Did breastfeeding save lives in Shelby County Tennessee?

Confused businesswoman looking at camera

It’s just the kind of evidence I have been looking for.

I’ve relentlessly hammered the point that although small studies suggest that breastfeeding has benefits in industrialized countries, population data has failed to demonstrate the predicted benefits in real world experience. The decreases in infant mortality of term babies, reduced incidence of various conditions and savings in healthcare costs have never materialized.

[pullquote align=”right” cite=”” link=”” color=”” class=”” size=””]The data suggest reverse causality: deathly ill newborns can’t breastfeed; that doesn’t mean newborns who don’t breastfeed become deathly ill.[/pullquote]

A new paper in Breastfeeding Medicine,Associations Between Breastfeeding Initiation and Infant Mortality in an Urban Population, analyzing data from a poor, urban area (Shelby County, Tennessee) purports to show that the benefits are real, not merely theoretical:

Initiation of breastfeeding was associated with a significant reduction in total infant mortality (OR=0.81, 95% CI=0.68–0.97, p=0.023). Neonatal mortality was also significantly reduced with any breastfeeding (OR=0.49, 95% CI=0.34–0.72, p=0.001). Postneonatal mortality was not significantly asso- ciated with breastfeeding initiation in the overall population (OR = 0.95, 95% CI = 0.78–1.17, p = 0.65), but was significant in the nonblack population (OR = 0.63, 95% CI = 0.41–0.98, p = 0.039). An association was observed between breastfeeding initiation and infant mortality from infectious disease (OR = 0.49, 95% CI = 0.32–0.77, p = 0.002).

The authors conclude:

In an urban area with high infant mortality and low breastfeeding rates, initiation of breastfeeding was significantly associated with reductions in overall infant mortality, neonatal mortality, and infection-related deaths. Breastfeeding promotion, protection, and support should be an integral strategy of infant mortality reduction initiatives.

But if you take a closer look, it’s just another example of breastfeeding researchers assuming causality without justification. And they leave out some critical information that threatens their conclusions.

The authors start with a very impressive set of graphs:


It shows that as breastfeeding rates rose in Shelby County, infant mortality declined. The decline among black women in particular is quite impressive.

How does that compare to the surrounding area?

Here’s a graph from another source, a report from the Shelby County Health Department:


Against the backdrop of an 18% decline in infant mortality in Tennessee, Shelby County experienced a remarkable 34% drop.

Was breastfeeding responsible? The authors of the new paper compared ever breastfed to never breastfed infants. In other words, if the baby was put to the breast at all, it was classified as ever breastfed, regardless of whether the baby was breastfed exclusively or breastfed for any length of time beyond the postpartum hospital stay.

They claim that breastfeeding is the cause of the reduction in infant mortality. Looking at their data suggests otherwise.


I added the red arrows to highlight the results that are statistically significant. They’re startling: breastfeeding appears to have its impact on neonatal deaths (deaths in the first 28 days), not postneonatal deaths; that’s the opposite of what you might expect since the purported benefits of breastfeeding are supposed to accrue over the long term, not immediately.

Indeed, when assessing the benefits of breastfeeding, many researchers have excluded the first month after birth entirely. As the influential 2004 paper Breastfeeding and the Risk of Postneonatal Death in the United States notes:

Because infants who are sick from birth may be unable to breastfeed and children who become ill later may stop, breastfeeding infants may seem healthier because illness, especially mortal illness, prevents breastfeeding rather than because breastfeeding prevents illness. The recommended methods for dealing with this problem are to exclude deaths that occur in the neonatal period and to assign feeding category by how the child was fed at some time before death occurred… These tactics do not exclude reverse causality completely, but they should minimize its effects.

Had the authors of the new paper followed these guidelines, they would not have much to write about, since the bulk of comparisons between never breastfed infants and ever breastfed infants showed no statistically significant difference. So although they reference these guidelines, they chose to add neonatal mortality back in though they excluded the first week post birth. Their statistically significant results occur during that period, the very time that reverse causality is most likely to affect results. The differences between ever breastfed and never breastfed infants who died before 28 days of life may simply reflect the fact that the babies who died were never well enough to breastfeed.

Looking at the causes of death raises further doubts.


Ever breastfed infants did NOT have a reduced risk of SIDS and there were so few deaths from NEC that the impact of breastfeeding could not be assessed. The main difference between ever breastfed and never breastfed infants was in deaths from infectious agents.

According to the authors:

Infectious etiologies coded on the death certificate identified the following organisms: Staphylococcus species; Streptococcus species; Klebsiella, Pseudomonas, and Escherichia coli species; cytomegalovirus; herpes; Candida; “viral”; Meningococcal species; “Whooping Cough”; and unspecified.

CMV is transmitted before birth and herpes during birth. Klebsiella, Pseudomonas, and systemic Candida are devastating infections that are often acquired by compromised infants in the NICU. It is likely, therefore, that in the case of these infectious agents, the babies were too sick to ever breastfeed. It wasn’t the lack of breastfeeding that led to the infections; it was the infections that led to the lack of breastfeeding.

So the data in this paper suggests that while breastfeeding was associated with fewer infant deaths, it did not cause fewer infant deaths.

Equally important, the authors failed to note that in 2006 Shelby County had embarked on a multi-pronged plan to reduce its high infant mortality rate.

This table appears across two pages in the presentation and I pasted them together:


As you can see, breastfeeding was one prong and indeed breastfeeding rates rose by 7.2% but there were other successes as well: the rate of early prenatal care rose; the rate of smoking in pregnancy dropped; and the teen pregnancy rate dropped by more than 21%. These factors almost certainly contributed to the drop in infant mortality in Shelby County but they aren’t even mentioned in new paper.

The bottom line is this:

1. The authors of the paper recognize that real world evidence (the same evidence I have been demanding from lactation professionals) is critical to demonstrating that the benefits of breastfeeding are real.

2. The authors claim that their data shows that increased breastfeeding caused decreased infant mortality. The evidence they provided shows the opposite: breastfeeding had almost no impact on postneonatal mortality and the effect on neonatal mortality almost certainly reflects that fact that deathly ill newborns can’t breastfeed, NOT that newborns who don’t breastfeed become deathly ill.

3. The authors disingenuously failed to mention a major multi-pronged effort to reduce infant mortality in Shelby County that produced critical changes — such as the steep drop in the teen birth rate — that almost certainly had as much or more to do with the decline in infant mortality than breastfeeding rates.

In other words, this paper did NOT show that breastfeeding provides real world benefits. We’re still waiting for that evidence.