Plausibility is never a substitute for empirical evidence.

Breastfeeding is a delicate issue. There are social, financial, and health interests. Everyone has convictions but few have data. As is often the case, the argument made seems a good deal more than plausible. What is better for a baby than a mother's milk?

But if we want to be rationalists, we always check plausible arguments against the actual data. Emily Oster does so in Everybody Calm Down About Breastfeeding. As a reporter for fivethirtyeight, she focuses on the numbers.

In the run-up to my son’s birth a couple of months ago, I spent a lot of time sitting in my midwife’s office staring aimlessly at the posters on the wall. My favorite one depicted two scoops of ice cream with cherries on top, strategically set to look like breasts. The caption underneath suggested that exclusive breastfeeding for six months would lower a child’s risk of obesity. Presumably the implication was that if you chose to breastfeed, your child could later eat ice cream with impunity.

It was a great visual, and given the current rate of obesity in the United States, a compelling argument. The only trouble is that there is no good reason to think it is correct. The one high-quality randomized controlled trial of breast-feeding did not show any impacts on childhood obesity.

Of course, it’s not just childhood obesity. The purported benefits of nursing (here is one list from the California Department of Public Health) extend to better mother-infant bonding, lower infant mortality, fewer infections in infancy, higher IQ, higher wages in adulthood, less cancer and on and on. If one takes the claims seriously, it is not difficult to conclude that breastfed babies are all thin, rich geniuses who love their mothers and are never sick a day in their lives while formula-fed babies become overweight, low-IQ adults who hate their parents and spend most of their lives in the hospital.

Good people honestly making plausible claims which they then wish the power to impose on others.

But is the claim true? Oster presents much research to suggest that at the very least we need to return a Scottish "Not Proven".

You might wonder: If this is correct, why all the mania about breastfeeding? Why all the policy focus? Why put out a poster trumpeting the value of breastfeeding against obesity if it is wrong? The simplest reason is that people (including policymakers) may not stop to think about which research they should believe — and, as I’ve noted, there are plenty of (flawed) papers that would point to all kinds of breastfeeding benefits. It often takes time for good research to trump bad — many people still think a low-fat diet is a good idea even though randomized trial data has not supported that belief.

It’s also the case that when it comes to our kids, we all want to believe that what we are doing is the best thing for them. Ensuring that your child gets exclusively breast milk for six months or a year can be very challenging, and it may be that we want to believe the sacrifice has some benefits. And it does. Just not nearly as many as we might think.

To actually learn about the impacts of breastfeeding, we need to rely on studies in which breastfeeding is assigned randomly (the best option) or, in the absence of that experiment design, studies that somehow fully adjust for differences across women.

This leaves us with a small but informative set of studies. In the first camp — the randomized trial camp — we have one very large-scale study from Belarus. Known as the PROBIT trial, it was run in the 1990s and continued to follow up as the children aged. 1 The study randomized women into two groups, one in which breastfeeding was encouraged and another in which it wasn’t, and found that the encouragement treatment increased breastfeeding rates. The trial has studied all sorts of outcomes, including infant and child health and cognitive development.

Given how much interest there is in this topic, it is perhaps surprising that we have only this one large randomized trial of breastfeeding. It’s not clear to me why this is the case. People may be so convinced of the benefits of breastfeeding that they see no need for further testing. Or it may be that a large enough study is too daunting and expensive to run. Whatever the reason, the randomized evidence is limited to this single case.

The other group of helpful studies are those of siblings — that is, studies that compare a sibling who has been breastfed with one who has not. These aren’t as good as randomized trials, since there must be some reason why one child was nursed and one was not, but they do get around the issue of mothers who breastfeed being different from those who don’t.

When people cite the benefits of breastfeeding, those benefits typically fall in one of three groups: early life health, later life health and IQ. So, what does the data say?

Breastfeeding and infant health

Among the most straightforward benefits claimed for breastfeeding are improvements in infant health. The randomized trial in Belarus evaluated a very large number of infant health outcomes. The results are decidedly mixed.

Infants in the treatment group — who, remember, were more likely to be breastfed — had fewer gastrointestinal infections (read: less diarrhea) and were less likely to experience eczema and other rashes. However, there were no significant differences in any of the other outcomes considered. These include: respiratory infections, ear infections, croup, wheezing and infant mortality.

In other words, the evidence suggests that breastfeeding may slightly decrease your infant’s chance of diarrhea and eczema but will not change the rate at which he gets colds or ear infections and will not prevent death.

Breastfeeding and child health and behavior
Many sources (for example, that poster in my midwife’s office) go further and claim that breastfeeding has health benefits for the child in the long term — lower obesity risk, better blood pressure and so on. And on top of this, there are the claimed benefits on behaviors — less hyperactivity, fewer behavior problems, more maternal attachment, etc.

Here, the evidence is not mixed. It rejects these claims across the board.

The PROBIT randomized trial is again the best source. The researchers analyzed the impacts of breastfeeding on allergies and asthma; on cavities; and on height, blood pressure, weight and various measures of obesity. They found no evidence of nursing’s impacts on any of these outcomes. They also found no evidence of impacts on child behavior issues, emotional problems, peer issues, hyperactivity or maternal-child connection.

These non-results are also present when we look at sibling data. This study, published last year in the journal Social Science & Medicine, compares siblings who were and were not breastfed on various health and behavior outcomes — obesity, asthma, hyperactivity, parental attachment, compliant behavior. They find no impact of breastfeeding on any outcome.

These sibling results also provide a sense of why less effective research methods would have shown impacts. The researchers demonstrate that if you simply compare breastfed to formula-fed infants, there are significant differences on virtually all outcomes. However, once you compare siblings — where differences among mothers are fully adjusted for — you no longer see any positive impacts of breastfeeding.

More detail in the article. Plausibility is never a substitute for empirical evidence.