I just did a little googling and that appears to be untrue. What's the source of your information?

The first hits from a Google search for <<random controlled trial breast formula milk>> were:

• Scholarly articles (Google pulls out a few of these at the top of its search results):
  • A study on promotion of breastfeeding in Belarus. Intervention appears to have increased breastfeeding and also substantially (and significantly) reduced the incidence of two of the three health problems they measured.
  • A study on how often breastfeeding passes on HIV from mother to child. (Conclusion: about 1/6 of the time; using formula is an effective way of reducing this. Interesting but not relevant here.)
  • A study on the effect of "peer support" on making mothers breast-feed for longer. Not relevant here.
• Ordinary search results:
  • First five hits are for the same study, looking at very premature babies. Conclusion was that mother's milk appears better than both donor breast milk and formula, and the latter two aren't much different.
  • Meta-analysis of studies comparing donor breast milk with formula. Conclusion is that unfortified donor breast milk (i.e., no extra nutrients added) leads to slower growth than formula, but formula leads to more necrotizing enterocolitis. Apparently donor breast milk is generally fortified nowadays.
  • Different meta-analysis looking specifically at necrotizing enterocolitis. Found only four relevant trials, from 20 years before. They were too low-powered to find significant results, but aggregating them finds a substantial and significant reduction in NEC risk for donor breast milk versus formula.
  • CBS News report on a study that found that giving a little formula to babies who lose a lot of weight shortly after birth may result in longer breast-feeding. Interesting but not obviously relevant.

And that's the end of the first page. So it looks to me as if

• There haven't been that many randomized controlled experiments -- fully half the first page of Google results were from a single one, and the meta-analyses seem to have found rather few, rather old, rather small studies.
• Some small old studies didn't find significant results.
  • But that seems to have been because of lack of power rather than because differences weren't there.
  • Aggregating those studies finds significant differences.
• There may be relevant differences between mother's milk and donor milk; comparing mother's milk to formula in RCTs is probably harder than comparing donor milk to formula. (Because of ethical difficulties in telling mothers not to breast-feed their children; because of practical difficulties telling mothers to breast-feed their children if they have trouble with that; because of difficulties in blinding. Maybe other reasons.)

Which seems like pretty much the reverse of what you say. But of course looking briefly at one page of Google results is not a proper scholarly literature review; would you like to tell us more?

http://www.bayesianrisk.com/chapters.html http://www.bayesianrisk.com/sample_chapters/Chapter%201%20There%20is%20more%20to%20assessing%20risk%20than%20statistics.pdf

I agree with Viliam & gym. This just points to the limits of statistical knowledge & that we need to supplement with other logical-experimental knowledge, such as arguments from evolution.

Risk cannot only be based on statistical knowledge, as chapter one of bayesian risk argues.