To determine whether breast-feeding protects infants from infections, a case-control study was conducted. The cases were previously healthy children who were admitted to Yale-New Haven Hospital for an infectious illness at or before 90 days of age. The controls were chosen from the log of births and matched to the cases for five important demographic variables. In addition, logistic regression models were used to adjust the results for other potential confounders. To minimize the potential surveillance bias that might occur if formulafed and breast-fed infants with the same degree of illness have a different probability of being hospitalized, the case-control pairs were stratified by the severity of the medical condition of the case at the time of hospitalization. For the 281 case-control pairs, the matched odds ratio was .50 (95% confidence interval .32, .77; P < .005), which indicates that breast-feeding is protective against infections. However, this apparent protective effect was diminished substantially when the data were stratified according to the severity of illness: the matched odds ratio for the 164 infants with serious illnesses was .79 (.47, 1.32; P < .50), and for the 117 infants with mild illnesses it was .17 (.03, .44; P < .001). These stratified results suggest that breast-feeding protects infants from hospitalization rather than from infections. Failure to consider the problem of surveillance bias may lead to erroneous conclusions about the protective effect of breast-feeding.
- Received August 1, 1985.
- Accepted February 10, 1986.
- Copyright © 1986 by the American Academy of Pediatrics