom any BF was associated with a 50% reduction in deaths (Table 1).
225 The above studies were from LMICs, but BF may also protect against deaths in HICs. A meta‐
226 analysis of 6 high‐quality studies showed that ever BF was associated with a 36% (95% CI 19;49%)
227 reduction in sudden infant deaths.21. Another meta‐analysis of 4 RCTs showed a 58% (95% CI
228 4;82%) decline in necrotizing enterocolitis,21 a condition with high case‐fatality in all settings.22
229 In terms of child morbidity, there is overwhelming evidence from tens of studies, mostly from
230 LMICs, that BF protects against diarrhoea and respiratory infections (Table 1), including three
231 randomized controlled trials.23 Approximately half of all diarrhoea episodes and one third of
232 respiratory infections would be avoided by BF. Protection against hospital admissions due to these
233 diseases is even greater: 72% and 57%, respectively.
234 The risks associated with breastmilk substitutes in terms of biological and chemical contamination
235 are discussed in webappendix 11.
236 Our reviews suggest important protection against otitis media in children under 2 years, mostly
237 from high‐income settings, but inconclusive findings for older children (Table 1). There was no
238 clear evidence of protection against allergic conditions: no association with eczema or food
239 allergies and some evidence of protection against allergic rhinitis under age 5 years. When the 29
240 studies on asthma were analysed, there was statistical evidence of a 9% (95% CI 2;15%) reduction,
241 but effects were smaller and non‐significant when analyses were restricted to 16 studies with
242 tighter control of confounding – reduction of 5% (95% CI ‐6;15%) – or to the 13 cohort studies –
243 6% reduction (95% CI ‐11;20%).
244 Our analyses of oral health outcomes (Table 1) found that BF was associated with a 68% reduction
245 (95% CI 60;75%) in malocclusions, based on 49 studies carried out mostly in LMICs. Most studies
246 were restricted to young children with deciduous teeth, but malocclusion in this age group is a risk
247 factor for malocclusion in permanent teeth.24,25 On the other hand, BF longer than 12 months and
248 nocturnal feeding were associated with 2‐3 fold increases in dental caries in deciduous teeth,
249 possibly due to lack of oral hygiene after feeding.
250 Information on BF and child growth was derived from 17 studies, including 15 RCTs, mostly from
251 middle‐income countries. Attained weight and length at around 6 months did not differ, but there

7
252 was a small reduction (‐0.06 Z scores; 95% CI 0.00;‐0.12) in body mass index or weight for length
253 among children in the BF intervention group, based on intent‐to‐treat analyses (Table 1).

254 Long term effects: Obesity, NCDs and intelligence

255
256 We updated existing meta‐analyses 26 on the associations between BF and outcomes related to
257 non‐communicable diseases (Table 1). Most studies are from high‐income settings.
258 Based on all 113 studies identified, longer BF was associated with a 26% reduction (95% CI 22;
259 30%) in the odds of overweight and/or obesity prevalence.27 The effect was similar in studies from
260 high and from low or middle‐income countries. The only BF promotion trial that reported on this
261 outcome did not detect an association; in this trial, there were important early differences
262 between intervention and comparison groups in terms of exclusive BF, but by age 12 months only
263 19% and 11% of children, respectively, were breastfed.28 29 A 2005 meta‐analysis30 of
264 breastfeeding and mean body mass index included 36 articles of which 11 included adjustment for
265 socioeconomic status, maternal smoking and maternal body mass index; their pooled effect did
266 not suggest an association with BF. In our review,27 23 high quality studies with sample sizes
267 greater than 1,500 and adjustment socioeconomic status, maternal BMI and perinatal morbidity
268 showed a pooled reduction of 13% (95% CI 6;19%). As mentioned above, the pooled analysis of
269 randomized trials of breastfeeding promotion in children suggested a small reduction (‐0.06 Z
270 scores; 95% CI 0.00;‐0.12) in body mass index or weight for length.
271 For type‐2 diabetes incidence, the pooled results from 11 studies indicate a 35% reduction (95% CI
272 14%;51%). Only three studies were deemed to be of high quality, indicating a potentially
273 important, but non‐significant reduction of 24% (95% CI ranging from a 60% reduction to a 47%
274 increase). The direction and magnitude of the association with diabetes are consistent with
275 findings on overweight. An earlier review indicated a possible protection against type‐1 diabetes,
276 based on a small number of studies.21
277 The meta‐analyses for systolic (43 studies) and diastolic (38 studies) blood pressure, and total
278 cholesterol (46 studies) provide no evidence of protective effects of BF.
279 BF was consistently associated with higher performance in intelligence tests among children and
280 adolescents, with a pooled increase of 3.4 IQ points (95% CI 2.3;4.6) based on 16 observational
281 studies that controlled for several confounding factors including home stimulation (Table 1). Nine
282 studies also adjusted for maternal intelligence, showing a pooled effect of 2.6 points (95% CI
283 1.3;4.0). A large randomized trial reported an increase of over 7 IQ points at 6.5 years of age,31 and
284 a similar effect was reported in a non‐randomized trial where preterm infants were fed formula or
285 breastmilk.32 Positive associations with attained schooling were reported from Great Britain,33 34
286 New Zealand35 and Brazil,36 but a joint analysis of four LMIC cohorts showed mixed results.37 A
287 recent publication from Brazil including 30‐years follow up suggested an impact of breastfeeding
288 on intelligence, attained schooling and adult earnings; the latter was largely mediated through
289 increased IQ.38
290 A 2007 review suggested that BF may reduce the risk of acute lymphocytic and acute
291 myelogenous leukemias. 21

292 Effects on the mother

293

8
294 Table 1 shows the results of new reviews on lactational amenorrhea, breast and ovarian cancer,
295 type 2 diabetes and osteoporosis. We also cite existing reviews on diabetes, weight retention and
296 maternal depression. Most studies are from HICs, except for lactational amenorrhea.
297 The role of BF in birth spacing is well recognized. In 2003, it was estimated that in countries where
298 continued BF is prevalent, such as Uganda and Burkina Faso, 50% more births would be expected
299 in the absence of BF.39 Our review (Table 1) confirms that increased breastfeeding, and in
300 particular exclusive or predominant BF, were associated with longer periods of amenorrhea. BF
301 promotion RCTs also confirm this effect.40
302 There is evidence of a robust inverse association between BF and breast cancer (Table 1). The
303 largest individual level analysis on this topic included about 50,000 cancer cases from 47 studies,41
304 representing about half those included in our meta‐analysis. Each 12‐month increase in lifetime BF
305 was associated with a reduction of 4.3% (95% CI 2.9;6.8%) in the incidence of invasive breast
306 cancer. This analysis included thorough adjustment parity and other confounders; nulliparous
307 women were excluded. The results did not vary significantly according to menopausal status. Our
308 meta‐analysis suggests a higher magnitude of protection, but when restricted to the 14 studies
309 with fine adjustment for parity and exclusion of nulliparae women, the reduction comparing
310 longer versus shorter BF durations was equal to 7% (95% CI 3%;11%).
311 The meta‐analysis of 41 studies on BF and ovarian cancer shows a 30% reduction associated with
312 longer BF (95% CI 25;36%). Confounding by parity may affect the results but socioeconomic
313 confounding is unlikely. The pooled reduction, based on studies with fine adjustment for parity
314 and exclusion of nulliparae women, was 18% (95% CI 14;42%).
315 We also reviewed the evidence on osteoporosis, finding no evidence of an association between BF
316 and bone mineral density in the four studies available (Table 1).
317 Recent reviews were available for maternal type‐2 diabetes, postpartum weight change and
318 depression. A meta‐analysis of six cohort studies on diabetes showed an odds ratio of 0.68 (95% CI
319 0.57;0.82).42 In light of this finding, one would also predict an association with overweight, but a
320 review of 54 articles on the possible role of BF on postpartum weight change was inconclusive.43
321 Few studies are available the long term association between nursing and adiposity. After the latter
322 review was published, an analysis of 740,000 British women with long‐term follow‐up showed that
323 mean BMI was 1% lower for every 6 months that the woman had breastfed.44 A qualitative review
324 of 48 studies showed clear associations between BF and reduced maternal depression,45 but it is
325 more likely that depression affects BF than the opposite.

326 Estimating lives saved for children and mothers

327
328 The Lives Saved Tool8 estimates that 823,000 annual deaths would be saved in 75 high‐mortality
329 LMICs in 2015 if breastfeeding was scaled up to near universal levels. This corresponds to 13.8% of
330 the deaths of children under two years of age. Among preventable deaths, 87% would have
331 occurred among infants under six months, due to a combination of high death rates and low
332 prevalence of exclusive BF.
333 We also estimated potential impact on breast cancer mortality (webappendix 7). Globally, existing
334 rates of BF avert an estimated 19,464 annual breast cancer deaths relative to a scenario in which
335 no women breastfed, using the estimates of protection from the pooled study (Table 1).41 The low‐
336 income regions with long BF durations (Africa and South Asia) account for 58% of currently
337 prevented deaths, despite only accounting for 36% of the global population included in this

9
338 analysis. We also estimate that an additional 22,216 lives a year would be saved by increasing BF
339 duration from current levels to 12 months per child in high‐income countries, and 2 years per child
340 in LMICs. We cannot model the same effect in all countries given the differences in data
341 availability, and the fact that very few children in high‐income countries are breastfed longer than
342 12 months. Latin America, CEE/CIS and the high‐income countries would benefit most because of
343 their higher incidence of breast cancer and also shorter BF durations.

344 Conclusions
345
346 As described by Vahlquist,1 the reproductive cycle includes pregnancy and BF. The latter has been
347 largely neglected by medical practice, leading to the assumption that breastmilk may be replaced
348 with artificial products without important detrimental consequences. This is particularly critical in
349 HICs, where fewer than one in every five children are breastfed by the age of 12 months. For each
350 doubling in national GDP per capita, BF prevalence at 12 months decreases by 10 percent points.
351 Modern epidemiology and biology confirm that relinquishing BF has major long‐term effects on
352 the health, nutrition and development of the child as well on women’s health. Possibly, no other
353 health behaviour can influence such varied outcomes in the two individuals who are involved: the
354 mother and the child. Recent findings on immunology, epigenetics, microbiome and stem cells will
355 likely be followed by other, even more exciting discoveries on the exquisite personalized medicine
356 provided by human milk.
357 Our global analyses show that over 80% of newborns receive breast milk in virtually all countries.
358 However, only about half are put to the breast within the first hour, even though such
359 recommendation was issued more than 25 years ago.46 As 60% of the world’s children are now
360 delivered by skilled assistants,9 further promotion of early initiation is possible. In most countries,
361 exclusive BF rates are well below 50%, and the correlation with the duration of any BF is only
362 moderate. This signals the need to tailor BF support strategies to specific patterns observed in
363 each country. In the poorest countries, late initiation and low rates of exclusive BF are the main
364 challenges. In middle and high‐income countries, short overall duration of BF represents an
365 additional challenge.
366 Our time trend analyses show that, for LMICs as a whole, exclusive BF increased by about 0.5
367 percent point a year since 1993, reaching 35% in 2013. In 2012, the 56th World Health Assembly
368 set as a target for 2025 to “increase the rate of exclusive breastfeeding in the first 6 months up to
369 at least 50%”.47 This requires a doubling of the recent annual increase, to more than 1 percent
370 point a year in the next decade, which is already the rate for the richest 20%. In light of the
371 benefits of exclusive BF and of current achievements by leading countries, one wonders if a more
372 ambitious target should not be aimed at. The Assembly did not set a goal for continued BF.
373 In terms of inequalities, we show that BF is one of the few positive health behaviours that is more
374 prevalent in poor than in rich countries. We also show that poor women breastfeed for longer
375 than rich women in LMICs, whereas in HICs the pattern is in the opposite direction. These results
376 suggest that BF patterns are currently contributing to reducing the health gaps between rich and
377 poor children in LMICs, which would be even greater in the absence of BF.

378 In LMICs, there are no inequalities between rich and poor mothers in exclusive BF rates. Time
379 trend analyses suggest that this is because rich mothers are adopting exclusive BF at a much faster
380 rate than the poor; only 20 years ago, the latter had substantially higher rates of exclusive BF. In

10
381 contrast, continued BF is still more common among the poor, but rates are dropping among these
382 while remaining stable among the rich. Protecting BF among the world’s poor is therefore a major
383 priority.

384 Our systematic reviews emphasize how important BF is for all women and children, regardless of
385 where they live and irrespective of whether they are rich or poor. Appropriate BF practices
386 prevent child morbidity due to diarrhoea, respiratory infections and otitis media. Where infectious
387 diseases are common causes of death, BF provides major protection, but even in high‐income
388 populations it lowers mortality due to causes such as necrotizing enterocolitis and sudden infant
389 death syndrome. Current evidence shows that BF enhances human capital by increasing
390 intelligence. BF also helps nursing women by preventing breast cancer. In addition, our review
391 suggests likely effects on overweight and diabetes in breastfed subjects, and ovarian cancer and
392 diabetes in mothers.

393 The only harmful consequence of BF we detected was an increase in dental caries among children
394 breastfed for over 12 months; in light of the many benefits of BF, this observation should not lead
395 to discontinuation of BF but rather to improved oral hygiene.

396 Findings from our systematic reviews are limited by the observational nature of most of the
397 available data on BF and by the limitations of meta‐analyses.48,49 Experimental data are scant
398 because BF promotion activities must be highly effective to change feeding patterns to an extent
399 that leads to measureable impact on short and long‐term outcomes.

400 Moreover, confounding may occur because BF is associated with higher socioeconomic position in
401 HICs. Our reviews included sub‐analyses of studies with tight control for confounding. Whenever
402 possible, we also carried out separate analyses of studies from LMICs, because the poor tend to
403 breastfeed for longer in these countries (webappendix 10), in contrast to what is observed in HICs.
404 Interpretation of associations is also affected by the fact that non‐breastfed infants receive
405 different diets in different countries, for example animal milk in most poor societies and formula in
406 middle and high‐income populations. The association between BF and overweight, for example, is
407 likely affected by the diet of infants who are not breastfed.

408 There is no consensus regarding whether breastfeeding protects against overweight and
409 diabetes,21,26,30 particularly because of potential residual confounding. Although the evidence is
410 not as strong as for infections or intelligence, we argue that there is growing evidence of
411 protection. Our meta‐analyses showed that the association persisted when restricted to higher
412 quality studies, and was also present in studies from low and middle income settings. The
413 association appears to be specific: for example we found no effect on blood pressure or blood lipid
414 levels, for which confounding patterns are similar. Last, the randomized trials of breastfeeding
415 promotion in infancy indicate a reduction in adiposity.
416 Scaling up BF practices to almost universal levels is estimated to prevent 823,000 annual deaths,
417 or 13.8% of all deaths of children under the age of 24 months in the 75 Countdown to 20159
418 countries. It should be noted that the target of 95% exclusive BF under six months is ambitious, as
419 currently the highest national prevalences are 85% for Rwanda and 76% for Sri Lanka. We also
420 used a target of 90% for any BF from 6‐23 months, but five countries already have levels that are
421 above this target (Nepal, Rwanda, Ethiopia, Burundi and Guinea). We acknowledge that these
422 targets are ambitious, but the estimates show the potential for lives saved if mothers and children

11
423 adhered to international recommendations. In spite of differences in methods, our estimates are
424 consistent with those from the 2013 Lancet Nutrition Series (804,000 deaths),50 but are higher
425 than those from the 2010 Global Burden of Disease study (540,000 deaths);51 in the latter, the
426 assumptions and methods are not sufficiently detailed to understand the reasons for the
427 discrepancy. BF is potentially one of the top interventions for reducing under‐five mortality, and
428 the modest changes in BF rates since 2000 have contributed to the fact that most LMICs failed to
429 reach the fourth Millennium Development Goal.52 We show that increasing exclusive BF should be
430 the first priority for reducing infant deaths.

431 As an example of the potential to save women’s lives, we estimated that current rates of BF
432 prevent almost 20,000 annual breast cancer deaths, and another 20,000 are preventable by
433 scaling up BF practices (webappendix 8).

434 To achieve its full impact, BF should continue up to the age of two years. Protection against
435 infectious diseases mortality and morbidity extends well into the second year of life; for example,
436 BF prevents half of deaths due to infections among children aged 6‐23 months. Protection against
437 otitis media, a common childhood illness throughout the world, also extends to two years and
438 possibly beyond. Studies of overweight and obesity show that longer durations of BF are
439 associated with lower risk, as do studies of IQ showing a clear dose‐response association with
440 duration. Breast cancer is reduced by lifetime duration of BF among women, with a 6% reduction
441 for each 12 months.41 Ethnographic research shows that total BF duration ranges between two
442 and four years in most traditional societies,53 and our review of the literature supports current
443 international recommendations about the total duration of BF, in rich and poor countries alike.

444 Data availability on BF patterns shows a highly unusual pattern. HICs seem to neglect BF to such an
445 extent that most of them are unable to report on reliable, standardized indicators. This is in sharp
446 contrast to the high data quality on LMICs, as a consequence of the regular conduct of
447 standardised surveys such as DHS and MICS. This is a rare inversion of the usual global distribution
448 of availability and coverage of health statistics.

449 At a time when the launch of the Sustainable Development Goals is approaching, we were able to
450 demonstrate how essential the protection, promotion and support of BF is for achieving many of
451 the goals by 2030. BF is clearly relevant to the 3rd goal on health, which includes not only maternal
452 and child health but also non‐communicable conditions and diseases such as breast cancer,
453 diabetes and overweight/obesity. It is relevant to the 2nd goal on nutrition, as well. The impact of
454 BF on intelligence and on human capital is relevant to the 4th (education), 1st (poverty) and 8th
455 (inclusive economic growth) goals. Finally, by helping close the gap between rich and poor, BF can
456 contribute to goal number 10 – reducing inequalities.

457
458
459
460

461

12
462 Panel: Search Strategy
463
464 Information on the associations between BF and outcomes in the child or mother was obtained
465 from 28 systematic reviews and meta‐analyses, of which 22 were commissioned for this article.
466 We reviewed the following conditions for young children: child mortality; diarrhoea incidence and
467 hospitalization; lower respiratory infections incidence, prevalence and hospitalization; acute otitis
468 media; eczema; food allergies; allergic rhinitis; asthma or wheezing; infant growth (length, weight
469 and BMI or weight/length); dental caries; and malocclusion. Among older children, adolescents
470 and adults, systematic reviews were carried out for systolic and diastolic blood pressure;
471 overweight and obesity; total cholesterol; type‐2 diabetes and intelligence. For mothers,
472 systematic reviews covered the following outcomes: lactational amenorrhea; breast and ovarian
473 cancer; type‐2 diabetes; postpartum weight change and osteoporosis.
474 Webappendix 5 describes the search terms.
475
476

477 Panel: Key Messages

478
479  Children who are breastfed have lower infectious morbidity and mortality, fewer dental
480 malocclusions and increased intelligence, which persists until later in life. There is also
481 growing evidence that breastfeeding protects against overweight and diabetes.
482  Breastfeeding also brings benefits to women, including prevention of breast cancer and
483 birth spacing, as well as likely reductions in diabetes and ovarian cancer.
484  High‐income countries show shorter breastfeeding duration than low and middle‐income
485 countries, but even in the latter only 37% of infants under six months are exclusively
486 breastfed.
487  Scaling up breastfeeding is estimated to prevent 823,000 child deaths and 20,000 breast
488 cancer deaths per year.
489  Modern biology is revealing novel mechanisms that characterize breast milk as the
490 ultimate personalized medicine for infants.
491  Breastfeeding promotion is important in rich and poor countries alike, and may contribute
492 to reaching the forthcoming Sustainable Development Goals.

493

13
494 Panel: Breastmilk: the most exquisite personalised medicine.
495
496 The nutritional advantages of breastfeeding and its protection against infection are well
497 recognised. Recently it has become apparent that critical imprinting events may be modulated,
498 with potential lifelong effects for the infant54. These may be mediated directly or through effects
499 upon the infant microbiome. The ability of the microbiome to regulate host responses in infancy
500 depends on individual species, which modulate T cell polarisation and immune regulation,
501 metabolic responses and adipogenesis, possibly even brain development and cognitive functioning
502 .55,56. Abnormal colonisation patterns have deleterious long term impact upon immune and
503 metabolic homeostasis. It is thus remarkable that a mother’s breastmilk transmits elements of her
504 own microbiome and immune responses, and also provides specific prebiotics to support growth
505 of beneficial bacteria.
506
507 Delivery mode initially determines whether the gut flora of the mother (vaginal delivery) or the
508 skin flora of the attendants (caesarean section) dominates the initial colonisers.57, which induce a
509 significant immune response in the infant. Feeding mode is the second fundamental determinant
510 of the infant microbiome. Breastfed infants maintain persistent microbial differences,
511 independent of delivery mode 58,59, due to the effects of human milk oligosaccharides (HMO).
512 Human milk contains a much wider variety of sugars than other mammalian milks: up to 8% of its
513 calorific value is provided in the form of undigestible HMO which function as prebiotics to support
514 growth of specific bacteria. They cannot be utilised by most enteric organisms, but notably
515 support growth of Bifidobacterium longum biovar infantis, which has co‐evolved to express the
516 enzymes needed for HMO utilisation.54. Importantly, there is significant inter‐individual variation
517 in maternal HMO production, which in turn underpins the pattern of flora acquisition by the
518 infant60. There is thus specificity of interaction between breast milk and the infant microbiome,
519 causing different bacterially‐induced effects upon the infant’s metabolism and immunity.
520 This specificity of interaction is further underpinned by the mother’s entero‐mammary axis. To
521 maintain her own gut homeostasis, the mother’s intestinal dendritic cells take up individual
522 bacteria from the lumen and transport these to gut lymphoid follicles 55, where T cells are
523 committed to a regulatory phenotype and B cells shifted towards IgA. Programmed dendritic cells
524 and lymphocytes re‐enter the circulation from there before homing back to the gut through
525 interaction between their induced 7 integrins and locally‐expressed mucosal vascular addressin
526 cell adhesion molecule (MAdCAM‐1). MAdCAM‐1 is expressed within mammary endothelium
527 during pregnancy, allowing selective uptake by the breast of gut‐programmed cells.61 The
528 consequences of enteromammary trafficking include release into colostrum and breast milk of
529 dendritic cells containing live maternal gut bacteria, T cells expressing gut‐derived 7 integrins and
530 plasma cells producing IgA specific for maternal gut bacteria. Breast milk thus contains a
531 dominance of immune cells of gut‐related phenotype ( cells, 7+ cells) that have matured within
532 the mother’s intestine.62 Breastmilk cytokines also vary depending upon the mother’s
533 immunological experiences. There is thus a coordinated input to the infant’s nascent mucosal
534 immune system, specific for the mother’s microbiome, in which individual bacterial types are
535 favoured and tolerogenic immune responses transmitted. Caesarean section, perinatal antibiotics
536 and failure to breastfeed are the three major factors that impact on this co‐evolved imprinting
537 process. An important study of flora acquisition and immune responses in primates identified clear
538 differences in both gut bacterial composition and mucosal immune responses in breastfed
539 compared to formula fed macaques , persisting to adult life.63

14
540 In addition to changes mediated through the flora, individual breastmilk components may directly
541 affect epigenetic programming of the infant 64. The usual adverse impact of peroxisome
542 proliferator‐activated receptor‐ (PPAR‐) polymorphisms upon adiposity and metabolism is
543 prevented by breastfeeding, possibly due to the content of PPAR modulating constituents such as
544 long‐chain polyunsaturated fatty acids (PUFA) and prostaglandin J 65. Protection against breast
545 cancer for the breastfeeding mother may also be mediated through PPAR modulation65.
546 Lactoferrin, a major breastmilk component, binds bacterial CpG motifs and blunts mucosal NF‐B
547 responses to the flora. Microvesicles called exosomes are secreted into breast milk, and may
548 inhibit atopic sensitisation dependent on maternal immune experience66. Breastmilk fat globules
549 contain numerous secreted micro‐RNA’s, whose expression is modulated by maternal diet, which
550 are predicted to target multiple genes within the infant (Munch). There is also evidence that
551 multipotential stem cells are secreted into breastmilk and may persist within the infant67.
552 Human breastmilk is thus not only a perfectly adapted nutritional supply for the infant, but
553 probably the most exquisitely specific personalised medicine that he/she is ever likely to receive,
554 given at a time when gene expression is being fine‐tuned for life. This is an opportunity for health
555 imprinting that should not be missed.
556

15
557 Table 1. Results of meta‐analyses on the associations between BF and outcomes in the child and mother.
Types of
Number
comparison Age range Pooled effect Confounding and effect
Outcome of Other biases Conclusions Reference
(breastfeeding of outcome (95% CI) modification
studies
categories)
EFFECTS ON CHILDREN, ADOLESCENTS OR ADULTS ACCORDING TO BF PATTERN
Mortality due Exclusive versus 3 < 6 mo OR: 0.68 All studies from LMICs, where Studies that avoided Consistent Sankar
to infectious predominant (0.52; 0.88) confounding by SEP would reverse causation evidence of major 201568
diseases Exclusive versus 3 OR: 0.35 likely underestimate the effect (BF being stopped protection. Few
partial (0.20; 0.61) of BF. Confounder‐adjusted because of illness) studies used the 4
Exclusive versus 2 OR: 0.07 studies showed similar effects. showed similar BF categories in
none (0.03; 0.16) effects. No evidence young infants, but
Any versus none 9 6‐23 mo OR: 0.48 of publication bias evidence from
(0.38; 0.60) but very few studies other studies
available. comparing any
versus no BF is
very consistent.
Diarrhea More versus less 15 <5 yr RR: 0.69 Most studies from LMICs, Few studies that Strong evidence of Horta
incidence BF (e.g. (0.58;0.82) where confounding would allowed for reverse major protection 201323
exclusive versus 23 < 6 mo RR: 0.37 likely underestimate an effect. causation also against diarrhea
non‐exclusive; (0.27; 0.50) Confounder‐adjusted studies showed protection. morbidity and
predominant 11 6 mo ‐ <5 yr RR: 0.46 showed similar effects. Three Publication bias is hospitalizations,
versus partial; (0.28; 0.78) BF promotion RCTs (not unlikely to explain particularly
Diarrhea partial versus 9 < 5 yr RR: 0.28 included here) showed the findings as among young
hospitalization none; any BF (0.16;0.50) protection against diarrhea results from large infants, based on
versus no BF, morbidity (pooled OR: 0.69 and small studies a large number of
etc.) (0.49; 0.96). were similar. studies.
Lower More versus less 16 < 2 yr RR: 0.68 Most studies from LMICs, Studies that avoided Strong evidence of Horta
respiratory BF (e.g. (0.60; 0.77) where confounding would reverse causation a reduction in 201323
infections exclusive versus likely underestimate the effect showed similar severe respiratory
incidence or non‐exclusive; of BF. Confounder‐adjusted effects. No evidence infections among
prevalence predominant studies showed similar effects. of publication bias. breastfed
Respiratory versus partial; 17 <2 yr RR: 0.43 The only available RCT showed children, based on
infections partial versus (0.33; 0.55) a RR of 0.85 (0.57; 1.27), a a large number of
hospitalization none; any BF non‐significant reduction in studies.
versus no BF, hospitalizations.
etc.)
Acute otitis More versus less 11 ≤2 yr OR : 0.67 Most studies carried out in High heterogeneity Consistent Lodge
media BF (ever versus (0.62, 0.72) HICs. Several studies failed to (I2= 84%) among the evidence of 201569

16
 never; exclusive 5 >2 yr OR: 1.21 adjust for important 5 studies of children reduction in acute
BF at 6 months (0.60; 2.45) confounders. >2 yr. Some otitis media
versus not evidence of during the first 2
exclusive BF at 6 publication bias, yr of life
months; any BF however study size associated with
for ≥ 3‐4 versus only accounted for a longer durations
≤ 3‐4 months) limited proportion of BF, based on 11
of the overall studies. No
heterogeneity. evidence of
protection after 2
yr.
Eczema More versus less 17 ≤2 yr OR: 0.95 About one‐third of the studies Some evidence of No evidence of an Lodge
BF (ever versus (0.85; 1.07) are from LMICs, and results publication bias, association 201569
never; exclusive 20 >2 yr OR: 1.09 are similar to those from HICs. with smaller pooled between BF and
BF at 6 months (0.99; 1.20) Few studies among young effect sizes in larger eczema or food
Food allergies versus not 10 ≤ 5yr OR: 1.07 children account for reverse studies. The 10 allergies.
exclusive BF at 6 (0.90; 1.26) causation. Several studies studies on food
months; any BF 4 >5yr OR: 1.08 failed to adjust for essential allergy in children ≤
for ≥ 3‐4 versus (0.73; 1.26) confounders. The protective 5yr were highly
2
Allergic rhinitis ≤ 3‐4 months) 5 ≤ 5yr OR: 0.79 effect of asthma was smaller heterogeneous (I = Possible
(0.63,0.98) and not significant in 16 88%). protection against
9 >5yr OR: 1.05 studies with thorough control allergic rhinitis
(0.99,1.12) for confounders (OR: 0.95; among children <5
95% CI 0.85; 1.06) and in the yr, based on only
13 cohort studies (OR: 0.94; 5 studies; no
95% CI 0.80; 1.11). There were evidence for those
too few studies to estimate older than 5 yr.
Asthma or 29 5‐18 yr OR: 0.91 association with asthma in Inconclusive
wheezing (0.85; 0.98) adults. evidence on the
association
between BF and
the risk of asthma
or wheezing.
Length Randomized 17 ≈6 mo Z scores : +0.03 Most studies are from middle Evidence of No evidence of an Giugliani
trials or quasi‐ (‐0;02; 0.08) income countries. publication bias for effect on BF 201570
Weight experiments 16 Z scores : +0.03 Confounding is unlikely BMI, with small promotion on
comparing (‐0;06; 0.12) because 15 of the 17 studies studies showing weight or length
BMI or children 11 Z scores : ‐0.06 were randomized trials. larger reductions. at 6 months of
weight/length receiving BF (‐0;12; 0.00) Analyses were based on intent age; some

17
 promotion to treat, so that low evidence
interventions compliance with BF promotion supporting a
with control may underestimate the reduction in BMI
children magnitude of the effect. or weight for
length.
Dental caries BF > 12 mo 4 <6 yr OR: 2.69 Most studies not control for Publication biases Consistent Lodge
versus ≤ 12 mo (1.28‐5.64) the introduction of sugary veer toward studies evidence that BF > 201569
BF on demand 6 <6 yr OR: 2.90 foods and drinks. Most with that show 12 mo and BF on
or nocturnal (2.33‐3.60) studies from HICs, where high association between demand have
feeding versus SEP would be expected to these breastfeeding detrimental
not [in breastfed negatively confound the beyond 12 months effects on
children] association. and on demand and deciduous teeth.
dental caries.
Malocclusion Never versus 41 Childhood, OR: 0.32 80% of the studies from Some evidence of Consistent Peres
ever BF; longer adolescence (0.25‐0.40) LMICs. Because malocclusions publication bias but evidence of a 201571
versus shorter and are not associated with SEP the association was major, two‐thirds
duration of adulthood. nor with any other known also present in the reduction in
exclusive BF; or determinant of BF patterns, it larger and better malocclusions in
longer versus is unlikely that these results designed studies. deciduous teeth
shorter duration are affected by confounding. among BF subjects
of any BF
Systolic blood Never versus 43 Childhood, mm Hg: ‐0.80 Three quarters of the studies Evidence of No evidence of a Horta
pressure ever BF; or adolescence (‐1.17; ‐0.43) from LMICs. Evidence of publication bias for reduction in blood 201527
longer versus and residual confounding as effect systolic blood pressure
shorter BF adulthood. was found in studies from HIC pressure studies. associated with BF
duration but not from LMICs.
Diastolic blood Never versus 38 Childhood, mm Hg: ‐0.24
pressure ever BF; or adolescence (‐0.50; 0.02)
longer versus and
shorter BF adulthood.
duration
Overweight Never versus 113 Childhood, OR: 0.74 In HICs, residual confounding Some evidence of Suggestive Horta
and/or obesity ever BF; longer adolescence (0.70; 0.78) by SEP is a possibility; publication bias with evidence of 201527
versus shorter and however, the effect size was larger effects in protection,
duration of adulthood. similar in studies from LMICs small studies, but including high‐
exclusive BF; or (one third of all studies). even large and well‐ quality studies
longer versus Twenty‐three high quality controlled studies and those from
shorter duration studies showed a smaller showed a 20% low or middle
of any BF pooled reduction of 13% (95% reduction in income settings.

18
 CI 6% to 19%). prevalence.
Total Never versus 46 Childhood, /mmol/L: ‐0.01 No evidence of heterogeneity No evidence of an No evidence of an Horta
cholesterol ever BF; or adolescence (‐0.05; 0.02) with nearly all studies showing association. association. 201527
longer versus and small effects. Three quarters
shorter BF adulthood. of the studies from HICs.
duration
Type‐2 diabetes Never versus 11 Childhood, OR: 0.65 Only 2 out of 11 studies from Small number of Limited evidence Horta
27
ever BF; longer adolescence (0.49; 0.86) LMIC showing 14% reduction; available studies; no of protection, 2015
versus shorter and residual confounding may evidence of based on 11
duration of adulthood. affect HIC studies. publication bias. studies.
exclusive BF; or
longer versus
shorter duration
of any BF
Intelligence Never versus 16 Childhood, IQ points: 3.44 In HIC (14 of the 16 studies), Some evidence of Consistent effect Horta
ever BF; or adolescence (2.30; 4.58) residual confounding by SEP is publication bias with of about 3 IQ 201527
longer versus and a possibility; however, the larger effects in points in
shorter BF adulthood. effect was also present in 2 small studies, but observational
duration studies from LMICs. One high even large studies studies; also
quality RCTs showed a showed an effect. present a large
significant increase in IQ of Nine studies with RCT on this topic.
over 7 points. adjustment for
maternal IQ showed
difference of 2.62
points (1.25; 3.98)
EFFECTS ON WOMEN WHO BREASTFED
Lactational Highest versus 13 Women RR : 1.17 Most studies from LMICs. No evidence of Consistent effect Chowdhury
amenorrhea lowest duration (<1 yr post‐ (1.04‐1.32) Residual confounding unlikely. publication bias. on prolonging 201572
of BF partum) Strongest effects when lactational
exclusive or predominant BF amenorrhea,
are compared with partial (RR: especially for
1.21) or no BF (RR: 1.23). exclusive or
predominant BF.
Breast cancer Highest versus 76 Adult OR : 0.81 Three quarters of the studies Some evidence of Consistent Chowdhury
lowest duration women (0.77; 0.86) are from HICs. Parity reduces publication bias but protective effect 201572
of BF the risk of breast cancer and is the association was of BF against
also associated with greater also present in the breast cancer in a
lifetime BF duration. Most larger and better large number of
studies failed to adjust designed studies. well‐designed

19
 appropriately for parity and studies, of a 4.3%
therefore tend to exaggerate reduction per 12
effect size. A thoroughly months of BF in
adjusted pooled analysis of 47 the better
studies shows an OR of 0.96 controlled studies.
41
for each 12 months of BF.
Ovarian cancer Highest versus 41 Adult OR: 0.70 Only 6 studies from LMICs. Some evidence of Suggestive Chowdhury
lowest duration women (0.64; 0.75) Confounding by parity may publication bias, evidence of a 201572
of BF affect the results but with smaller pooled protective effect
socioeconomic confounding is effect sizes in the 22 of BF.
unlikely. Studies with fine studies with
adjustment for parity and samples greater
exclusion of nulliparous than 1,500 women ‐
showed less protection with OR: 0.76 (0.69;
OR: 0.82(0.75‐0.89) 0.84).
Osteoporosis Highest versus 4 Adult SDS: ‐0.132 All studies from HICs. High Not assessed due to Insufficient Chowdhury
72
(distal radius) lowest duration women (‐0.260;‐0.003) heterogeneity in the distal small number of evidence 2015
Osteoporosis of BF 4 SDS: ‐0.142 radius analyses with the studies.
(femoral neck) (‐0.426; 0.142) largest study showing no
association and smaller studies
showing protection; for
femoral neck, none of the
studies found an association.
Type‐2 diabetes Highest versus 6 Adult RR: 0.68 Several confounding factors Small number of Limited evidence Aune
lowest duration women (0.57; 0.82) were adjusted for. Significant available studies; no of protection 201342
of BF protection also found for 3‐ evidence of against type‐2
and 12‐month increases in BF publication bias. diabetes among
duration. Five of the 6 studies women who
are from HICs. All six showed breastfed for
protection. longer periods,
based on 6 studies
Postpartum Qualitative 45 Women Not estimated Studies were highly variable. Not assessed in the The role of Neville
weight change review (<2 yr post‐ because of Most studies found no published review. breastfeeding on 201443
partum) different association. Of the five studies post‐partum
outcome with high methodological weight change is
measures at quality, four reported uncertain.
variable post‐ beneficial effects. Nearly all
partum ages. studies from HICs.
558

20
559

21
560 Notes:
561 In several reviews, the summary effect sizes represent the pooled results of studies comparing longer versus shorter BF durations (either never versus ever BF;
562 exclusive BF for more than x months versus less than x months; or any BF for more than x months versus less than x months). Separate results for each type of
563 categorization are available in the web annexes.
564 Acronyms ‐ HIC: high‐income countries; LMIC: low and middle‐income countries; SEP: socioeconomic position; OR: odds ratio; RR: risk ratio; RCT: randomized
565 controlled trial; SDS: standard deviation scores.
566
567
568

22
569 F
Figures
570
571 FFigure 1. Percen
nt of children who
w receive any breastmilk at 12
1 months of agge, by country (data from 153 countries, 1995
5‐2013).
572 CCountries withoout data are shoown as terrain view.
v

573

23
574 Figure 2
2. BF indicattors by cou
untry incomme group in 2010 (analyyses of natiional surveyys using
575 standarrd indicatorrs, weightedd by national populations of child
dren under two years. Data
576 on up to
o 153 counttries.
577

578
579

Lancet B
Breastfeedingg Series 2015
5 – Paper 1 Paage 24
580 Figure 3
3. National‐‐level propoortions of children age
ed 12‐15 mo onths who are breastffed,
581 accordinng to exclussive BF in children age
ed 0‐5 montths. Countrries are colo
or‐shaded
582 accordinng to the re
egions of thhe world (UNICEF classsification). Based
B on th
he most reccent
583 survey ffrom 117 coountries (20 000‐2013).
584

585

Lancet B
Breastfeedingg Series 2015
5 – Paper 1 Paage 25
586 Figure 44. National and wealth
h quintile‐sppecific timee trends in exclusive
e BF 0‐5 month hs (solid
587 lines) an
nd continueed BF 12‐155 months (ddotted liness) , 1993‐20013, weighte ed by natioonal
588 populattions of chilldren underr two yearss at the timee of the surrvey. Analyses restricted to 66
589 countriees with info
ormation onn household wealth.
590

591
592

Lancet B
Breastfeedingg Series 2015
5 – Paper 1 Paage 26
593 REFERENCES
594
595 1. Vahlquist B. Introduction. Contemporary patterns of breast‐feeding Report of the WHO
596 Collaborative Study on Breast‐feeding. Geneva: World Health Organization; 1981.
597 2. Jelliffe DB, Jelliffe EFP. Human milk in the modern world. Oxford: Oxford University Press;
598 1978.
599 3. Jelliffe DB, Jelliffe EFP. Human milk in the modern world (letter). British Medical Journal
600 1978; 2(6151): 1573.
601 4. Task Force on the Assessment of the Scientific Evidence Rrelating to Infant‐Feeding Practices
602 and Infant Health. Report of the Task Force on the Assessment of the Scientific Evidence Relating to
603 Infant‐Feeding Practices and Infant Health. Pediatrics 1984; 74(4 Pt 2): 579‐762.
604 5. Rollins NC, et al. Lancet Breastfeeding Series: Why invest, and what it will take to improve
605 breastfeeding practices in less than a generation. Lancet 2014; (submitted).
606 6. World Health Organization. Indicators for assessing infant and young child feeding practices.
607 Part I: definition. Geneva: World Health Organization.; 2008.
608 7. Royston P, Sauerbrei W. Multivariable Model‐Building: A Pragmatic Approach to Regression
609 Analysis based on Fractional Polynomials for Modelling Continuous Variables. Chichester, UK.: Wiley;
610 2008.
611 8. Walker N, Tam Y, Friberg IK. Overview of the Lives Saved Tool (LiST). BMC public health 2013;
612 13 Suppl 3: S1.
613 9. Requejo JH, Bryce J, Barros AJ, et al. Countdown to 2015 and beyond: fulfilling the health
614 agenda for women and children. Lancet 2014.
615 10. Roberts TJ, Carnahan E, Gakidou E. Can breastfeeding promote child health equity? A
616 comprehensive analysis of breastfeeding patterns across the developing world and what we can
617 learn from them. BMC medicine 2013; 11: 254.
618 11. Berger‐Achituv S, Shohat T, Garty BZ. Breast‐feeding patterns in Central Israel. The Israel
619 Medical Association journal : IMAJ 2005; 7(8): 515‐9.
620 12. Haggkvist AP, Brantsaeter AL, Grjibovski AM, Helsing E, Meltzer HM, Haugen M. Prevalence
621 of breast‐feeding in the Norwegian Mother and Child Cohort Study and health service‐related
622 correlates of cessation of full breast‐feeding. Public Health Nutr 2010; 13(12): 2076‐86.
623 13. Jones JR, Kogan MD, Singh GK, Dee DL, Grummer‐Strawn LM. Factors associated with
624 exclusive breastfeeding in the United States. Pediatrics 2011; 128(6): 1117‐25.
625 14. Kohlhuber M, Rebhan B, Schwegler U, Koletzko B, Fromme H. Breastfeeding rates and
626 duration in Germany: a Bavarian cohort study. The British journal of nutrition 2008; 99(5): 1127‐32.
627 15. Labbok M, Taylor E. Achieving Exclusive Breastfeeding in the United States: Findings and
628 Recommendations. Washington, DC: United States Breastfeeding Committee; 2008.
629 16. Ministere des Solidarites de la Sante et de la Famille, Institut National de la Sante, Institut
630 National de la Sante et la Recherche Medicale. Enquete Nationale Perinatale 2003. Paris: MSSF; 2005.
631 17. Oakley LL, Henderson J, Redshaw M, Quigley MA. The role of support and other factors in
632 early breastfeeding cessation: an analysis of data from a maternity survey in England. BMC pregnancy
633 and childbirth 2014; 14: 88.
634 18. Al‐Sahab B, Lanes A, Feldman M, Tamim H. Prevalence and predictors of 6‐month exclusive
635 breastfeeding among Canadian women: a national survey. BMC pediatrics 2010; 10: 20.
636 19. Almquist‐Tangen G, Stromberg U, Holmen A, et al. Influence of neighbourhood purchasing
637 power on breastfeeding at four months of age: a Swedish population‐based cohort study. BMC public
638 health 2013; 13: 1077.
639 20. WHO Collaborative Study Team on the Role of Breastfeeding on the Prevention of Infant
640 Mortality. Effect of breastfeeding on infant and child mortality due to infectious diseases in less
641 developed coutries: a pooled analysis. Lancet 2000; 355: 451‐55.

Lancet Breastfeeding Series 2015 – Paper 1 Page 27

642 21. Ip S, Chung M, Raman G, et al. Breastfeeding and Maternal and Infant Health Outcomes in
643 Developed Countries. . Rockville, USA: Agency for Healthcare Research and Quality, 2007.
644 22. Holman RC, Stoll BJ, Curns AT, Yorita KL, Steiner CA, Schonberger LB. Necrotising enterocolitis
645 hospitalisations among neonates in the United States. Paediatr Perinat Epidemiol 2006; 20(6): 498‐
646 506.
647 23. Horta BL, Victora CG. Short‐term effects of breastfeeding: a systematic review of the benefits
648 of breastfeeding on diarhoea and pneumonia mortality. Geneva: World Health Organization, 2013.
649 24. Onyeaso CO, Isiekwe MC. Occlusal changes from primary to mixed dentitions in Nigerian
650 children. The Angle orthodontist 2008; 78(1): 64‐9.
651 25. Peres KG, Peres MA, Hallal PC, Menezes AMB, Thomson WM, Broadbent J. Deciduous‐
652 dentition malocclusion predicts orthodontic treatment needs later: findings from a population‐based
653 birth cohort study. Am J Orthod Dentofacial Orthop 2015; 147: 492‐98.
654 26. Horta BL, Victora CG. Long‐term effects of breastfeeding: a systematic review. Geneva: World
655 Health Organization, 2013.
656 27. Horta et al. Systematic review of effects of breastfeeding on intelligence, blood pressure,
657 overweight/obesity, cholesterol, and diabetes. Acta Paediatrica (submitted).
658 28. Kramer MS, Chalmers B, Hodnett ED, et al. Promotion of Breastfeeding Intervention Trial
659 (PROBIT): a randomized trial in the Republic of Belarus. Jama 2001; 285(4): 413‐20.
660 29. Kramer MS, Matush L, Vanilovich I, et al. Effects of prolonged and exclusive breastfeeding on
661 child height, weight, adiposity, and blood pressure at age 6.5 y: evidence from a large randomized
662 trial. The American journal of clinical nutrition 2007; 86(6): 1717‐21.
663 30. Owen CG, Martin RM, Whincup PH, Davey‐Smith G, Gillman MW, Cook DG. The effect of
664 breastfeeding on mean body mass index throughout life: a quantitative review of published and
665 unpublished observational evidence. The American journal of clinical nutrition 2005; 82(6): 1298‐307.
666 31. Kramer MS, Aboud F, Mironova E, et al. Breastfeeding and child cognitive development: new
667 evidence from a large randomized trial. Archives of general psychiatry 2008; 65(5): 578‐84.
668 32. Lucas A, Morley R, Cole TJ, Lister G, Leeson‐Payne C. Breast milk and subsequent intelligence
669 quotient in children born preterm. Lancet 1992; 339(8788): 261‐4.
670 33. Richards M, Hardy R, Wadsworth ME. Long‐term effects of breast‐feeding in a national birth
671 cohort: educational attainment and midlife cognitive function. Public Health Nutr 2002; 5(5): 631‐5.
672 34. Martin RM, Goodall SH, Gunnell D, Davey Smith G. Breast feeding in infancy and social
673 mobility: 60‐year follow‐up of the Boyd Orr cohort. Archives of disease in childhood 2007; 92(4): 317‐
674 21.
675 35. Horwood LJ, Fergusson DM. Breastfeeding and later cognitive and academic outcomes.
676 Pediatrics 1998; 101(1): E9.
677 36. Victora CG, Barros FC, Horta BL, Lima RC. Breastfeeding and school achievement in Brazilian
678 adolescents. Acta Paediatr 2005; 94(11): 1656‐60.
679 37. Horta BL, Bas A, Bhargava SK, et al. Infant feeding and school attainment in five cohorts from
680 low‐ and middle‐income countries. PloS one 2013; 8(8): e71548.
681 38. Victora CG, Horta BL, de Mola CL, et al. Association between breastfeeding and intelligence,
682 educational attainment, and income at 30 years of age: a prospective birth cohort study from Brazil.
683 The Lancet Global Health; 3(4): e199‐e205.
684 39. Becker S, Rutstein S, MH. L. Estimation of births averted due to breast‐feeding and increases
685 in levels of contraception needed to substitute for breast‐feeding. J Biosoc Sci 2003; 35(4): 559‐74.
686 40. Kramer MS, Kakuma R. Optimal duration of exclusive breastfeeding. The Cochrane database
687 of systematic reviews 2012; 8: Cd003517.
688 41. Breast cancer and breastfeeding: collaborative reanalysis of individual data from 47
689 epidemiological studies in 30 countries, including 50302 women with breast cancer and 96973
690 women without the disease. Lancet 2002; 360(9328): 187‐95.

Lancet Breastfeeding Series 2015 – Paper 1 Page 28

691 42. Aune D, Norat T, Romundstad P, Vatten LJ. Breastfeeding and the maternal risk of type 2
692 diabetes: a systematic review and dose‐response meta‐analysis of cohort studies. Nutrition,
693 metabolism, and cardiovascular diseases : NMCD 2014; 24(2): 107‐15.
694 43. Neville CE, McKinley MC, Holmes VA, Spence D, Woodside JV. The relationship between
695 breastfeeding and postpartum weight change‐‐a systematic review and critical evaluation.
696 International journal of obesity (2005) 2014; 38(4): 577‐90.
697 44. Bobrow KL, Quigley MA, Green J, Reeves GK, Beral V. Persistent effects of women's parity
698 and breastfeeding patterns on their body mass index: results from the Million Women Study.
699 International journal of obesity (2005) 2013; 37(5): 712‐7.
700 45. Dias CC, Figueiredo B. Breastfeeding and depression: a systematic review of the literature.
701 Journal of affective disorders 2015; 171: 142‐54.
702 46. World Health Organization. Protecting, promoting and supporting breast‐feeding: the special
703 role of maternity services. A Joint WHO/UNICEF Statement. . Geneva: World Health Organization;
704 1989.
705 47. World Health Organization. Essential Nutrition Actions: Improving Maternal, Newborn, Infant
706 and Young Child Health and Nutrition. Geneva: World Health Organization, 2013.
707 48. Greenland S. Can meta‐analysis be salvaged? Am J Epidemiol 1994; 140(9): 783‐7.
708 49. Bailar JC, 3rd. The promise and problems of meta‐analysis. N Engl J Med 1997; 337(8): 559‐
709 61.
710 50. Black RE, Victora CG, Walker SP, et al. Maternal and child undernutrition and overweight in
711 low‐income and middle‐income countries. Lancet 2013; 382(9890): 427‐51.
712 51. Lim SS, Vos T, Flaxman AD, et al. A comparative risk assessment of burden of disease and
713 injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990‐2010: a systematic
714 analysis for the Global Burden of Disease Study 2010. Lancet 2012; 380(9859): 2224‐60.
715 52. Requejo J, Bryce J, Victora CG, and the Countdown to 2015 writing team. Countdown to 2015
716 and beyond: fulfilling the health agenda for women and children. New York: UNICEF and the World
717 Health Organizaiton, 2014.
718 53. Dettwyler KA. A time to wean: the hominid blueprint for the natural age of weaning in
719 modern human populations. . In: Macadam PS, Dettwyller KA, eds. Breastfeeding: biocultural
720 perspectives. New York: Walter de Gruyter, Inc.; 1995: 39‐73.
721 54. Gura T. Nature's first functional food. Science 2014; 345(6198): 747‐9.
722 55. Hooper LV, Littman DR, Macpherson AJ. Interactions between the microbiota and the
723 immune system. Science 2012; 336(6086): 1268‐73.
724 56. Mayer EA, Knight R, Mazmanian SK, Cryan JF, Tillisch K. Gut microbes and the brain: paradigm
725 shift in neuroscience. The Journal of neuroscience : the official journal of the Society for Neuroscience
726 2014; 34(46): 15490‐6.
727 57. Dominguez‐Bello MG, Costello EK, Contreras M, et al. Delivery mode shapes the acquisition
728 and structure of the initial microbiota across multiple body habitats in newborns. Proceedings of the
729 National Academy of Sciences of the United States of America 2010; 107(26): 11971‐5.
730 58. Azad MB, Konya T, Maughan H, et al. Gut microbiota of healthy Canadian infants: profiles by
731 mode of delivery and infant diet at 4 months. CMAJ : Canadian Medical Association journal = journal
732 de l'Association medicale canadienne 2013; 185(5): 385‐94.
733 59. Yatsunenko T, Rey FE, Manary MJ, et al. Human gut microbiome viewed across age and
734 geography. Nature 2012; 486(7402): 222‐7.
735 60. De Leoz ML, Kalanetra KM, Bokulich NA, et al. Human Milk Glycomics and Gut Microbial
736 Genomics in Infant Feces Show a Correlation between Human Milk Oligosaccharides and Gut
737 Microbiota: A Proof‐of‐Concept Study. Journal of proteome research 2014.
738 61. Bourges D, Meurens F, Berri M, et al. New insights into the dual recruitment of IgA+ B cells in
739 the developing mammary gland. Molecular immunology 2008; 45(12): 3354‐62.
740 62. Latuga MS, Stuebe A, Seed PC. A review of the source and function of microbiota in breast
741 milk. Seminars in reproductive medicine 2014; 32(1): 68‐73.
Lancet Breastfeeding Series 2015 – Paper 1 Page 29
742 63. Ardeshir A, Narayan NR, Mendez‐Lagares G, et al. Breast‐fed and bottle‐fed infant rhesus
743 macaques develop distinct gut microbiotas and immune systems. Science translational medicine
744 2014; 6(252): 252ra120.
745 64. Verduci E, Banderali G, Barberi S, et al. Epigenetic effects of human breast milk. Nutrients
746 2014; 6(4): 1711‐24.
747 65. Verier C, Meirhaeghe A, Bokor S, et al. Breast‐feeding modulates the influence of the
748 peroxisome proliferator‐activated receptor‐gamma (PPARG2) Pro12Ala polymorphism on adiposity in
749 adolescents: The Healthy Lifestyle in Europe by Nutrition in Adolescence (HELENA) cross‐sectional
750 study. Diabetes care 2010; 33(1): 190‐6.
751 66. Torregrosa Paredes P, Gutzeit C, Johansson S, et al. Differences in exosome populations in
752 human breast milk in relation to allergic sensitization and lifestyle. Allergy 2014; 69(4): 463‐71.
753 67. Hassiotou F, Hartmann PE. At the dawn of a new discovery: the potential of breast milk stem
754 cells. Advances in nutrition 2014; 5(6): 770‐8.
755 68. Sankar J et al. Systematic review of breastfeeding and child mortality. Acta Paediatrica
756 (submitted).
757 69. Lodge et al. Systematic review of breastfeeding and allergic disease, otitis media and dental
758 caries. Acta Paediatrica (submitted).
759 70. Giugliani et al. Systematic review of breastfeeding and infant nutrition. Acta Paediatrica
760 (submitted).
761 71. Peres et al. Systematic review of breastfeeding and malocclusion. Acta Paediatrica
762 (submitted).
763 72. Chowdhury et al. Systematic review on the effects of breastfeeding on maternal health. Acta
764 Paediatrica (submitted).
765

Lancet Breastfeeding Series 2015 – Paper 1 Page 30