The protective effect of human milk against acute infections has been addressed extensively (Brown et al. 1989, Cushing and Anderson 1982, Eager et al. 1984, Fergusen et al. 1981, France et al. 1980, Glass and Stoll 1989, Howie et al. 1990, Launer et al. 1990, López-Alarcón et al. 1992, Victora et al. 1987, Wright et al. 1989). A number of studies conducted in the last decade report a reduced incidence of infectious diseases in breast-fed infants compared with those who are fed other milks. Most of these studies claim that human milk provides such a protection against diarrhea (Brown et al. 1989, Cushing and Anderson 1982, France et al. 1980, Glass and Stoll 1989, Howie et al. 1990, Victora et al. 1987). However, whether breast feeding protects against acute respiratory infection (ARI) as well remains unclear.

Although there is no doubt that some factors present in human milk provide resistance against infections, the barrier effect of breast-feeding against the fecal-oral contamination that causes diarrhea is of great importance because high levels of environmental contamination are strong determinants for diarrhea in developing countries (Butz et al. 1984). In the case of acute respiratory infections, the barrier effect does not exist; thus, the mechanisms for such a protective effect must include substances with anti-microbial activity (Andersson et al. 1986, Fishaut et al. 1981, Ryan-Poirier and Kawaoka 1993) or that modulate the immune response of the infant (Goldman et al. 1996). The evidence of protection against respiratory tract infections (Howie et al. 1990, Pisacane et al. 1994, Wright et al. 1989) relates more to lower incidence rates, but little attention has been given to the effects on the severity and duration of individual episodes.

To investigate whether human milk is protective against acute respiratory infection as well as diarrhea, we designed a follow-up study to test the hypothesis that incidence and duration of both acute respiratory infections and diarrhea are lower in breast-fed than in formula-fed infants.

SUBJECTS AND METHODS

Infants were selected if they were singleton, full-term, healthy, weighed >2500 g at birth and were born in this prenatal care center. Subjects with congenital malformations and those who moved out of the neighborhood were excluded.

Diarrhea was defined as the presence of three or more liquid or semiliquid stools per day accompanied or not by blood, mucus or fever. The total number had to exceed the usual number of daily bowel movements.

Individual episodes were defined as an event separated from another event by at least 2 d free of symptoms.

Incidence was described as the number of episodes/100 d of observation. Prevalence was calculated as the percentage of days ill in each month. Relative risks and 95% confidence intervals were computed for each month of age, stratifying the subjects according to their feeding mode for that particular month. Fully breast-fed infants were considered to be not exposed. The risks were contrasted by the chi-square test, and confidence intervals calculated by Miettinen's method (Albhom and Norell 1987). Duration of individual episodes was computed as the number of consecutive days with a defined illness. Episodes were pooled for 1-4 mo for ARI and for 2-6 mo for diarrhea and compared by feeding practice using Student's t test.

In another analysis, infants were stratified on duration in months they received full breast-feeding. Those fully breast-fed for the whole 6-mo follow-up period were the comparison group. The effect of duration of full breast-feeding was examined using a multiple linear regression model in which potential confounders were included: where the y term was alternately the number of episodes or the number of days ill with ARI and diarrhea; X₁ = total duration of full breast-feeding in months; X₂ = maternal age; X₃ = years of education; X₄ = number of individuals sleeping in the same room (crowding); and X₅ = marital status. An index of the socioeconomic level of the family (based on the availability of tap water, sewage, refrigerator and paved floor) and the presence of siblings (yes = 1, otherwise = 0) were introduced as covariates in the analysis. Because maternal age was correlated with the number of siblings living in the same household (r = 0.52, P = 0.008), it was removed from subsequent regression models.

A logistic regression model was carried out to examine the coincidence of an episode of diarrhea occurring immediately after solid food was introduced. ANOVA, median test (Siegel 1990) and ² were used, where appropriate, to contrast maternal, infant and household characteristics among feeding practice groups.

RESULTS

Table 1. Maternal, child, and household characteristics of the sample by duration of breast-feeding [View Table]

Table 2. Feeding practices in fully, partially breast-fed and formula-fed infants from 0 to 6 mo of age [View Table]

Relative risks and confidence intervals for incidence and prevalence of acute respiratory infection are presented in Tables 3 and 4. The probability of suffering an episode of ARI was higher for formula-fed than for fully breast-fed infants during the first 4 mo. Thereafter, differences were not significant. The risks for the partially breast-fed infants were intermediate between those of formula-fed and fully breast-fed, except for the last 2 mo of follow-up. The probability of having a larger percentage of days ill was also significantly higher for formula-fed than for breast-fed infants from birth up to 4 mo.

Table 3. Incidence of acute respiratory infections by category of feeding practice and age in infants 0-6 mo [View Table]

Table 4. Prevalence of acute respiratory infections by category of feeding practice and age in infants 0-6 mo [View Table]

The mean duration of episodes occurring from birth to 4 mo of age (n = 111) was higher for the formula-fed than for the fully breast-fed infants (6.4 ± 3.6 vs. 5.1 ± 3.5 d, respectively, P < 0.02). Antibiotics were prescribed in 20 out of 233 episodes recorded (11 formula-fed and 9 fully or partially breast-fed infants).

The number of episodes of ARI was negatively associated with duration of full breast-feeding (r = 0.17, P < 0.05) and the number of siblings (r = 0.16, P < 0.04). That is, infants who never received breast-milk and had at least one sibling were more likely to have an episode of ARI than those breast-fed for at least 1 mo. The number of days ill with ARI tended to be associated with duration of full breast-feeding. This association became stronger when crowding was removed from the model (r = 0.16, P = 0.04). That is, infants who never received breast milk were more likely to have longer episodes of ARI than those breast-fed for at least 1 mo (Table 5).

Table 5. Effect of duration of breast-feeding and potential confounders on the number and duration of acute respiratory infections and diarrhea episodes in infants 0-6 mo of age [View Table]

Relative risks and confidence intervals for incidence and prevalence of diarrhea are presented in Tables 6 and 7. The probability of suffering an episode of diarrhea was higher for the formula-fed than for the fully breast-fed groups from 2 to 6 mo of age. The likelihood of having a larger percentage of days ill was also higher for the formula-fed than for the fully breast-fed infants during the entire follow-up. The relative risks for the partially breast-fed group did not show any trend.

Table 6. Incidence of diarrhea by category of feeding practices and age in infants 0-6 mo [View Table]

Table 7. Prevalence of diarrhea by category of feeding practice and age in infants 0-6 mo [View Table]

The mean duration of episodes of diarrhea was also higher for the formula-fed than for the fully breast-fed infants (6.2 ± 4.4 vs. 3.8 ± 2.2 d, respectively, P < 0.001). Antibiotics were prescribed in 21 out of 95 episodes registered (seven formula-fed and 14 fully or partially breast-fed infants).

The number of episodes and the number of days ill with diarrhea were both associated with duration of full breast-feeding (r = 0.17, P = 0.02 and r = 0.19, P < 0.008, respectively). Maternal education, marital status, crowding, number of siblings, and socioeconomic level were not significantly correlated with diarrhea (Table 5). In the logistic analysis, only duration of full breast-feeding was correlated with the number of episodes of diarrhea; introduction of solid food and occurrence of an episode of diarrhea were not associated (odds ratio = 0.86 and 1.16, CI = 0.75-0.98 and 0.88-1.52, respectively).

DISCUSSION

These results are consistent with recent studies (Beaudry et al. 1995, Brown et al. 1989, Howie et al. 1990, Pisacane et al. 1994, Wright et al. 1989) claiming a protectve effect of breast-feeding against ARI. Efforts were made to control for known confounders mentioned in these studies. An association between breast-feeding and respiratory infections was found in a retrospective study carried out in Canada. In that study, diagnosis of otitis media and of upper and lower respiratory infections was based on a questionnaire mailed to the mother (Beaudry et al. 1995). Two more studies, one combining prospective and retrospective assessments (Wright et al. 1989) and the other a case-controlled, hospital-based study (Pisacane et al.1994), also found a lower incidence of wheezing (but not of non-wheezing respiratory illness) and pneumonia and bronchiolitis, respectively, in breast-fed infants. However, hospital-based studies may bias the results because more severely ill infants are more likely to be included. In addition, retrospective studies are not the best design to use in looking for causality; therefore the results of these studies must be confirmed with better-designed studies. In 1989, Brown et al. conducted a longitudinal study in a poor neighborhood in Lima, Peru, which found that infants were protected against ARI for the first 20 wk postpartum. Although maternal smoking seemed to be uncommon in that community, it was not controlled in the study. In addition, the authors claimed that family size was comparable across feeding practice categories, but its effect was not analyzed explicitly. Protection by breast-feeding against ARI has been also reported in longitudinal studies conducted in developed countries. However, in most of these studies, the association vanished when controlling for confounders, i.e., maternal smoking and family size (Fergusen et al. 1981, Taylor et al. 1982). On the contrary, in a well-designed, longitudinal study, Howie and co-workers (1990) found that breast-fed infants were protected against ARI for the first 13 wk of age. Such protection lasted beyond the end of breast-feeding up to 40-52 wk of age if infants were breast-fed for at least 13 wk. Statistical significance persisted after allowing for confounders.

Several methodological efforts help to strengthen the conclusions in the present study. Community-based surveillance avoided the bias imposed by differences in maternal judgment to obtain health care, or in hospitalization rates according with feeding practices. Detection bias was minimized by frequent and regular surveillance, alternately at home and in the clinic, and by daily reporting of the health status of infants by their mothers. Participant mothers were instructed and supervised by the field personnel so that compliance was 95%. We were not able to find any published validation for such a procedure in developing countries, but there is merit in the fact that it is based on written information during an active surveillance. Misdiagnoses are probably unusual because it is unlikely to misclassify a respiratory infection event as a diarrhea event, except when respiratory infections present with vomiting. In addition, a physician blinded to the feeding practice category assigned the final diagnosis based on maternal observations, nurses records, and pre-established definitions of illness. Assessment of diagnosis concordance between the assigned diagnosis by the researcher and the diagnosis written in the medical records was possible only in a limited number of events (50 out of 325). Although the concordance was almost universal, such evaluation does not seem valid because medical care probably was more likely to be demanded for severe than for mild cases. Women who are breast-feeding may be more likely to detect any episode of illness because of their recognized mother-child attachment. This might result in a detection bias leading to an underestimation of the protective effects of breast-feeding.

The effects of breast-feeding on the prevalence of ARI have been addressed by Brown et al. (1989), who demonstrated fewer days ill with ARI in breast-fed than in formula-fed infants. However, the reported lower percentage of days ill may reflect a lower number of episodes instead of a shorter duration. Our study provides evidence for a lower prevalence of ARI as well as a shorter duration of individual episodes in breast-fed infants, by examining the percentage of days ill and the mean duration of individual episodes.

In this study, none of the participant mothers smoked cigarettes. Smoking habits of the fathers were not investigated. Smoking is highly prevalent in the male population of Mexico City. However, in this area of the city, most working males are away from their homes for 12 h/d or longer because of the lengthy commuting time to their work.

The effect of family size was analyzed using both the number of siblings and the presence of crowding, because each of those two variables affects the morbidity of acute infections differently. These variables functioned as effect modifiers. For instance, the protective effect of duration of full breast-feeding on incidence of ARI was present only for those infants who had siblings, and the association of duration of breast-feeding with the number of days ill with ARI was present only when crowding was removed from the analysis; crowding was not significantly associated by itself.

The mean birth weight, by feeding practice categories, was not different among groups, allowing us to assume that nutritional status was comparable at base line. Hence, any difference in the morbidity rates cannot be attributed to a nutritionally impaired immune response of the children (Chandra 1989). Thus, differences in morbidity might be attributed to the feeding mode. The prescription of antibiotics was infrequent (12% of all events) and unrelated in the analyses to the percentage of days ill.

Babies shifting from one feeding mode category to the other caused the formula-fed group to be a mixture of those who were formula-fed since birth and those with a previous history of breast-feeding. Thus, the risk rates for the formula-fed group were probably underestimated, especially in the later months of follow-up.

For the purpose of this study, the definition of ARI included symptoms and signs of both upper and lower respiratory tract infection, but episodes of lower respiratory tract infections were not detected. Consequently, our results refer to upper respiratory illnesses. The nature of the protective mechanisms against respiratory infections are not as well understood as the mechanisms protecting against diarrhea. There is growing evidence that some bioactive components in human milk protect against respiratory infections. Such is the case for a specific secretory immunoglobulin (sIgA) against syncytial viruses resulting from the bronchomammary pathway (Fishaut et al. 1981) and for an ₂-macroglobulin-like substance that inhibits influenza and parainfluenza viruses (Ryan-Poirier and Kawaoka 1993). Some milk oligosaccharides may inhibit the attachment of Steptococcus pneumoniae and Haemophilus influenzae to host cell surface receptors (Andersson et al. 1986). Lactoferrin may inhibit the growth of gram-positive and gram-negative bacteria by limiting iron availability (Sanchez et al. 1992), by interfering with virus receptor-mediated binding to target cells (Van Berkel et al. 1995), by modulating complement activation (Ferenec and Viljoen 1995), and stimulating natural killer cells (Bezault et al. 1994). Therefore, there is enough evidence to support the belief that respiratory infections in breast-fed infants are less severe than in their formula-fed counterparts.

The protective effect of breast-feeding against diarrhea reported by others was also confirmed in our study. The lowering effect of breast-feeding on the percentage of days ill with diarrhea has been previously reported by Brown et al. (1989), but the finding here of reduced duration of individual episodes may be novel.

A unique form of detection bias resulting in overdiagnosis of the occurrence of diarrhea can occur because stools of very young breast-fed infants are typically much looser and more frequent than those of nonbreast-fed infants (Bauchner et al. 1986). To avoid such a bias, the outcome event definition was as recommended by the World Health Organization (WHO 1984). In addition, we compared the number of bowel movements per day occurring during a given episode to the number of daily bowel movements registered by the mother.

The feeding mode categories in the present study were based on the type of milk consumed regardless of whether water, other liquids, or solid food was introduced. Nonexclusive forms of breast-feeding (i.e., inclusion of other liquids or food) increase the risk of diarrhea (Brown et al.1989). That means that the protection breast milk brings against acute infections may have been underestimated in our study. However, introduction of solid foods in this study was not associated with the appearance of an episode of diarrhea, probably because introduction of liquid or solid food and the daily energy intake were evenly distributed among groups. Popkin et al. (1990) found no differential protection between exclusive and full breast-feeding in a large longitudinal study on almost 3000 infants from both rural and urban areas of the Phillipines.

In conclusion, we present evidence of a protective effect of breast-feeding against acute respiratory infections and diarrhea, which reduced the incidence, percentage of days ill and duration of individual episodes in breast-fed infants. Causal inferences between feeding practices and the risk of ARI and diarrhea cannot be made with our design, because ethical limitations did not permit the random assignment of breast- or formula-feeding to infants. However, the consistency of our results with those of others, the effect of breast-feeding intensity, and the dose-response effect of breast-feeding on the incidence and prevalence of ARI and diarrhea, strongly suggest that such a relationship is plausible.