Incorrect Outcome Variable?

To the Editor.—Forste et al1 conclude that increases in breastfeeding among black women would narrow the gap between black and white infant mortality. They have incorrectly identified their outcome variable and reached conclusions not supported by their research.This is not a study of infant mortality (deaths in the first year of life). By excluding infant deaths in the first month of life, they have studied something akin to postneonatal mortality (deaths after day 27). High black infant mortality is largely caused by the increased incidence of preterm delivery in black women (compared with white women). Most infant mortality among black preterm babies occurs in the first 27 days.In addition, the authors have provided no evidence of the direction of causality in the association between breastfeeding and mortality, and they have failed to distinguish between the presence of a risk factor and the impact of removing that risk factor. Even if breastfeeding were associated with infant mortality and the relationship were causal, there is no evidence that changing the breastfeeding rate would result in a decrease in infant mortality.The authors may have rediscovered the protective effect of breastfeeding on mortality. Black postneonatal mortality is higher than white; some of this is attributable to higher black sudden infant death syndrome (SIDS) than white SIDS. And breastfeeding protects against SIDS.2,3In Reply.—Rosenberg, Acuna, and Mather are correct in that we do not measure infant mortality according to the standard measure of deaths during the first year. We exclude deaths during the first month because the National Survey of Family Growth does not provide enough detail to determine whether infants who died during the first month lived long enough for the mother to have breastfed. Had we included deaths during the first month, we would have overestimated the benefits of breastfeeding to infant survival. The analyses would have wrongly attributed many deaths occurring shortly after birth to a lack of initiating breastfeeding.Rosenberg, Acuna, and Mather are correct in that what we estimate is more accurately described as postneonatal mortality. It would be inappropriate to apply our findings to neonatal mortality. Although black infant mortality is high during the first month, as noted by Rosenberg, Acuna, and Mather, we still found the mortality rate for black infants to be 1.5 times higher than the rate for whites during the postneonatal period. This is substantially lower than the infant mortality difference of 2.4 reported by the Census Bureau for 1995. This difference of 1.5, however, reflects a large racial disparity that merits our consideration.In examining the effects of breastfeeding on this difference in mortality, we acknowledge, as with all nonexperimental research, that it is difficult to draw firm conclusions about cause-and-effect. Without the benefits of an experimental setting, we are only able to present, based on a national sample, a broad look at the relationship between ever breastfeeding and infant mortality. It is unlikely that breastfeeding research will ever have randomly assigned treatment groups, thus the need to rely on nonexperimental evidence. Including breastfeeding in our analyses does reduce the size of the coefficient indicating the racial difference in mortality. This finding presents a potential avenue for consideration in addressing the racial gap in infant mortality, an avenue that Rosenberg, Acuna, and Mather acknowledge may be related to factors such as the protective effect of breastfeeding on SIDS. It is our expectation, that having found such a pattern using national data, that additional research will be able to confirm and clarify the causal mechanisms between breastfeeding and infant mortality.