Thursday, February 25, 2016

The Paleo Mouse Study

The media have recently been publicising a new study, claiming that a Paleo diet causes weight gain, glucose intolerance and insulin resistance (for example [1, 2, 3, 4, 5, 6, 7, 8] (you get the idea)

“We are told to eat zero carbs and lots of fat on the Paleo diet. Our model tried to mimic that, but we didn’t see any improvements in weight or symptoms. In fact, they got worse.” [9]

The Study

The study in question wanted to look at how a low carbohydrate, high fat (LCHF) diet affects body weight, glucose tolerance, insulin sensitivity and pancreatic function.  To do this, they used a mouse model called the New Zealand Obese (NZO) mouse and fed one group of mice a LCHF diet and the other group a standard low fat rodent chow diet (see supplementary tables 1 and 2 respectively) [10]

Low Carb, High Fat
Standard Chow
Energy Density (MJ/kg)

The LCHF group gained fat and had worse glucose intolerance and insulin sensitivity than the chow group (the chow group weren’t great either as this is a model of obesity and insulin resistance, but the LCHF group was definitely worse) [10]

* The senior author commented that “The moral of the story is that calories matter. If you eat more calories, you will put on more weight” [9], but this is actually kind of the opposite of what this study found.  The LCHF group gained a lot more fat even though if anything they are consuming fewer calories (figure 1) [10].  I’m simply noting this as an interesting and contradictory observation.  It tells you something about the model, but unfortunately will likely be used by calorie deniers

Mice vs Humans

This animal study is a long way from being able to make such conclusions you see in the paper and in the media.  It’s not an RCT in humans and so generalising this effect to humans assumes these mice respond the same way to different diets as humans do.  This assumption is implied by the senior author:

“The researchers used mice for the study, because their genetic, biological and behavioural characteristics closely resemble those of humans.” [9]

But certain strains of rodents in particular simply do not do well on high fat diets.  One of the most common mouse strains in mice obesity/diabetes research is the C57Bl/6 mice that were bred to develop obesity and the metabolic syndrome on a high fat diet (see quote below) [11].  While the NZO mouse strain used for this study lacks a protein involved in fat metabolism (and this genetic defect isn’t seen in humans) [12] and becomes obese and diabetic with a high fat diet, regardless of whether the fat comes from lard, sunflower oil or fish oil [13]

“An increase in dietary fat content has been shown to produce diabetes and obesity in various strains of mice and in rats. The B6 mouse is a particularly good model of the human metabolic syndrome because it develops a syndrome of obesity, hyperinsulinemia, hyperglycemia, and hypertension, when allowed ad libitum access to a high-fat diet, but remains lean and physically normal when restricted to low-fat chow. In marked comparison to B6, other strains, such as the A/J mouse or the C57BL/KsJ (KsJ), are relatively resistant to these effects of a high-fat diet” [11]

So the results of this study shouldn’t surprise anyone who has read a few papers on diet induced obesity in rodents, least of all the president of the Australian Diabetes Society.  It doesn’t really matter whether the major fat source is saturated fat, monounsaturated fat or omega 6 polyunsaturated fat, a high fat diet based on either of these fairly reliably induces obesity, impaired glucose tolerance and insulin resistance in mice [14, 15, 16, 17, 18, 19] (though may differentially affect things like lipogenic and inflammatory gene expression and gut bacteria).  Coconut oil doesn’t seem to share this effect [20] and fish oil tends to not cause these problems either and is often protective [21, 22, 23, 24, 19].  This is also the case in a number of rat strains [25]

However, there have been 10 large (≥ 100 participants) and long term (≥ 12 months) RCTs in humans that have compared the effect of LCHF diets with low fat diets and these consistently find that they are equal to or better than low fat diets for weight loss (which would probably come out as significant in a meta-analysis), and equal to low fat diets for lowering glucose and/or insulin levels [26].  In addition, based on what I’ve so far, the low carb groups are more successful in many of the shorter trials.  But apparently: "Low-carbohydrate, high-fat diets are becoming more popular, but there is no scientific evidence that these diets work.” [6]

Similarly there have been a few RCTs in humans that have compared the effect of Paleo diets with conventional dietary advice and these consistently find that they are equal to [27, 28, 29, 30] or better than [31, 32] conventional dietary advice for weight loss, and equal to [28, 29, 30, 31] or better than [27] conventional dietary advice for lowering glucose and/or insulin levels.  A meta-analysis of the first four Paleo RCTs found that the Paleo diet beats conventional dietary advice in reducing waist circumference and equal to conventional dietary advice in reducing fasting glucose [33]

* It’s important to note that the Paleo diets used in these RCTs were never very low carbohydrate diets like the one used in the mouse study (39% [27], 32% [31], 29% [28], 32% [29], 54% [30], 23% [32])

** The senior author recommends the Mediterranean diet, but the amount (~40% of total calories) and type (olive oil/MUFA) of fat promoted in the Mediterranean diet causes obesity, glucose intolerance and insulin resistance in mice [15, 17]

Their LCHF Diet Sucks

Another issue with generalising to humans is what they fed to the mice.   The LCHF diet contained: cocoa butter, casein, sucrose, canola oil, ghee and cellulose (and added vitamins and minerals).  The standard chow diet contained: wheat, wheat by-products, fish meal, tallow/vegetable oil blend, soybean meal, skim milk powder, yeast, molasses, limestone (and added vitamins and minerals) [10]

The LCHF diet is extremely processed, whereas the chow diet is ok and looks exceptionally good in comparison, with it at least being based on whole foods.  The chow diet is not all that dissimilar from conventional dietary advice (which is a bit of stretch), but the LCHF diet certainly doesn’t resemble anything that someone would eat, particularly someone following a Paleo diet.  Who cares about the macronutrients, where is the unprocessed food?  Where are the meats, fruit and vegetables that are staples on Paleo?

The whole idea behind Paleo is that what animals would normally eat in the wild is what they are more likely to be best/more genetically adapted to, and therefore more likely to have better health outcomes when on such diets.  In this respect Paleo advocates would expect the chow group to be leaner and healthier than the high fat group based on the chow diet being based on whole foods and being more appropriate for mice

* The premise of this study is that the senior author has this idea that "We are told to eat zero carbs and lots of fat on the Paleo diet”.  This simply isn’t true even among advocates of very low carb diets like Pete Evans (this study is probably just a reaction to him).  Most Paleo advocates either leave macros up to each person to decide or recommend a lower carb diet usually around 20-40% of total calories (which is similar to Loren Cordain’s average of 22-40% [33]).  (I’m probably around 40%, which is why I don’t weigh few hundred kilos now :P “To put that in perspective, a 100 kilogram person on a Paleo diet could pile on 15 kilograms in two months.”)

Why do Animal Research Then?

This brings up the question of why do animal research at all for chronic disease?  Animal research can and should be used to make good guesses for the mechanisms of diseases or improving our understanding of basic physiology

  • Animal models can be used to study organs that we wouldn’t have easy/any access to in humans (such as the brain, liver and heart)
  • Diseases can be induced by diet or with chemicals (such as streptozotocin for diabetes) whereas this would be quite unethical in humans
  • Experimental treatments can be tested in animals, such as bariatric surgery, novel potentially therapeutic substances and gene therapy (which is often preceded by gene knockouts, overexpression and/or underexpression studies)

As mice clearly respond differently to humans from dietary interventions, using a high fat diet in mouse studies is simply a model of diet induced obesity, as opposed to a chemical or genetic cause of obesity

Such animal models of mouse high fat diet induced obesity should be used where appropriate, not to be prescriptive for human needs, but because the vast majority of humans develop obesity and type 2 diabetes from dietary causes rather than genetic or chemical causes.  So an animal model of diet induced obesity, rather than chemical or genetic obesity would likely be more suitable model.  It’s not perfect, as it is after all, just a model

I don’t see much value in this study.  It is simply characterising (the β-cell function data) a potential model of obesity and the metabolic syndrome (the NZO mouse) that may be less informative than other models.  What got it published, let alone in Nature, was likely the authors and the narrative that could be sold to the media.  As other people have said, the way this study have been publicised is disgraceful

No comments:

Post a Comment