Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Infection and disease: The impact of early weaning

Cesar G. Victora



Although the protective effect of breastfeeding against infectious diseases has been long suspected, it has only recently been adequately quantified This article reviews the available epidemiologic evidence. Breastfeeding provides marked protection against diarrhoeal morbidity and severity (as measured by its duration, level of dehydration, or case-fatality rate), and mortality. This is true for acute watery diarrhoea as well as for dysentery and persistent diarrhoea. A dose-response pattern is observed: the risk level for children receiving both breastmilk and artificial milk is between that of exclusively breasfed children and that of completely weaned children. Even the introduction of water or herbal teas to a previously exclusively breasfed infant increases the risk of morbidity and mortality. The protection is greatest for young infants, and the period immediately after weaning seems to carry the highest risk. Although breastfeeding does not seem to have an appreciable effect on the incidence of acute respiratory infections as a whole, there is evidence of an effect on the incidence of pneumonia, on the incidence of hospitalization for pneumonia, and on mortality from respiratory infections. The level of protection against pneumonia, however, is lower than that against diarrhoea. Breastfeeding also protects against onus media and other infections, including neonatal sepsis, meningitis, and bacteraemia, although the number of relevant studies is small. A simulation exercise shows that a 40% reduction in the prevalence of nonbreastfeeding would prevent up to 15% of diarrhoea deaths and 7% of pneumonia deaths occurring in regions with a short breastfeeding duration, such as urban Latin America. In regions where breastfeeding durations are longer, the emphasis should be on maintaining these high rates and increasing the proportion of young infants who are exclusively breastfed. The present review confirms that breastfeeding promotion is an essential child survival strategy.



The protective effect of breastfeeding against infectious diseases has long been known. Several studies from Europe in the eighteenth and nineteenth centuries indicated that nonbreastfed infants were two to three times more likely to die than those who were breastfed [1, 2]. In the last 10 to 20 years, however, there has been a resurgence of interest in breastfeeding, and a large number of new studies have been produced.

Recent studies have shown improvements in sampling and in the operational diagnoses of different infectious diseases, as well as in data processing and analysis. In addition, a number of recent publications have addressed key methodological issues that may distort the findings of studies of breastfeeding and disease [3-6]. These issues include self-selection, reverse causality, and confounding.

Self-selection may occur when infants are weaned because they were not thriving on breastmilk. Those who continue to be breastfed, therefore, may represent a subgroup of particularly healthy infants. Reverse causality takes place when an infection leads to weaning, rather than the opposite. Confounding results when early weaning is associated with demographic, socio-economic, or environmental characteristics that may also affect morbidity. Such confounding factors may lead to a spurious association between breastfeeding and protection from infection. Some recent studies have attempted to avoid these pitfalls through specific improvements in their design and analysis [5]. Unless otherwise noted, the studies reviewed here were adjusted for confounding variables. Several investigators have also attempted to account for reverse causality and self selection.

Virtually all studies of breastfeeding and infection have observational designs, such as case-control, cohort, or crosssectional studies. Intervention trials would solve many of the methodological problems inherent to these designs, but the random allocation of mothers and infants to a breastfeeding or a control group is not ethically justified. In addition, ensuring compliance with a recommended feeding practice is extremely difficult, and the sample sizes required for studies with morbidity outcomes, in the presence of low compliance, would be enormous.

Another problem is that studies have addressed different types of disease outcomes. Some investigators have studied overall morbidity from, for example, diarrhoea or respiratory infections. Others have concentrated on more severe diseases, such as persistent or dehydrating diarrhoea or pneumonia. Finally, some studies have assessed mortality. The risk factors for acquiring an infectious disease may be quite different from the prognostic factors leading to a severe episode and finally to death. Therefore, an apparent lack of consistency between different studies may be due to the fact that different disease outcomes were being investigated.

Other participants in this Workshop have addressed the mechanisms through which breastfeeding protects against infections, as well as the epidemiologic literature from developed countries. I will concentrate on the epidemiologic evidence supporting this protective role, based on studies from less-developed countries. It is important to separate studies from developed and less-developed areas, because the effects of breastfeeding are modified by a number of socio-economic, environmental, and dietary factors. For example, greater protection is likely to occur where there is poverty, crowding, or unsafe drinking water or where weaning foods are contaminated or of low energy density. Studies from developed countries, therefore, cannot be extrapolated to the rest of the world.

This review makes no claim to be exhaustive, as a number of comprehensive summaries of the literature on this topic are available [7-12]. Rather, I will use examples from a few investigations-with a strong bias towards studies in which I was involved -to highlight the key epidemiologic and public health issues.


Impact on diarrhoea

Feachem and Koblinsky [8] reviewed 35 studies on diarrhoeal incidence from 14 countries. There was a clear protection associated with breastfeeding, but this varied according to the child's age. The relative risks for nonbreastfed infants compared with all those who were breastfed, either exclusively or par tially, were 3.0 up to two months of age, 2.4 at three to five months, and about 1.4 thereafter. According to this review, protection after infancy was not evident.

FIG. 1. Risk of dehydration (odds ratio) from diarrhoea in infants in Brazil in 1992 according to type of milk

Epidemiologic studies show that breastfeeding protects against shigellosis and cholera. Breastmilk also contains antibodies against a large number of other gastrointestinal pathogens [13]. It also protects against neonatal necrotizing enterocolitis [14].

Breastfeeding appears to provide even higher protection against more severe forms of diarrhoea. In a study we carried out in Brazil, dehydrating diarrhoea (defined by the presence of a persistent skinfold plus at least two other signs of dehydration) was 6.0 times more frequent among non-breastfed infants than among exclusively breastfed infants (fig. 1) [15]. If a child already had diarrhoea, the risk of developing dehydration was 3.3 times greater for the nonbreastfed infant [16].

Another indication of severity is the duration of the diarrhoeal episode. When 400 Brazilian infants were followed from birth to the age of six months, those weaned during the first week of life had a risk of diarrhoea three times greater than those who were exclusively breastfed (including those also receiving water or herbal teas). When the analysis was restricted to episodes of persistent diarrhoea lasting 14 days or more, the relative risk was 5.1.

Yet another indicator of severity is case fatality. In Rwandan children hospitalized with diarrhoea, the casefatality rate was three to four times lower for those who were breastfed [17l. Breastfeeding, therefore, seems to provide even greater protection against severe diarrhoea than against mild diarrhoea.

The evidence from mortality is also compelling. Several studies from the first half of this century showed a protective effect of breastfeeding. For example, Newsholme [18] compared infants who died of "epidemic diarrhoea" in Brighton between 1903 and 1905 with a population-based control group. Reanalysis of his data with logistic regression (fig. 2) shows that after adjustment for age, infants receiving condensed milk were 91 times more likely to die than those on breastmilk. The odds ratio for infants on cow's milk was 40. These relative risks are probably inflated by reverse causality and other biases.

FIG. 2. Risk of death (odds ratio) from diarrhoea in infants in England in 1903-05 according to feeding pattern

FIG. 3. Risk of death (odds ratio) from diarrhoea in infants in Brazil in 1984-85 according to feeding pattern

FIG. 4. Risk of dehydration (odds ratio for incidence) from diarrhoea in infants in Brazil in 1992 according to time since weaning

FIG. 5. Prevalence of diarrhoea (relative risk for percentage of days ill) in infants under two months of age in Peru in 1988 according to feeding pattern. Arrow shows doubling of prevalence when infant receives liquids in addition to breastmilk

Our study from Brazil showed odds ratios for diarrhoeal mortality of 14.2 for totally weaned infants and 4.2 for those on both breastmilk and artificial milk, compared with those on breastmilk alone [19].

Figure 3 shows the results of this study for several feeding patterns. Children receiving only artificial milk were at a particularly high risk. The addition of any supplements, either artificial milk or food, resulted in at least a two- to threefold increase in the risk for breastfed infants.

In this study, deaths were broken down according to the clinical characteristics of the diarrhoeal episode. Nonbreastfed infants were 21 times more likely than exclusively breastfed infants to die of acute watery diarrhoea, 10 times more likely to die of persistent diarrhoea, and 3 times more likely to die of dysentery [20].

The greater vulnerability of young infants was confirmed by this study, because the protection was even higher before two months of age (odds ratio of 23 for the nonbreastfed infant under two months). A related issue is the especially high vulnerability of infants who were recently weaned. Figure 4 shows additional results from the abovementioned study of dehydrating diarrhoea [15, 16]. The risk was highest for infants who had recently been weaned and decreased thereafter.

A major issue raised in the last 10 years is the importance of exclusive breastfeeding for the prevention of diarrhoea. Figure 5 shows the results of a Peruvian longitudinal study in which the percentage of days on which a child had diarrhoea was related to feeding patterns [21]. There was a doubling in the prevalence when the child received liquids such as water or herbal teas in addition to breastmilk. The diarrhoea mortality study from Brazil also showed that each additional fluid feeding resulted in a 40% increase in the risk of death from diarrhoea among breastfed infants [19, 22]. More recently, a study from the Philippines showed that giving fluids to breastfed infants two months of age resulted in a two- to threefold increase in the incidence of diarrhoea [23]. Both the Peruvian and the Filipino studies showed that the benefits of breastfeeding were considerably less after the first six months.


Impact on respiratory Infections

Fewer studies have addressed the effect of breastfeeding on respiratory infections. This may be partly because of the difficulties in operationalizing and standardizing the diagnoses of these infections under field conditions.

A recent review of studies from less-developed countries showed that although breastfeeding did not seem to provide important protection against acute respiratory infections as a whole, there was evidence of an effect on the incidence of pneumonia and of hospitalizations for pneumonia, as well as on mortality from respiratory infections [24]. For example, figure 6 shows the results of two recent case-control studies from two Brazilian cities in which the outcome was radiologically confirmed pneumonia [25, 26]. The risks for partially breastfed and non-breastfed children were clearly greater than those for exclusively breastfed children.

FIG. 6. Risk of pneumonia (odds ratio for incidence) in children under two years of age in two Brazilian sites, Porto Alegre and Fortaleza, in 1993-95 according to type of milk

FIG. 7. Risk of otitis media (odds ratio for incidence in last month) in infants six months of age in Brazil in 1993-94 according to duration of breastfeeding

FIG. 8. Incidence of neonatal infections in a randomized trial of human milk (HM) versus formula in India in 1984 [30]

Breastfeeding also affects the case-fatality rate of respiratory infections. In Rwanda non-breastfed children were twice as likely as breastfed children to die of pneumonia [17]. Non-breastfed Brazilian infants were 3.6 times more likely to die of acute lower respiratory tract infections than those who received breastmilk and no artificial milk. Infants receiving both human and artificial milk had an intermediate odds ratio of 1.6 [19, 27].

An interesting but somewhat unexpected finding from two of the Brazilian studies deserves further investigation. Regardless of the type of milk consumed, infants who received non-milk food complements were 2.3 times less likely to develop pneumonia [25] and 3.2 times less likely to die [27].

The protection provided by breastfeeding against otitis media is well established, mostly through studies from developed countries [6, 9]. Figure 7 shows the results of a recent study of six-month-old Brazilian infants [28]. There is a clear gradient in the risk of medically confirmed otitis media in the preceding month, according to the duration of breastfeeding. It is interesting to compare these findings with those shown in figure 4, where the most recently weaned babies showed the highest risk of dehydrating diarrhoea. This contrast, along with the finding that the protection against otitis media is also well documented in developed countries, suggests that the mechanism may involve the anatomical development of the upper respiratory system rather than any environmental risk factors.


Impact on other infections

Breastfeeding is known to protect against other infections, including neonatal sepsis, meningitis, and bacteraemia [11, 13]. Despite one study that failed to find protection against mortality from measles [29], the Rwandan study showed that case fatality from measles was reduced by one-third for breastfed infants [17].

The only randomized intervention trial examining the relation between breastmilk and infection that I am aware of was carried out in India. This study compared the effects of different milk combinations given by bottle to high-risk newborns [30]. Septicaemia, conjunctivitis, diarrhoea, and umbilical sepsis were the most common infections observed. Figure 8 shows that there were separate effects of pasteurization and of formula introduction on neonatal infection rates.

The Brazilian case-control study of mortality [19] included deaths from meningitis, skin infections, measles, whooping cough, neonatal sepsis, and tuberculosis. Because of the small number of cases in each category, these were grouped together. Compared with infants receiving breastmilk without any artificial milk, non-breastfed children were 2.5 times more likely to die of these infections. Infants receiving both breastmilk and artificial milk were at lower risk (odds ratio, 0.4), but this difference was not statistically significant.


Summary of the evidence

This review began with a discussion of methodological problems affecting observational studies and argued that experimental, community-based studies would in general not be feasible or ethical. The Indian trial described above was an exception, since it was a hospital-based study for highrisk neonates who were all bottle-fed [30]. In some circumstances, however, one may carry out quasiexperimental studies in the community advantage of existing interventions. In the Brazilian city of Guarujá, a lactation centre has been well established for several years [31]. Mothers giving birth in the city hospitals are invited to join the centre, and about half of them eventually attend at least once. This is obviously not a random process, because mothers who attend may differ from those who do not, but in this study it was possible to interview the mothers in the hospital before they were invited to join the centre and to collect a large amount of baseline data. There were some differences between attenders and non-attenders in social and demographic variables. There were no differences in reported maternal attitudes towards breastfeeding, but attenders were more likely to have received breastfeeding advice during pregnancy. These confounding factors were handled later in the data analysis.

All mothers were followed up until the child was six months old. At four months, 42% of attenders were exclusively breastfeeding, against 12% of nonattenders. At six months of age, after adjusting for the above-mentioned confounding variables, the incidence of diarrhoea in the previous two weeks among non-attenders was 1.7 times that among attenders, a significant difference (fig. 9). Nonattenders also had almost twice as many hospital admissions as attenders for diarrhoea since birth, but this difference was not significant (p = .09). This study brings us closer to an experimental design. Because the comparison is based on attendance and not on successful breastfeeding, reverse causality bias is avoided, as well as some of the self-selection bias (in fact, mothers facing difficulties may be more, rather than less, likely to attend). Self-selection and confounding are further reduced by obtaining prior information about attitudes and practices regarding breastfeeding at the time of delivery, and by adjusting in the analyses for any differences. It is reassuring that the protection provided by breastfeeding against diarrhoea is confirmed in a stricter study design.

The main findings of this review may be summarized as follows. Breastfeeding reduces the incidence of diarrhoea. In addition, for children who have already developed diarrhoea, breastfeeding reduces the severity of the episode, whether measured by duration of the episode, risk of dehydration, or case-fatality rate. As a consequence, breastfeeding strongly protects against diarrhoeal mortality. There is a clear effect of age, with the greatest protection among young infants. There is also evidence that recently weaned infants may be more vulnerable. Although partial breastfeeding results in a reasonable degree of protection against diarrhoea, exclusive breastfeeding for young infants is essential to achieve the highest protection. This is because the mere addition of water or herbal teas to the diet sharply increases the incidence of diarrhoea and mortality due to diarrhoea.

The protection afforded by breastfeeding against respiratory infections is not well established. However, studies from less-developed countries clearly show a protective effect on severe respiratory infections, as measured by the incidence of pneumonia and by hospitalizations and mortality due to respiratory infections. There is also consistent evidence of protection against otitis media. A dose-response trend is clear in several studies, with partially breastfed children having intermediate levels of risk. Nevertheless, the protection is several times smaller than that provided by breastmilk against diarrhoea. This finding is consistent with biological knowledge, because in addition to its systemic effects, breastmilk has a direct effect on the gastrointestinal immune system and flora, as well as leading to the avoidance of possibly contaminated weaning foods.

Evidence of the impact on infections other than diarrhoea and respiratory infections is not as clear-cut, but several studies suggest some level of protection.

On the basis of simulations we recently performed for the World Health Organization [24], our group has calculated the likely impact of breastfeeding promotion on mortality from diarrhoea and respiratory infections, the two major killers of children in the world. Data on three typical populations that differed in the duration of breastfeeding were taken from the WHO Collaborative Study on Breastfeeding (table 1) [32]. Based on the literature, it was estimated that full or partial breastfeeding would lead to a 50% reduction in deaths from respiratory tract infections and a 66% reduction in deaths from diarrhoea among children aged under 18 months. It was further assumed that it would be possible to reduce the prevalence of non-breastfeeding by 40%, a reasonable figure based on results from intervention studies. Table 2 shows the proportions of deaths among infants under five years of age that would be prevented through this intervention.

FIG. 9. Risk of diarrhoea (odds ratio for diarrhoea in the last two weeks) and hospital admission in infants six months of age in Brazil in 1993 in the control and intervention (lactation centre) groups [31]

Breastfeeding promotion would obviously have very limited impact in areas where most children are already breastfed, such as rural Africa, South Asia, and Central America. It would have a substantial impact, however, in areas where early weaning is common. These include many urban and rural areas in South America and South-East Asia as well as some of the largest African cities. These calculations are conservative, because the benefits of partial breastfeeding are not considered, but they serve to illustrate the potential life-saving effect of breastfeeding promotion.

The Bible says, "Blessed is the womb that bore you and the breast that you sucked" (Luke 11:27). Recent research is confirming how correct this statement is.



I would like to thank my colleagues Sandra Fuchs and Fernando Barros for their contribution to the Brazilian studies reviewed here.

TABLE 1. Percentage of children not breastfed according to age in populations with different durations of breastfeeding

Age (mo)

Duration of breastfeeding
















Source: ref.32

TABLE 2. Expected percentage reduction in mortality among children under five years of age resulting from a 40% decrease in the prevalence of non-breastfeeding in populations with different durations of breastfeeding

Cause of death Duration of breastfeeding









0.5 %

3.3 %

7.0 %



1. Wickes IG. A history of infant feeding. Arch Dis Child 1953;4:233-40, 416-22.

2. Knodel J. Van der Valle E. Breast feeding, fertility and infant mortality. An analysis of early German data. Pop Stud 1967;21:109-31.

3. Habicht JP, DaVanzo J. Butz WP. Does breastfeeding really save lives, or are potential benefits due to biases? Am J Epidemiol 1986;123:279-90.

4. Sauls HS. Potential effect of demographic and other variables in studies comparing morbidity of breast-fed and bottle-fed infants. Pediatrics 1979;64:523-7.

5. Victora CG. Case-control studies of the influence of breastfeeding on child morbidity and mortality: methodological issues. In: Atkinson SA, Hanson LA, Chandra RK, eds. Breastfeeding, nutrition, infection and infant growth in developed and developing countries. St. John's, Canada: ARTS Biomedical Publishers and Distributors, 1990:405-18.

6. Bauchner H. Leventhal JM, Shapiro ED. Studies of breast-feeding and infection. How good is the evidence? JAMA 1986;256:887-92.

7. Jelliffe DB, Jelliffe EFP. Human milk in the modern world. Oxford: Oxford University Press, 1978.

8. Feachem RG, Koblinski MA. Interventions for the control of diarrhoeal diseases among young children: promotion of breastfeeding. Bull WHO 1984;62:271-91.

9. Kovar MG, Serdula MK, Marks JS, Fraser DW. Review of the epidemiologic evidence for an association between infant feeding and infant health. Pediatrics 1984;74:615-38.

10. Jason JM, Nieburg P. Marks JS. Mortality and infectious disease associated with infant-feeding practices in developing countries. Pediatrics 1984;74:702-27.

11. Cunningham AS, Jelliffe DB, Jelliffe EFP. Breastfeeding and health in the 1980s. A global epidemiological review. J Pediatr 1991;118:659-66.

12 Victora CG. Breastfeeding, morbidity and mortality. In: Chandra RK, ed. Proceedings of the Conference on Nutrition and Immunology. St. John's, Canada: ARTS Biomedical Publishers and Distributors, 1992: 63-72.

13. Hanson LA, Ashraf R. Carlsson B. Mattsby-Baltzer I, Motas C, Hahn-Zoric M, Mata L, Herias V, Cruz JR, Lindblad BS, Karlberg J, Jalil F. The Second John Soothill Lecture: breastfeeding, infections and immunology. In: Chandra RK, ed. Nutrition and immunology. St. John's, Canada: ARTS Biomedical Publishers and Distributors, 1992:45-60.

14. Lucal A, Cole TJ. Breast milk and neonatal necrotizing enterocolitis. Lancet 1990;336:1519-23.

15. Fuchs SC. Risk factors for dehydrating diarrhea: a case-control study. Ph.D. thesis, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, 1993.

16. Victora CG, Fuchs SC, Kirkwood BR, Lombardi C, Barros FC. Breastfeeding, nutritional status and other prognostic factors for dehydration among young children with diarrhoea: a casecontrol study. Bull WHO 1992;70:467-75.

17. Lepage P. Munyakazi C, Hennart P. Breast feeding and hospital mortality in children from Rwanda. Lancet 1981;i:40911.

18. Newsholme A. Domestic infection in relation to epidemic diarrhoea. J Hyg 1906;6:139-48.

19. Victora CG, Smith PG, Vaughan JP, Nobre LC, Lombardi D, Teixeira AM, Fuchs SC, Moreira LB, Gigante LP, Barros PC. Evidence for a strong protective effect of breast-feeding against infant deaths due to infectious diseases in Brazil. Lancet 1987;2:319-22.

20. Victora CG, Huttly SR, Fuchs SC, Nobre LC, Barros FC. Deaths due to dysentery, acute and persistent diarrhoea among Brazilian infants. Acta Paediatr 1992; S381:7-11.

21. Brown KH, Black RE, Romana GL, Kanashiro HC. Infant feeding practices and their relationship with diarrhea and other diseases in Huascar (Lima), Peru. Pediatrics 1989;83:3140.

22. Victora CG, Smith PG, Vaughan JP, Nobre LC, Lombardi D, Teixeira AM, Fuchs SC, Moreira LB, Gigante LP, Barros PC. Infant feeding and deaths due to diarrhoea: a case-control study. Am J Epidemiol 1989;129: 103241.

23. Popkin BM, Adair L, Akin JS, Black R. Briscoe J. Flieger W. Breastfeeding and diarrhea! morbidity. Pediatrics 1990;86:87482.

24. Victora CG, Kirkwood BR, Ashworth A, Black RE, Rogers S. Sazawal S. Campbell H. Gove S. Interventions for preventing pneumonia among young children: nutritional risk factors. Geneva: WHO, Child and Health Development Division, 1996.

25. Victora CG, Fuchs SC, Flores A, Fonseca W. Kirkwood B. Pneumonia among Brazilian children: a hierarchical analysis. Pediatrics 1994;93:977-85.

26. Fonseca W. Kirkwood BR, Victora CG, Fuchs SC, Flores JA, Misago C. Risk factors for childhood pneumonia among the urban poor in Fortaleza, Brazil: a case-control study. Bull WHO 1996;74:199-208.

27. Victora CG, Smith PG, Vaughan JP, Barros FC, Fuchs SC. Risk factors for deaths due to respiratory infections among Brazilian infants. Int J Epidemiol 1989;18:901-8.

28. Amaral JJF. Prevalencia e fatores de risco pare infecçao respiratória aguda em crianças aos seis meses de vida em Pelotas. Master's dissertation, Universidade Federal de Pelotas, Brazil, 1995.

29. Aaby P. Bukh J. Lsse IM, Smits AJ, Smedman L, Jeppsson O. Lindeberg A. Breastfeeding and measles mortality in GuineaBissau. Lancet 1981;ii:1231.

30. Narayanan I, Prakash K, Munthy NS, Gujral W. Randomized controlled trial of effect of raw and pasteurized human milk and of formula in the incidence of neonatal infection. Lancet 1984;ii:1111-3.

31. Barros FC, Semer TC, Tonioli Filho S. Tomasi E, Victora CG. The impact of lactation centres on breastfeeding patterns, morbidity and growth: a birth cohort study. Acta Paediatr 1995;84:1221-6.

32. World Health Organization. Contemporary patterns of breastfeeding. Geneva: WHO, 1981:31-8.

Contents - Previous - Next