Breast feeding and infant mortality.

The evidence linking bottle feeding to infant and early childhood mortality has been reviewed. Ecological studies of national time trends in infant mortality do not parallel breast feeding trends in those countries, and indicate that falling death rates are more likely to be related to better health care facilities and social conditions. Direct studies of deaths provide some contradictory findings; meta-analyses are not informative because of the many differences in statistical and sample methodology. The methodology exhibited in most studies is more likely to have over- rather than under-estimated a relationship between bottle feeding and infant mortality. Retrospective analyses must take account of changes in feeding pattern due to early signs of illness. Prospective population studies able to account for large numbers of potential confounders provide the best estimates, especially if proportional hazards models are used. Two such studies have been carried out--both showed protective effects of breast feeding.