(SBM) [11]. This difference could occur, for instance, if expression of the breast removed more high-fat hind milk than would be obtained by the baby if the breast was not fully emptied during feeding.
In order to study SBM, a milk sampling system was devised by modifying a clinical nipple shield worn on the breast during breastfeeding. The modified nipple shield contained a milk sampling line so that milk could be sampled continuously during a breastfeed, and it also contained a flowmeter in the tip. Initial research using the nipple shield sampling system showed that SBM fat content was around 2.5 g/100 mL versus a figure of around 4.0 g/100 mL obtained in a vast number of prior studies on EBM composition [8, 11]. Thus, if valid, our data suggested that using EBM, it was possible to overestimate milk fat content by 60% compared to SBM obtained during normal feeding. We estimated the energy content of SBM to be 58 kcal/100 mL compared to around 71 kcal/100 mL based on over 1,500 prior publications. This would equate to a methodological error in measuring milk energy content of over 20% when studying EBM versus SBM.
When these data on SBM were published, they were too radically different to those published previously using EBM to be widely accepted. So, to confirm our findings on the energy content of breast milk, we used the doubly labeled water method in a novel way. In this method, 2 naturally occurring stable isotopes: deuterium (heavy hydrogen, 2H) and heavy oxygen (18O) are given orally producing enrichment of these isotopes in body fluids. Decline in these isotope enrichments back towards baseline is measured in urine or saliva over several days. The slope of the decline in 2H can be used to measure water output (since hydrogen is lost as water), which in steady state reflects water intake, and, from this, milk volume intake can be derived. The decline in 18O is faster since oxygen can be lost in both water and carbon dioxide. Hence, the difference in the decline in 2H and 18O is the CO2 production rate, from which energy expenditure can be derived – and hence metabolizable energy intake. Thus, over several representative days, milk volume intake and energy intake can be estimated, and, by dividing the latter by the former, the energy content of breast milk is derived without any recourse to breast milk sampling. This approach produced values for energy content of breast milk according to postnatal age of 57–61 kcal/100 mL, thus confirming our previous values using the nipple shield system [12]. Later work has confirmed our finding that breast milk energy content had been greatly overestimated in past EBM studies.
One importance of these findings is that formula manufacturers based their products, and still do, on the composition of EBM, which emerges as the wrong model.
In addition to errors in prior estimation of breast milk fat and energy, breast milk protein content was also overestimated by using analytic methodology developed by the dairy industry. In cow’s milk (CM), there is little nitrogen that is not part of protein so that it is possible to estimate protein content by multiplying nitrogen content by a constant (6.38) [13]. This was used inappropriately for human breast milk in which there is a high content of nonprotein nitrogen (e.g., urea), which should not be counted as protein. Thus, more recent work shows that the true protein content of breast milk is significantly lower than previously thought [14], and this is one reason why infant formulas have substantially higher protein content than breast milk.
Thus, as a result of flawed methodology for breast milk compositional analysis, a generation of babies were fed on formulas modeled on EBM composition unphysiologically high in fat, energy, and protein. This constituted in an inadvertent experiment in early overfeeding, and animal studies since the 1960s show early overfeeding increases later cardiovascular risk factors [15].
The higher nutrient intake of the formula-fed infants is believed to be a major factor in the faster early growth of formula- rather than breastfed infants. So, does it matter that formula-fed babies grow faster? In 2004, based on our nutritional intervention trials and animal evidence, we published our postnatal growth acceleration hypothesis, which proposes that faster early growth increases the risk for later obesity and CVD [15]. In that publication, the known increased risk of obesity and cardiovascular risk markers with formula feeding was proposed to relate to the faster growth rate. Since then, over 60 studies, including randomized trials, have supported the postnatal growth acceleration hypothesis.
Thus, flaws in research on breast milk composition were indirectly partly responsible for the major modern epidemic of CVD and obesity – a salutary example of the importance of methodology in science. The field has now become a priority for research on both breastfeeding and formula feeding.
The Benefits of Breastfeeding Revisited
Arguably, the main platform for the global promotion of breastfeeding is the scientific evidence for its clinical benefits. However, with few exceptions, the comparison of breast- and formula-fed babies has not been based on randomized trials that would prove causation, but rather on observational associations.
Initially, the main outcomes of interest were infection and cognition, but these outcomes are potentially highly confounded by the differences in the populations (statistically) that choose to breastfeed or formula feed. As an example, cognitive benefits in breastfed babies have been described in a number of studies since 1929, but in 2006, Der et al. [16] concluded from a meta-analysis and study of sibling pairs that there was no cognitive benefit due to breastfeeding, and the previous positive findings were explained by the higher maternal IQ in those who chose to breastfeed. This study emphasizes the ever-present potential for confounding in epidemiological studies where there are major demographic differences between the groups compared, though the study by Der et al. [16] was also nonrandomized.
Today, a wide variety of beneficial outcomes has been linked beneficially to breastfeeding [17], including CVD and obesity risk, atopic disease, IQ, brain size, infection, cancer, sudden infant death, celiac disease, and type I and II diabetes – but again these beneficial outcomes have only been epidemiologically associated with breastfeeding and not determined experimentally, leaving uncertainty over causation.
The challenge then is how better-quality evidence can be obtained, given the constraint that randomized trials, for instance comparing the outcome of breastfeeding versus formula feeding, are generally precluded on ethical grounds.
The Preterm Infant as a Model
The area I shall focus on here is the use of the preterm infant as a model. Whilst accepting that the spectrum of diseases and the sensitivity to early nutrition is somewhat different in preterm and term infants, neonatal care is an area where it has been ethically possible to conduct numerous strictly randomized trials of EHM feeding versus exposure to CM. My argument is that if a wide range of important outcomes in preterm human infants are favorably impacted by HM feeding, this would indicate that the weaker observational data on the benefits of breastfeeding in full-term infants are more likely to be causal – especially when the same outcomes (e.g., infection, allergy, cardiovascular risk, or cognitive development) can be studied in both the preterm and term populations.
Preterm Trials Comparing Exclusive Human Milk Feeding versus Exposure to Cow’s Milk
There are 3 categories of randomized controlled trials (RCTs) that provide evidence on the benefits of HM or adverse impact of CM.