Wednesday, September 30, 2015

The Paleo Meta-Analysis

There are currently a few RCTs where the Paleo diet is compared to conventional dietary advice

Lindeberg et al (2009) and Jonsson et al (2010) (blog) Paleo vs. Mediterranean
Jonsson et al (2009) and Jonsson et al (2013) (blog) Paleo vs. American Diabetes Association
Mellberg et al (2014) (blog) Paleo vs. Nordic Nutrition Recommendations
Boers et al (2014) (blog) Paleo vs. Dutch Health Council Guidelines
Masharani et al (2015) (blog) Paleo vs. American Diabetes Association
Pastore et al (2015) (blog) Paleo vs. American Heart Association
As well as some other trials that don’t have a control group (blog and blog)

Several weeks ago a meta-analysis was published that pooled the results of the first four trials  (link).  Masharani et al and Pastore et al weren't included because they were both published after the systematic review for the meta-analysis.  The meta-analysis isn’t open access, but has also been discussed by Dan Pardi and Adel Moussa, and I suggest you check those links out

Overall, the trials tend to greater improvements in the Paleo group, but in many cases this wasn’t significant and it could be argued that the trials were likely underpowered.  This is an advantage of a meta-analysis, by pooling similar studies together you increase the sample size (power) and may find a statistically significant result from several non-significant trends.  In fact, that’s pretty much what the meta-analysis showed (see bottom of the post)

The meta-analysis has been criticised by all the usual people for various reasons.  While I agree that it’s a bit early to do a meta-analysis, as it could only include 4 trials, this meta-analysis serves an important purpose to draw interest for further RCTs rather than being definitive evidence that grains/dairy are bad because you can’t draw those conclusions with the way the Paleo diet trials are designed due to the multiple differences in the diets

You’ll notice that some of the measures have high heterogeneity (I2 > 50%).  This is to be expected as there were some differences in study design such as slightly different diets (particularly carbohydrate as it relates to HDL-C and triglycerides), the time to follow up (which varied quite a lot), participant characteristics and whether the diets were isocaloric or not.  Again, this is a problem with only 4 trials and ideally you would have a meta-analysis with enough trials (at least 10-20) to analyse the effects of these using subgroups

I also agree that basically any diet can produce short term weight loss and some metabolic improvements.  Where Paleo seems to shine is its ability, at least anecdotally, to dramatically improve things like autoimmune diseases, miscellaneous GI and skin issues, and some other chronic diseases.  Hopefully this meta-analysis also draws interest for RCTs in these areas as well

I found it surprising that when considering the potential ‘biological foundations of Paleolithic nutritional benefits’, the meta-analysis introduced the Carnivore Connection hypothesis, which is that insulin resistance is an adaptation to ancient hunter-gatherers consuming low carbohydrate diets.  I don’t agree with that completely as the Paleo and conventional dietary advice groups in the RCTs didn’t always differ in carbohydrate consumption by that much (see table) although even a 10% TE reduction in carbohydrate should favourably affect HDL-C and triglycerides.  Also, I don’t agree with the Carnivore Connection hypothesis (blog and blog)

% Carbohydrates
Paleo
Conventional
Difference
Lindeberg
39
47
-8
Jonsson
32
42
-10
Mellberg
29
44
-15
Boers
32
50
-18
Masharani
54
58
-4

Finally, it’s important to recognise that the ‘Paleo diets’ used in the RCTs often differ by varying degrees to the way most Paleo diets are implemented.  Not this this is necessarily good or bad.  I think it’s appropriate to use this as evidence provided you explain the reasons for the differences (blog).  But it’s certainly not appropriate to use this as evidence for very low carbohydrate diets or nuts (both of which have been done before)

Sunday, September 13, 2015

List of Posts on the Clinical Trials included by SFA vs. PUFA Meta-Analyses

In the following table I have listed the clinical trials included in meta-analyses related to saturated fat and/or polyunsaturated fat and the incidence of coronary heart disease.  The table lists which the trials included by the meta-analyses and how they are categorised in Ramsden et al (2010) and Hooper et al (2012)*.  I have also added hyperlinks to the blog posts on the trials.  Due to size constraints in the table I abbreviated the meta-analyses as follows:

H12     Hooper et al (2012)
Har      Harcombe et al (2015)
H15     Hooper et al (2015)

* Ramsden et al categorised trials as replacing SFA with just n-6 PUFA (6) or with both n-6 and n-3 PUFA (6+3).  Hooper et al (2012) categorised trials as involving fat modification (M) or both modifications and reductions in fat (M,R).  For the National Diet Heart Study Hooper et al (2012) drew several comparisons between the experimental and control groups in the first and second studies.  Some of the experimental diets were modified fat while others were both modified and reduced fat and were categorised accordingly (Both)



SM
Moz
Ram
H12
Cho
SH
Har
H15
X


X
(6)
X
(M)

X
X
X







X

X
X
X
(6+3)
X
(M)
X
X
X
X




X
(Both)




X
X
X
(6+3)
X
(M)
X
X
X
X
X
X
X
(6+3)
X
(M)
X


X
X
X
X


X






X
(6)
X
(M)
X
X
X
X



X
(M)




X
X
X
X
(6)
X
(M)
X




X
X

X
(M,R)
X
X


X
X
X
X
(6+3)
X
(M,R)
X
X


X








X








X



X
(M,R)













X








X







X



X
(M,R)