A newly published RCT seems at first glance to challenge the overall verdict of the literature that consuming breast milk improves intellectual development, particularly for preterm infants. (O’Connor DL, et al. Effect of Supplemental Donor Human Milk Compared With Preterm Formula on Neurodevelopment of Very Low-Birth-Weight Infants at 18 Months: A Randomized Clinical Trial. JAMA. 2016;316(18):1897-905).

180 infants were randomized to receive preterm formula whenever there was insufficient maternal breast milk. The other group of 180 received pasteurized donor breast milk whenever there was not enough breast milk. Infants received their assigned supplement until 90 days or discharge, whichever occurred first, and continued to get the assigned diet if they were back-transferred to a level 2 nursery before going home.

The study was well-designed and carried out, but unfortunately vastly underpowered. All NICUs now try and give mother’s milk to the babies, which is always the first choice, so nearly 30% of the babies in each group never received a supplement, which makes them uninformative for the primary outcome, which was Bayley Testing at 18 months of age.

Of those that did receive a supplement, the median proportion of total enteral feeds for each infant consumed as mother’s milk  was 58.4% [IQR, 13.6%-96.0%]  for the donor milk group vs 63.3% [IQR, 9.6%-97.2%] for the formula group. So in other words some of the babies received as little as 3% of their feeds as the assigned diet, and about half received around 40% or more.

I think its fairly obvious that receiving only a tiny proportion of your feeds as formula is highly unlikely to have a measurable effect on developmental delay. The only babies that you might think could possibly have an effect would be those with a significant percentage of their milk being donor milk, rather than formula.

As you can see from those figures above, half of those who did get the assigned diet received more than 40% of their feeds as something other than mother’s milk. We could guess that 40% might be enough to show an effect if there was one, if we use that as the threshold, there were only approximately 60 babies per group, of whom 92% had follow up, or about 55. The power of detecting an effect on development with only 55 per group is rather low.

I understand the need to do intention to treat analysis, but in a study such as this, where it was expected that 30% of the babies would never receive the assigned intervention, (and not because of protocol deviations, but because the mother is able to produce enough milk) you could easily argue that the main analysis should be among those who did actually need a supplement.

This is not the same as, for example a study comparing planned vaginal breech delivery to planned cesarean, where the clinical question is “what should I plan for the delivery of this baby?” The clinical question here was, “if we need to supplement, what should we do it with?” and it is a question that can be posed when you get to the point of there not being enough breast milk. Enrolling mothers and getting consent could be done soon after birth, but actually randomizing and collecting data could be done only when the baby needs a supplement, which would increase the power of the study.

As it is, there are a limited number of babies in the trial who are informative for the outcome, which means that a substantial benefit (or harm) of supplementation with donor milk, on the primary outcome, could have been missed.

Secondary outcomes are also underpowered, but even with this lack of power, there was a significant reduction in NEC, from 6.6% (stage 2 or more) with formula supplementation to 1.7%, with donor milk as a supplement. Other outcomes including late onset sepsis were not different between groups.

There are all sorts of goodies in human milk for preterm babies, and this study confirms that NEC is less frequent when you supplement with donor milk compared to preterm formula. It wouldn’t be surprising that the impact on NEC required less cow’s milk than an impact on developmental delay, so a large relative difference between groups, for this outcome, is biologically plausible.

Also of note, most of the babies in this study were exposed to cow’s milk protein. Both in the formula group, but also in the breast milk groups, as milk was fortified with cow’s milk based breast milk fortifier, even if the baby only received their own mother’s milk.

The conclusions state the following:

Results from the present study suggest no advantage of feeding nutrient-enriched donor milk compared with preterm formula, as a supplement to mother’s milk, on neurodevelopment of VLBW infants at 18 months’ corrected age.

But that isn’t quite right, the study suggests no advantage of a strategy of being ready to receive nutrient enriched donor milk if needed, compared to a strategy of getting formula if needed. That comparison showed no overall impact on Bayley scores, but did show a benefit in terms of NEC.

The results are consistent with a substantial effect of breast milk on neuro-development, (as well as being consistent with no effect or with a negative effect).

To be honest, I don’t think we need to do more studies like this, concentrating on helping mothers to express their milk, and when necessary supplementing with donor milk, should be the standard of care. This study confirms the benefits in terms of NEC, and didn’t show any downside of breast milk supply to preterm babies.