Breastfeeding Medicine

Physicians blogging about breastfeeding

The Breastfeeding and Obesity Controversy

with 8 comments

Fact or fiction: Is breastfeeding actually linked with a lower risk of childhood obesity? The common thinking for the past several years has been “yes,” based on comprehensive analyses from the US Agency for Healthcare Research and Quality and the World Health Organization. However, two recent articles have disputed these conclusions. Considerable media attention has surrounded new data from the Promotion of Breastfeeding Intervention Trial (PROBIT), published in the Journal of the American Medical Association (JAMA) on March 13, and in a January 31 article in the New England Journal of Medicine (NEJM) on obesity myths by Casazza et al. We discuss both articles’ conclusions, as the public tries to make sense of all the conflicting information.

In PROBIT, hospitals in Belarus were randomized to Baby-Friendly type maternity practices or usual care. The investigators then looked at over 17,000 healthy breastfed infants for their health outcomes, and found no effect on childhood obesity in childhood or early adolescence.

However, just because this study found no link does not mean there is no link between breastfeeding and lower risk of obesity. This is because the PROBIT study design only supported the detection of extremely large differences in obesity rates attributable to exclusive breastfeeding. It was not designed to be able to detect any moderate or smaller differences. Furthermore, the study did NOT compare breastfed infants to fully formula-fed infants, further diminishing the chance of finding any differences in health outcomes. Because exclusively formula-fed infants were not even included in PROBIT, one cannot draw any conclusions about the safety of formula feeding in terms of obesity risk.

A strength of PROBIT is that it is a randomized controlled trial, considered the gold standard in scientific research. Its design means that the investigators could minimize the chance that the health effects of breastfeeding were due to other healthy behaviors that breastfeeding families might have engaged in, rather than breastfeeding itself. However, PROBIT has other limitations, most notably that all the breastfeeding rates in both arms of the study were quite low. The hospitals in the intervention arm did markedly increase their breastfeeding rates: exclusive breastfeeding at 3 months was from 6% in the usual care group, and 43% in the intervention group, while exclusive breastfeeding at 6 months was 0.6% in the usual care group to 7.9% in the intervention group. Any breastfeeding at 12 months was 11.4% in the usual care group and only 19.7% in the intervention group.

In other words, few infants in either group were still breastfeeding at one year, with only 1/5 of all infants still breastfeeding even in the intervention group. This means that any health effect found would have to be enormous to show up at all. The previous work from PROBIT found that the population of obese children did not change in the intervention group. The new PROBIT study uses the same populations as the original publication, so it is completely expected that it would also find no effect of breastfeeding on childhood obesity, as in their original study.

Indeed, a 2011 analysis of PROBIT showed that the PROBIT study was far too small to detect a significant association. In other words, because it is a population based study with very low breastfeeding rates, PROBIT does not have the sample size to show the level of protection that has been shown in other studies.

Most studies showing an effect of breastfeeding on obesity relate it to longer durations of breastfeeding, generally at least 6 months. There was a dose-response relationship, meaning more breastfeeding meant less obesity, and this is one piece of evidence that suggests that breastfeeding actually helps cause lower rates of obesity, and is not just merely associated with it. A well-respected 2007 meta-analysis from the US Agency for Health Care Research and Quality found good evidence of an association between breastfeeding for at least 3 months and obesity in later life, taking into account multiple confounding factors. A WHO meta-analysis from later in 2007 also found good evidence to link longer breastfeeding with lower rates of obesity. These authors addressed the issue of publication bias, noting that only studies that show an effect are published while studies that fail to show an effect are not published. Despite this, they concluded that there was still a convincing effect of breastfeeding on obesity, even taking into account publication bias, socioeconomic confounders and parents’ obesity. In many of the 33 studies examined, the effects were seen with breastfeeding in durations in the range of 3 to 9 months. Although a 2010 sibling comparison study found a significant relationship when comparing breastfed with non-breastfed siblings within the same family-unit,8 re-analyses of the same data by other authors found only a “trivially small” relationship.

The New England Journal article had a short section on obesity and breastfeeding in its discussion on “obesity myths,” but has been widely quoted in the media as debunking any association between breastfeeding and a lower risk of obesity later in life. However, the authors’ look at obesity was quite cursory, and in fact does not discount the possibility that breastfeeding itself may lower a child’s risk for obesity.

The New England Journal authors cite the earlier work published from PROBIT as evidence of no association between breastfeeding and obesity.5,10 The New England Journal article highlights the publication bias of the WHO study, meaning studies that show an effect of breastfeeding are more likely to be published than studies that do not. More pointedly, the NEJM authors do not comment on the AHRQ analysis at all. They point out that the Egger test of population bias showed that smaller studies in the WHO analysis tended to have the largest effects. However, the Egger test is only one test of publication bias—the alternative Begg test had a p-value of 0.96, indicating that publication bias was not a significant issue. Owen et al pointed out that even though the effect size was larger in small studies, the association was still present in larger studies and was statistically significant. Even a 2008 article critical of the WHO study still found an effect of breastfeeding on obesity,12 but it was smaller than the WHO concluded.

The New England Journal authors picked one statistic that supports their conclusion and ignored the other statistics reported in their other “obesity myths” in the same article. Notably, the authors disclosed funding from many groups with a conflict of interest, including the Mead Johnson Nutrition (makers of Enfamil infant formula), Global Dairy Platform and other dairy associations (most infant formula is made from dairy products). Authors also received funding Kraft Foods, McDonald’s, Coca-Cola, PepsiCo, and Jenny Craig, all of which markedly diminishes the credibility of any scientific research on obesity. It’s unclear why the esteemed New England Journal would choose to publish an article with such a cursory review and so many obvious conflicts of interest.

Finally, the mechanism by which breastfeeding may be linked to lower rates of obesity may be multi-factorial. In addition to the breastmilk itself, some evidence suggests that consuming milk from a bottle may be linked with a caregiver’s desire to empty the bottle, which may contribute to overfeeding and obesity. Children fed exclusive at the breast have to rely on internal cues to determine if they are full or not. (Imagine eating a plate of food with your eyes closed—only your stomach tells you when you’ve had enough.) Research on infant feeding practices needs to consider both the type of food (formula vs. breastmilk) and delivery (bottle-fed versus fed at the breast) when estimating a link with childhood obesity.

So, after all this discussion, the media may still be asking: “Is breastfeeding actually linked to a lower risk of childhood obesity?” Our answer: It is premature to conclude that there is no independent relationship between breastfeeding and childhood obesity. However, this may be the wrong question to ask. Rather, we should be asking, “Is exclusive/predominate formula feeding linked with an increased risk of childhood obesity?” From current available data, any effects of breastfeeding on childhood obesity are likely not large, and tend to be most noted when formula feeding is compared with longer durations of breastfeeding.

Melissa Bartick, MD, MSc is an internist at Cambridge Health Alliance and Assistant Professor at Harvard Medical School. Nathan Nickel, MPH, PhD, is Nathan Nickel is a post-doctoral research fellow at the Manitoba Centre for Health Policy (MCHP).

Posts on this blog reflect the opinions of individual ABM members, not the organization as a whole.

References

Written by melissabartick

March 15, 2013 at 5:30 pm

8 Responses

Subscribe to comments with RSS.

  1. Well said Melissa. Thank you

    Karen Peters

    March 15, 2013 at 7:26 pm

  2. [...] the study design.  For this I owe a debt to Drs. Melissa Bartick and Nathan Nickel, who wrote an excellent analysis for the Academy of Breatfeeding Medicine blog.  Their critique boils down to these [...]

  3. With much thanks to Nathan and Melissa, I think there are two vital issues in the interpretation of this study that are not emphasized:
    1. The entire population was not heavy. In fact, the average BMI of 18 is on the low side.
    2. The Breastfed arm are taller
    3. In looking at the percent obese, it should be analyzed by studying the statistical variance from 15%, rather than comparing one group to the other. The question should be: did either group show any statistical difference from normal distribution? My guess, looking at these numbers, is that the answer is no.

    Miriam Labbok, MD

    April 10, 2013 at 2:24 pm

  4. With much thanks to Nathan and Melissa, I think there are three vital issues in the interpretation of this study that are not emphasized:
    1. The entire population was not heavy. In fact, the average BMI of 18 is on the low side.
    2. Those in the breastfed arm are taller
    3. In looking at the percent obese, it should be analyzed by studying the statistical variance from 15%, rather than comparing one group to the other. The question should be: did either group show any statistical difference from normal distribution? My guess, looking at these numbers, is that the answer is no.

    Miriam Labbok, MD

    April 10, 2013 at 2:24 pm

  5. [...] suggests may have a protective effect. Yet that’s exactly what the authors did, causing many to question the integrity of not just the article but of the journal [...]

  6. […] suggests may have a protective effect. Yet that’s exactly what the authors did, causing many to question the integrity of not just the article but of the journal […]

  7. I remember an article by von Kreis that got lots of media coverage about breastfeeding protecting against obesity. A few years later, I reviewed another article by von Kreis from a similar population, probably about factors affecting childhood obesity, and in that analysis, breastfeeding had shown no effect, so they had left it out of their model. So one article showing a positive association got highly cited, and one that had found a negative association barely mentioned it and got no coverage. This is the kind of thing that skews published data towards positive rather than null findings.

    CarolynS

    March 4, 2014 at 8:39 pm


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

Join 681 other followers

%d bloggers like this: