To drink cows milk or not to drink cows milk – that is the question

 

Righty-ho so let’s break down this CNN article by Wayne Drash which claims “Drinking non-cow milk linked to shorter kids, study suggests”.  The plan here is to assess some research about a very specific effect of cows milk on the height of young children so I’m not going to go into a bunch of other stuff around cows milk – just addressing the points of this research.

Remember from this article I wrote on “How to tell if that new research study your friend posted on Facebook should be shared or deleted” these were the red flags we needed to look out for:

  1. A self reported behaviour survey – especially one where the participant is being asked to recall something they did a long time ago
  2. The lack of a control group, test/treatment group (and in some cases) a placebo group
  3. Non randomised
  4. Non Blinded or Non Double Blinded
  5. Small sample sizes
  6. Non peer reviewed – i.e published in a dodgy journal
  7. Statistical significance without IMPORTANCE

So let’s review.  Firstly, we have to look BEYOND the article to the research itself.  The only link the article had was to the American Journal of Nutrition – there was no original source link but I found the original research paper by searching “Jonathon Maguire non-cow milk” (the name of the researcher and a keyword on the research topic) on trusty ol’ Google.

This is the link to the abstract for the research in the American Journal of Clinical Nutrition: http://ajcn.nutrition.org/content/early/2017/06/07/ajcn.117.156877.abstract

Sometimes you can take the full title of the abstract and do another Google search and find the full version posted somewhere (e.g. Association between noncow milk beverage consumption and childhood).  Very often I can find it on Researchgate which I have access to. Unfortunately I tried that in this case and it hasn’t been around long enough for any other sources to be hosting the full version.  If you’re a uni student, your uni email should give you access to a whole host of credible journals and so you’ll generally be able to access the full version.

Ok so for now we’ll have to work with what we’ve got – the research abstract.

  • Self reported behaviour survey: YES
    So we need to approach with caution.  The abstract says its a cross sectional study of kids enrolled in an existing research program called the “Applied Research Group For Kids”.  I looked that up and found this: https://www.ncbi.nlm.nih.gov/pubmed/24982016.  Looks pretty solid and as a longitudinal study, they’re basically asking a bunch of questions on a regular basis so participants aren’t having to remember what they did ages ago.  The program is also listed on the US National Library of MedicineNational Institutes of Health so I feel pretty safe about this.
  • Lack of control, test and placebo group: N/A
    Not really relevant here as its a longitudinal study which is just collecting a bunch of different data over time.
  • Non randomised: N/A
    As above, not relevant per above
  • Non Blinded or Non Double Blinded: N/A
    Not relevant as participants were not assigned to control, treatment or placebo groups because they weren’t testing any particular treatment
  • Small Sample Sizes: NO
    The sample size is 5,034.  This sample size is ok but I would use caution in applying these results more broadly because these are 24-72 month old Canadian kids, the majority of them are of caucasian background. Now given that 75 percent of African Americans and American Indians and 90 percent of Asian Americans are lactose intolerant – lactose intolerance develops in Asian and African genetic heritage in much higher rates than caucasian – this study may not be applicable in those cases. As a bit of background to lactose tolerance/intolerance, basically when you’re a baby you have a bunch of enzymes called “Lactase” which essentially helps your body to break down the sugar molecules in milk called “Lactose” and some of us don’t keep producing those enzymes once we’re done with breastfeeding! (some more info on this here).
  • Non Peer Reviewed: NO
    Nope this was published in a good journal.  Here are the results for the journal’s credibility.
  • Statistical Significance Without Importance: YES
    This is really the biggest problem with this whole study.  This is the assumption that the researchers have made “Cow milk consumption in childhood has been associated with increased height, which is an important measure of children’s growth and development”.

    That is true, but it’s also a misleading statement because they are not defining how much height is good.  Saying that “height is an important measure of child development ipso facto a taller child is a healthier child” is a fallacy.  This is like saying “Vitamin A is good for me so more is better”.  Well, that’s not true.  Too much Vitamin A can cause dizziness, nausea, headaches, coma and oh yes…death.Now I’m sure these researchers have good intentions, but they seem to have ignored previous research as a baseline or benchmark for their assumptions.

    Sure, western society holds up males in particular as being “better” for being taller.  But does that mean they’re healthier? Nope. All evidence points to shorter being healthier. This is info from the main US govt health site on this topic https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1071721/.  This overview of research to date on the topic is well worth the 15 minute read if you’ve got the time. Seriously, don’t take my word for it.

    Furthermore height is the main mediator of higher risk of hip fractures later in life for men – and is determined by their earlier intake of milk.  Here is the original study abstract for the above statement here: https://www.ncbi.nlm.nih.gov/m/pubmed/24247817/.

    The concern I have is how the researchers of the kids and nonmilk study choose to describe the background to their research:
    “Many parents are choosing noncow milk beverages such as soy and almond milk because of perceived health benefits. However, noncow milk contains less protein and fat than cow milk and may not have the same effect on height.”

    Is it true that soy and almond milk have perceived health benefits (that may, or may not be true depending on what research has shown vs what is widely believed). Yes that’s pretty true.

    And is it true that soy milk contains less protein than cows milk?  Let’s take a look…

  • 1 cup of Soy milk protein: 8 grams
  • 1 cup of Cow milk protein: 8 grams
  • 1 cup of Almond milk protein: 1 gram

    Gee…that doesn’t seem right…have they averaged out protein across Soy and Almond Milk perhaps to give “nonmilk” drinks an overall lower protein number? Because by my lay-person eye it looks to me as if they have the same amount of protein.  Of course milk may have other nutrient soy does not have and I’m not disputing that but why say something when it’s not true across the board?
    Surely you’d make sure you were looking specifically at who drank almond milk vs soy milk as a non milk alternative and understand the difference between both.
    Let’s ask a more important question.  Is it fat OR protein that contribute to height?
    Erm, neither exactly. Protein is part of it, but it’s actually calcium as well as the Insulin-like Human Growth Factor hormone that is present in cows milk at a much higher dose than that of human breast milk which is thought to contribute to height gain.

    And lets as a final and important question.  If western societies continue to promote health through consuming animal products through infancy through adolecence when more than than half of the world’s population is intolerant, what are we saying?  That only those who are lactose tolerant can be truly healthy? I’d love to see more studies that divide people into three groups:

  • Group 1: Conventional treatment (e.g. milk)
  • Group 2: Plant based supplement that meets all the same / similar nutritional properties that are thought to affect the variable (e.g. perhaps this is soy + other factors not present in soy as fortification)
  • Group 3: None or “on the market” options

Why are we only doing studies that suit a caucasian genetic makeup?

So what does all the above mean for this study ?
Well, the findings of association itself are not wrong.  But the assumption that is being created in the way the introduction is phrased are disturbingly misleading.  What should happen next?

As the abstract itself states at the end “Future research is needed to understand the causal relations between noncow milk consumption and height”.

This is absolutely true because milk consumption (or the specific properties within it) may not be the cause of different height growth.  Epigentic forces could be contributing to the outcome here.  i.e. mothers and fathers who drink noncow milk and plan to give noncow milk to their children could have a variety of other different diet habits that have a different impact at the point of conception and methylation.

Recent epigenetic research shows for instance that mothers who take less than around 300 grams of protein per day during the early stages of pregnancy alter the DNA methylation status at the site of the Vitamin E receptor gene and this contributes to the child’s percentage fat mass later in life.  I wonder, does it impact their height too?  Because if it does, this could be a critical factor in helping to explain this Canadian research as having an epigenetic cause.  It is possible that women who choose to drink noncow milk, may also have a lower protein intake in general which in turn contributes to these epigenetic changes upon conception!  PS that is just a completely untested hypothesis of mine…but it’s not entirely crazy I don’t think..

But furthermore, future studies should first seek to establish what is considered to be a bottom level AND top level healthy height in childhood through to adulthood.  And ensure that studies around the benefits for shorter adult height are fully considered in this assessment.

Only then is it really possible to start drawing insight around what types of foods/drinks to give children and give them the best chance of health throughout their lives.

—-

Lastly, while trying to find images for this post I came across the cutest awkward cow toy I have ever seen. 😀  I think I’m going to buy one!

Advertisements

How to tell if that new research study your friend posted on Facebook should be shared…or deleted.

Last week CNN posted an article with the title “Drinking non-cow milk linked to shorter kids, study suggests“.

Let me start by saying, that this statement is in fact correct…but not for all intents and purposes.  What does that mean?  Let’s start at the beginning.

In the media model, article views equals the ability to sell more advertising space which equals revenue for shareholders.  In the science model, replication of a well designed study by other well designed studies and producing the same result (a process that can take years, sometimes decades!) equals an answer that may then be worthy of writing an article about. These organisational models are living in parallel universes where time between them runs at different speeds.  And this is a shame for the consumer.

Imagine a world where as a consumer, you had access to an instant meter of how valid the results of any research study was according to some universally accepted scoring criteria so you weren’t at risk of consuming erm…trash.

Before you read another click-baiting, crowd-pleasing, over-shared, under-researched article, I’m going to jot down a few things for you to take note of, or to take a few extra minutes to research after you read any article reporting on a new scientific finding.  I challenge you NOT to either share the research nor commit the findings to memory until you’ve availed yourself of the facts surrounding the research design.

The most important question to ask is: Did the research study control for confounding variables?


A confounding variable happens when a researcher can’t tell the difference between the effects of different factors on a variable.  There are so many different things that can have an impact on the results of the study and so understanding what “data noise” to remove is critical to making sure that pattern in the data that a researcher might see is unlikely to be due to chance alone.

When reviewing the validity of research results, these are some of the red flags when it comes to research design:

    • Self reported behaviour surveys
      Humans can barely remember what they did on the weekend let alone on a daily or weekly basis five years ago!  That’s not to say that these studies aren’t relevant, simply that the evidence from them would not be considered as strong as say a study where the experimental design had people follow a pattern of behaviour (for the control and placebo groups) across a specific period of time and followed up with them regularly for self reporting across that time period.
    • The lack of a control group, test group (and in some cases) a placebo group
      A control group is a group of participants to whom the treatment isn’t applied, in the test group the variable that the researcher wants to test is introduced and in the placebo group, the participants think they are part of the test group, but they are receiving some sort of alternative to the treatment that will not yield the expected result.  The human brain and body are pretty powerful…when we think we’re getting we can actually experience improvements that don’t really exist!

      However, a placebo group is not always feasible depending on what is being tested so a bit of common sense needs to be applied here. For instance, if you were trying to test some sort of effect related to drinking water, you could have a control group who didn’t drink water, but given everyone knows what water tastes like, attempting to create a “water placebo” would be pretty tough. But a control and test group should be the bare minimum!  And in cases of medication where a placebo can be easily applied, there should ALWAYS be a placebo group.

    • Non randomised
      If the research is experimental in nature (and not survey based), and the report doesn’t say it’s randomised, then it probably isn’t.  A randomised experiment means that participants in the experiment (those put into either control, test or placebo groups) were randomly assigned assigned to them.  i.e. that the researchers weren’t in control of choosing who would be assigned to which group.  If they are, they can introduce all sorts of unconscious bias that could affect the results.
    • Non Blinded or Non Double Blinded
      A blinded study is where the participant in the research is unaware which group (i.e. test, control or placebo) they have been assigned to.  A double blind study is where neither the participant in the research, nor the researcher themselves, knows which group the participants have been assigned to.  That means the researcher might only see a number in place of an individual’s name and details when seeing the results. And the experiment may be designed in such a way that those responsible for data collection, and perhaps physically collecting data from the participants, do not communicate with the researcher (or may not even be known to them).
    • Small sample sizes
      A “sample” is basically a little portion of the broader “population”.  A population in research doesn’t have to mean the population of a country, it may just be the population within a particular category pertinent to the research.  For example “people with Diabetes”, or “people who have been treated for depression”, or “women who have given birth to at least one child”.  The sample size has, because there are random effects that can occur in small samples that even themselves out when you test the same thing on a larger sample size.

      Most good research might start with a smaller sample size to test an initial hypothesis (theory).  They’ll release initial results but caution that due to small sample sizes, more research should be done to see if the results can be replicated on a broader scale so that it can be .  If this is the first research in a particular area to come out and it’s got a small sample size you MUST treat it with caution. It means that it is essentially “baby research” it’s not fully formed yet nor capable of making truly informed conclusions about its own existence!

 

  • Non peer reviewed – i.e published in a dodgy journal 
    Yep, not all journals are created equal.  A good piece of research should be published by a journal that has a process whereby other scientific peers review the research methodology before accepting it for publishing.  Sometimes good journals will create a single-blind process for doing this – meaning that those reviewing the research don’t know who the author is.  That’s important – because humans have an innate bias to trust people who are perceived to be more credible, despite there potentially being a lack of evidence to support that trust.

    If it has been archived or cited here: https://www.ncbi.nlm.nih.gov that’s a good start.  Apparently this tool https://harzing.com/resources/publish-or-perish helps you figure out how many times the article has been cited in journals (although I’ve not used it before) and this one helps you figure out the ranking of the journal: http://www.scimagojr.com/journalrank.php?area=2700&type=j.

    Monash Uni have a bunch of good links and info about assessing journal quality here including:

  • Statistical significance without IMPORTANCE
    Statistical significance is generally agreed that there is either a 5% or lower (sometimes 1% for more rigorous research) probability that the results obtained were due to chance versus the variable being tested.

    That’s a great first step for sure, but significance does not mean importance. Once the study has met the above criteria, ask yourself one, final and very important question “Is this question the right question to be asking, and is the assumption that underpins the question being asked a correct assumption?”

 

In the next article I’ll use this cnn article to test drive some of the knowledge above.