Early, limited data for early, limited formula use

A small study published in Pediatrics suggests that supplementing newborns with small quantities of formula may improve long-term breastfeeding rates. The results challenge both dogma and data linking supplementation with early weaning, call into question the Joint Commission’s exclusive breastfeeding quality metric, and will no doubt inspire intimations of a formula-industry conspiracy. Before we use this study to transform clinical practice, I think it’s worth taking a careful look at what the authors actually found.

First, I think it’s very important to be clear about what the authors meant by “early limited formula.” The authors used 2 teaspoons of hypo-allergenic formula, given via a syringe, as a bridge for mothers whose infants had lost > 5% of their birthweight and mom’s milk had not yet come in. At UNC, we use donor milk in a similar way, offering supplemental breast milk via a syringe as a bridge until mom’s milk production increases.

This is very different from the way that formula supplementation is handled in many US hospitals. We know that in the US overall, 1/4 of breastfed infants are given formula by day 2 of life, and that number reaches as high as 40% in some areas. Typically, when a family member expresses interest in giving the baby some formula, a hospital staff member plunks a 6-pack of 2-oz bottles of ready-to-feed formula in the baby’s bassinet with no instruction about how much to feed. A neonate whose stomach holds one to two teaspoons gets 2 ounces (12 teaspoons) of milk poured into him. The baby then sleeps for the next four to six hours, like someone who’s just over-indulged at a Thanksgiving buffet. In this scenario. Mom doesn’t get any breast stimulation, and family members all express relief that “finally the baby is happy.” When baby finally wants to eat again, there are five more convenient, ready-to-feed, six-hour-nap-inducing bottles sitting in the bassinet. This does not tend to help mothers breastfeed successfully. I worry that the headlines from this study — such as “How Formula Could Increase Breastfeeding Rates (TIME)”  and  “How Formula Can Complement Breastfeeding (NYT)” —  will translate into “a six pack of formula back in every bassinet!”

It’s also notable that there were more multiparous women in the ELF group than in the control group (70 vs 50%)— and multiparty was a powerful predictor of exclusive breastfeeding. At three months, 78% of multiparous moms were exclusively breastfeeding, vs. 33% of primiparous moms. Based on these numbers, if you treated both groups exactly the same way, you’d expect 65% of the ELF moms to be exclusively breast-feeding at 3 months (.70 * .78 + .30 * .33 = .65), compared with 55% of the control group moms (.50 * .78 + .50 * .33). In a larger study, one would adjust for this difference statistically, but this sample was too small to allow for such adjustment.

In a future study, researchers could avoid this issue by randomizing first-time moms and experienced moms separately, ensuring equal numbers in each intervention group. Bigger numbers would also increase the precision of the estimated difference between the two groups. The authors report that 15/19 moms in the ELF group were exclusively breastfeeding at 3 months, vs. 8/19 in the control group. Those are very small numbers. Intuitively, 150/190 is more precise than 15/19 — and if the authors reported an exact confidence interval, they would have reported that 15/19 is 79%, but the precision on that estimate (the 95% CI) ranges from 54-94 percent. For the control group, 8/19 = 42%, with a 95% CI of 20-67 percent. So it’s possible that the real difference between the groups is much smaller — or much bigger — than the one described in the paper.  A larger study would improve that precision.

I wonder as well about 5% weight loss as the threshold for intervention. For a pilot study, the goal is to find out whether it’s feasible to enroll patients, so researchers tend to use fairly loose inclusion criteria. In fact, it may be that waiting a bit longer, or for a higher threshold for weight loss, better identifies the dyads who need a bridge. I’d be hesitant to use this study to change practice to “syringe feed every baby down 5%”

A larger study would also test whether the intervention would work the same way “in real life.” A study doctor or nurse administered the intervention, staying at the patient’s bedside to teach her how to use the feeding syringe or use the soothing techniques. Study personnel are, by definition, deeply invested in doing the intervention in a way that ensures continued breastfeeding, and probably chose their words very carefully so as to maintain mom’s confidence in her ability to breastfeed. For this approach to successfully scale to a non-study setting, the authors will need to define the language, context, and training that they used and teach it to other care providers so that it achieve the desired outcome.

I also wonder a bit about the potential for bias during the control intervention, in which study physicians or nurses taught mothers about infant soothing techniques. The people doing this piece of the intervention did this study because they thought that ELF might be beneficial. In the course of a 15-minute conversation with the control mother about soothing a hungry baby without supplementing, they may have telegraphed that belief to the mother, possibly making her more likely to give up breastfeeding than she might have had she never been in the study at all. This is a challenge in any randomized trial — you have to give the control group “something,” and it’s always possible that that “something” affects the outcome.

Future studies should also explore the effect of offering expensive, hypo-allergenic formula to infants, compared with using donor milk. The rationale for using the hydrolyzed formula is that this would reduce the risk of formula exposure for the infant, and this is supported by the TRIGR study, which found that infants at high risk of Type I diabetes who received regular formula were more likely to develop islet cell antibodies than infants who received hydrolyzed formula. However, having a health care provider recommend hydrolyzed formula could lead parents to continue buying this product long-term. I wonder about whether this approach could be co-opted as a marketing strategy for high-priced formula. I could imagine formula manufacturers selling “syringe-ready” feedings for newborn nurseries branded with their logo and product line. Using donor milk would obviate this issue entirely.

Finally, one of the underlying questions in this type of research is why it’s so common for babies to lose more weight than expected in the first few days of life. Some of this may reflect obstetrical practices — there are a few studies that suggest when moms large volumes of IV fluid during labor, some of the extra water goes to their babies. That extra water registers as an exaggerated birthweight, and as they pee off the extra fluid, they appear to lose too much weight, when they actually are doing just fine.

Laurie Nommsen-Rivers has done some ground-breaking work looking at the prevalence of delayed lactogenesis — i.e. Milk coming in more than 72 hours postpartum — and found that this is incredibly common in the U.S., but not in other parts of the world. She posted the following comment on a blog I wrote last fall:

In our prospective study of 431 first-time mothers residing in Sacramento, CA, 44% experienced onset of stage II lactogenesis after 72 hours, but only 1.7% still had not experienced stage II lactogenesis at the end of the first week postpartum (Nommsen-Rivers, et al, Am J Clin Nutr, 2010). In contrast, a companion study lead by Susana Matias in peri-urban Peru found that only 17% of first-time mothers experienced onset of stage II lactogenesis after 72 hours and NONE were still waiting for their milk to come in after 1 week (Matias, et al, Mat Child Nutr, 2009). Another interesting contrast comes from rural Ghana: in a study published by Otoo, et al (JHL), in which <5% of a cohort delivering at a baby-friendly hospital in rural Ghana experienced onset of stage II lactogenesis beyond 72 hours. All of these studies used similar, validated measures of stage II lactogenesis.

One of the key risk factors for US mothers in her study was maternal body mass index, with 54% of women with a BMI > 30 experiencing delayed onset of lactogenesis, compared with 31% of normal BMI women. Older maternal age and higher education were other risk factors. A central question for future research is why so many American mothers start making milk so much later than women in other parts of the world.

The bottom line is that this is a complex issue.  We know that the onset of mature milk production comes later in American mothers than in women in other parts of the world, and we need to develop and test strategies that can support long-term breastfeeding success. This small pilot study suggests that bridging mothers with a syringe of a breast milk substitute may ultimately help them establish breastfeeding.  The next step is to replicate these findings, ideally using donor human milk, to test whether these results hold up in a larger sample.

Alison Stuebe, MD, MSc, is a maternal-fetal medicine physician, breastfeeding researcher, and assistant professor of Obstetrics and Gynecology at the University of North Carolina School of Medicine. She is a member of the board of the Academy of Breastfeeding Medicine.

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