Because the effects of breastfeeding can be evaluated only by observational methods, studies of the relation between breastfeeding and illnesses are subject to potential limitation by misclassification of exposure and outcome and by confounding. Bauchner et al. (15) listed four standards for evaluation of breastfeeding studies in relation to these limitations: avoidance of detection bias, definition of the outcome event (e.g., respiratory illnesses), definition of breastfeeding, and adjustment for potential confounding factors. In this study, we limited detection bias, i.e., information bias, through prospective, frequent, standardized surveillance of all subjects. We categorized breastfeeding according to specified criteria and ascertained the breastfeeding status of each participant on a 2-week basis. We assessed respiratory symptoms every day by using a symptom diary and categorized symptoms into illnesses and validated the agreement between calendar diary-defined illnesses and diagnoses made by clinicians (22). We gathered information on potential confounding factors and used multivariate statistical models to adjust for these factors. By restricting participants to infants from households without smokers, we eliminated the possibility of confounding by exposure to environmental tobacco smoke, also a cause of LRI (26).
Although our study met the standards offered by Bauchner et al. (15), the findings should not be extended to all infants because of the selection criteria for study participants. The infants who were excluded from our study, those living in homes with smokers or attending day care on a full-time basis, are now a majority of US infants. It is possible that our cohort was at lower risk than the infants excluded by these criteria and that the effect of breastfeeding would be different in infants at higher background risk.
We found that the overall incidence of respiratory illnesses, including URIs and LRIs, was not affected by breastfeeding (tables 2 and 3). However, breastfed infants had fewer illnesses classifiable as LRIs during the first 4 months of life (figure 2). LRIs were of shorter duration in breastfed infants, but the difference in duration was not statistically significant. The duration of all respiratory illnesses combined was significantly shorter in infants who were fully breastfed (figure 3). We interpret this pattern of reduced incidence of LRIs and shorter duration of all respiratory illnesses as evidence that breastfeeding reduces the severity of respiratory illnesses without lowering the incidence. Other studies of breastfeeding have focused on the incidence of illnesses and have used classification schemes that were not sufficiently precise to examine transfers from one diagnostic category to another, as from URI and LRI. Other studies have not addressed the effect of breastfeeding on illness duration.
Our findings are consistent with others in regard to the reduction of LRI by breastfeeding and the age dependence of the effect of breastfeeding. In the Tucson Children's Respiratory Study, breastfeeding reduced the risk of all LRIs (27) and of respiratory syncytial virus-associated LRIs (28), primarily during the first 4 months of life, and breastfeeding was associated with a lower proportion of wheezing illnesses as well. The Tucson Study ascertained respiratory illnesses through contact with a health care provider, and consequently, the incident LRI cases would tend to be more severe than our study. In another prospective cohort study conducted in Scotland, infants breastfed during the first 13 weeks of life had a significantly lower incidence of LRI during the first 13 weeks of life and during weeks 40-52 (29). Most studies having sufficiently large numbers of infants showed protection against LRIs (27), particularly illnesses due to respiratory syncytial virus (9, 28).
Duncan et al. (30) found that infants who were fully breastfed for 4 months had half as many episodes of acute otitis media as did infants who were not breastfed. In a retrospective study of Pima Indian children, breastfeeding provided protection through age 8 months against having medical care for an URI or otitis media (31). We did not find an effect of breastfeeding on URIs at any age and did not ascertain otitis media in this study.
The consistency of our findings with those of these earlier reports suggests that the benefits of breastfeeding vary with age and do not extend beyond the first 6 months of life. This age pattern may reflect the relatively larger proportion of total dietary intake received as breast milk during the first months of life and the higher concentration of immunoglobulin A, the major component of local mucosal immunity, early in lactation (32).
We did not find persistent benefit beyond age 6 months for breastfed children or any indication that the reduced illness risk during the initial months of life was followed by either subsequent benefit or increased risk.
The weight of evidence from this and other studies is that infants less than 6 months old have fewer LRIs if they are fully breastfed than if they are not. For this reason, among others, promotion of breastfeeding is a worthwhile strategy for prevention of LRIs in young infants, the age group most likely to require hospitalization (33) and to suffer long-term consequences of such illnesses (34).