Pass on ultra-pasteurization?

A friend of mine recently asked for advice on introducing milk to her one year old. Specifically, she had heard that ultra-pasteurized milk was less nutritious than (regular) pasteurized milk; the friends who told her this advocated introducing organic, pasteurized milk instead. Spoiler alert: a lot of organic milk is ultra-pasteurized, so advocating organic milk that is not ultra-pasteurized is a little bit of a contradiction. Anyway, my science-minded friend did a quick search and couldn't find any evidence comparing pasteurized to ultra-pasteurized milk, so I promised her that I would look into it.

First, we need to know what the difference is between the process of pasteurization versus ultra-pasteurization. According to the FDA, pasteurization involves heating milk to a certain temperature and maintaining it at that temperature for a designated amount of time; for example, milk could be heated to 72 degrees Celsius for 15 seconds, to 89 degrees Celsius for 1 second, etc. This process is intended to kill contaminating pathogens, such as E. coli and salmonella. Ultrapasteurization involves heading milk to at least 138 degrees Celsius for at least 2 seconds; this results in extended shelf-life and even allows milk to be stored without refrigeration for months. Ultrapasteurization is common in Europe, where the milk in grocery stores is not kept refrigerated. According to SABIC (cited in Oupadissakoon, 2007) ultra-high temperature milk accounts for 54% of the market in Europe compared to 41% for pasteurized milk. In the U.S., ultra-pasteurized milk is not common and mainly exists in the form of organic milk (which is not marketed as ultra-pasteurized even though it often is).

One of the first studies that I came across in my search seemed to directly address the very question that my friend was asking. AlHanhal, Al-Othman, and Hewedi (2001) report that ultra-high temperature milk treatment results in loss of nutrients like the amino acid lysine. [Side note: this is the same amino acid that was manipulated in Michael Crichton's Jurassic Park dinosaurs.] However, according to these authors, some studies show that lysine loss is relatively small unless milk is stored at high temperatures for a period of months, and others report that effects of ultra-pasteurization on vitamin content is negligible. AlHanhal and colleagues designed an experiment with rats to specifically compare the nutritional value of ultra-high temperature milk and pasteurized milk. Their ultra-high temperature milk was heated to 137 degrees Celsius for 4 seconds whereas the pasteurized milk was heated to 73 degrees for 15 seconds. They found that nutritional quality of the protein (measured by apparent digestibility, true digestibility, biological value, and net protein utilization) in the ultra-pasteurized milk was the same as that in the pasteurized milk; however, two of the four nutritional quality measures decreased in the ultra-pasteurized milk after storage for 3 and 6 months.

Looking at the data, however, raises some concerns for me. To understand those concerns, you need to know something about data variability. In a randomized experiment, different people (or rats, in this case) are assigned to different conditions. In this particular experiment, rats were assigned to receive either pasteurized or ultra-pasteurized milk; additionally, some rats in the ultra-pasteurized condition received milk that was fresh whereas others received milk that had been stored for 3 or 6 months. Now, not every rat in a single group will have the exact same levels of nutritional quality even though they are receiving the same "treatment." In other words, each rat in the pasteurized milk condition will have a slightly different nutritional quality value from the other rats in that group; this is the concept of variability. The more different a rat's value is from the other rats in its same group, the higher the variability of that group. This variability is captured by a statistic called the standard deviation. Standard deviation tells us how much a group member differs from the others in the group on average. When we try to see whether groups are different from one another (i.e. is the pasteurized group different from the ultra-pasteurized group), we compare whether the variability between the groups (i.e. the amount that ultra-pasteurization rats differ from the pasteurization rats) is larger than the variability within the groups (i.e. the amount that all the rats in the ultra-pasteurization group differ from one another). In order to do that statistical test, we have to make some assumptions about the data; if those assumptions are not met, then running that particular statistical test would not be valid. One of the assumptions is called "homogeneity of variance", which is a fancy way of saying that we assume the variability within all of our groups is the same; in other words, the variability of the pasteurization rats should be the same as the variability of the ultra-pasteurization rats. If we look at the data in the article, this assumption appears to be violated, and the researchers make no mention of checking for homogeneity of variance.

How can I tell that the assumption might be violated? I look at the numbers the researchers present, which include the averages for each group as well as the standard deviations; I'm choosing just one of the nutritional quality measures (net protein utilization) to illustrate this point. In the fresh conditions, the average for pasteurized was .80 with a standard deviation of .05 whereas the average for ultra was .81 with a standard deviation of .01; these are not statistically different from one another and the standard deviations are similar (so the homogeneity of variance assumption is met). But researchers only found differences in nutritional quality between pasteurized and ultra in the 3 and 6 month storage conditions, so let's look at those numbers. In the 6 month condition, net protein utilization for pasteurized milk was .80 with a standard deviation of .02 whereas the average for the ultra-pasteurized group was .75 with a standard deviation of .17; the fact that .02 is very different from .17 is what makes me suspect that the homogeneity of variance assumption was violated in the stored conditions, which means that I'm not putting much stock in these results.

Finally, the authors conclude that the "reduction in nutritional quality of protein in ultra-high temperature milk stored at high temperature...may be of a significant value in feeding young children who might be dependent on this type of milk in their diet." It is a very large leap to take a 4-5% decrease in two nutrition measures in rats and generalize it to advice for giving milk to children! A single study like this, in my opinion, does not warrant any change in societal behavior. First of all, it would be necessary to demonstrate that changes in these same nutritional measures observed in rats actually take place in humans as well. Second, it would be important to consider whether decreases in these specific nutritional measures have any observable impact; in other words, what implications does a reduction in biological value and net protein utilization have, if any?

So, it appears that whatever changes in milk protein occur during the ultra-pasteurization process do not translate to differences in nutrition; the only potential difference emerged after 3-6 months of storage for the ultra-pasteurized milk, and even that difference is called into question by sketchy statistical analysis. Ideally I'd like to base by conclusions on multiple studies, but this is the only one that I found directly comparing pasteurized to ultra-pasteurized milk in terms of nutrition in animals.

Even if there is no nutritional difference in pasteurized vs. ultra-pasteurized milk, other factors like taste may influence whether parents wish to give ultra-pasteurized milk to their children. Anecdotally, the ultra-pasteurization process does result in milk that tastes different from standard pasteurization; I can speak to this from experience drinking ultra-pasteurized milk in Europe. Additionally, Oupadissakoon (2007) reports that American consumers (who are accustomed to pasteurized milk rather than ultra-pasteurized) give lower taste ratings to ultra-pasteurized milk compared to Thai consumers (who consume ultra-pasteurized milk on a regular basis). So, are American children more likely to prefer pasteurized milk to ultra-pasteurized milk? A study by Chapman and Boor (2001) attempts to answer this question. They tested 6-11 year old children from New York and found that children did indeed give higher liking ratings to pasteurized milk (6.0 on a 7-point scale) compared to ultra-pasteurized milk (4.7 on a 7-point scale).

So, at the end of the day, what should my friend do when it comes to giving her child pasteurized vs. ultra-pasteurized milk? Well, there doesn't seem to be any evidence that ultra-pasteurized milk is less nutritious (at least in any meaningful way) compared to pasteurized milk. However, it is possible that a child would prefer the taste of pasteurized to ultra-pasteurized milk, thus making him/her more likely to accept that milk and drink it regularly.