Monday, November 30, 2015

Does Dietary Protein Impair Insulin Sensitivity?

If you read some diabetes literature you can easily get the sense that everything is bad: protein, fat and insulin cause insulin resistance, while carbohydrates spike blood glucose and insulin levels.  Today I’m going to look at protein and insulin resistance

In many ways there’s a fairly reasonable case for protein impairing insulin sensitivity:

·         Infusion of amino acids impairs insulin sensitivity [1] [2]
·         People with insulin resistance tend to have elevated levels of most amino acids, particularly the branched chain amino acids (leucine, isoleucine and valine) [3] [4] [5] [6], but not glycine [4] [7] [8] [9]
·         Mice fed a high protein low carbohydrate diet developed insulin resistance (the high protein diet was a bit ridiculous at 60% protein, but the ‘moderate’ protein group had at 33% protein had higher insulin levels but similar glucose tolerance to the low protein group at 5% protein) [10]

There’s quite a reasonable mechanism for this.  Amino acids (particularly leucine) activate mTOR.  mTOR forms a complex with Raptor called mTORC1.  mTORC1 activates S6K1.  S6K1 promotes protein and lipid synthesis.  But S6K1 is part of a negative feedback loop as it inhibits the activation of insulin receptor substrate 1 (IRS-1), leading to an impairment in insulin sensitivity [11] [12] [13] [14]

The black circle represents an inhibition

However, whether high protein diets impair insulin sensitivity in clinical trials is fairly mixed

·         A few studies find no difference with a higher protein diet [15] [16] [17] (while the latter of these found an improvement with the high carbohydrate, high fibre diet)
·         Whey protein supplementation improved insulin sensitivity with no change in fasting glucose [18]
·         Unlike the hypocaloric high carbohydrate diet, the hypocaloric high protein diet didn’t result in improvements in fasting glucose, insulin sensitivity and HbA1c [19]
·         The Zone diet (high protein), Akins diet (high fat) and conventional dietary advice (high carbohydrate) all reduced fasting insulin in insulin resistant obese women [20]

t and there are alternative explanations to the observations above

·         Amino acid infusion is hardly a physiological state
·         The higher levels of amino acids seen in people with insulin resistance could be a consequence of insulin resistance rather than a cause of it (see this review paper [21])
·         Higher protein diets (33%) and/or casein (their source of protein + supplemental methionine) are biologically inappropriate for mice 

So I’m not sure whether dietary protein impairs insulin sensitivity.  It’s not a good idea to jump to conclusions based on mechanisms, animal studies and biomarkers.  To answer this question it would be ideal to have a clinical trial to manipulate all three macronutrients because replacing protein with carbohydrate (which is the most common method) only tells you about the effect of protein relative to carbohydrate

Sunday, November 29, 2015

Nutrition Science is Controversial, but That's OK

Nutrition/diet science is a pretty controversial field.  At times it can seem a bit ridiculous, but that’s probably each of us perceives at least one of the controversies to be a non-issue and/or clearly resolved by the current body of evidence

But for the average person, who appears to have both low nutrition and scientific literacy, this can be quite confusing.  For this reason people may decide to opt out of engaging with nutrition science altogether or use that as a cover to justify to themselves and/or others why they are continuing to eat a diet that they understand to be probably not all that good

Achieving a consensus in nutrition was the aim of a recent conference by Oldways, which set out to find a common ground between diet advocates and nutrition scientists with some differing views

I’m sympathetic to the idea of a consensus, or a list of things almost all people agree on.  My rationale for this is that at the end of day I imagine almost everyone agrees that whole minimally processed foods are almost always superior to highly refined/processed foods regardless of macronutrient composition/etc and that getting sufficient micronutrients is also very important.  And then when we look what the average person eats (with 40% of calories coming from ‘extras’ [1]) and their nutrient intakes [2] it’s clear that this should be the first step and that engaging the average person with even a basic level of nutrition science will be a good thing

But the problem with nutrition science/dietary guidelines is that so much of the evidence is based on mechanisms and observational studies rather than randomised controlled trials.  As a result nutrition science is often quite speculative.  If someone asked ‘is it healthier to eat oats or eggs for breakfast’ there is enough evidence to have an intelligent conversation, but I don’t think we are close to answering that question definitively.  Even if we had a meta-analysis of long term RCTs there’s a component of individual variation as well and ignoring this can lead to the argument that ‘if X work for you, then you didn’t do X properly’ (the no true Scotsman fallacy).  All of this would be ok if there was an intelligent conversation and acknowledgment of the limitations of the evidence, but this doesn’t tend to happen

Furthermore, what this ‘consensus’ looks like and the lengths people went to fit this consensus (see Boyd Eaton’s presentation) is also a problem.  The consensus is essentially the dietary guidelines, a Mediterranean diet and a vegetarian diet, which wasn’t surprising considering the people involved.  I don’t subscribe to the idea that the dietary guidelines made us fat and sick, but I think they can be improved

To some extent the presence of a consensus can lead to dogma and also be used to shut down discussion of how current dietary guidelines can be improved.  While this seems to be a greater issue in more politicised topics, it still happens in nutrition science (for example).  In this respect the controversial nature of nutrition science can be seen as advantage over other more stagnant fields

The internet opens the doors to huge number of viewpoints on a wide range of issues, some of which may not necessarily be as simple as being right or wrong.  Some parts of nutrition science are simple and usually these are the areas with a lot of agreement (whole foods > processed foods, and adequate nutrition), while other parts of the nutrition science are complex and these are often the areas of controversy (diet and disease relationships).  Particularly as we’re living in the information age, we should stop telling people what to think, but rather how to think, so they can then go engage with the information out there and be better able to identify which are the logical arguments backed by evidence and which aren’t 

I also recommend reading George Henderson’s post on this topic

Sunday, November 22, 2015

Is Saturated Fat Associated with Cholesterol Levels in Cohort Studies? Part 2

In part 1 I mentioned that people have argued that in ‘observational studies performing a multivariate analysis that includes cholesterol levels is over-adjustment for the relationship between saturated fat and coronary heart disease and then listed many that cohort studies haven’t actually reported the baseline relationship between saturated fat and cholesterol levels.  The saturated fat intake and cholesterol levels in the studies that reported them are presented in the table below

Quintile/Quartile
1
2
3
4
5
P Value
HPFS [5] Table 1
(g/day & mmol/l)
15.9
21.5
24.8
27.7
32.4

5.2
5.3
5.3
5.3
5.3
???
(g/day & mmol/l)
5.0
8.5
11.9
18.3


4.81
4.82
4.95
5.07

< 0.001
JPHC [27] Table 1
(g/day & mg/dl)
9.6
13.4
16.3
19.4
24.9

203±36
206±36
207±35
208±35
209±34
< 0.001
Kuopio [29] Suppl 2
(g/day & mmol/l*)
32.3
42.4
53.0
66.8


6.06
6.11
6.24
6.38


* Calculated from average HDL-C and total-C:HDL-C values.  Association between saturated fat with both LDL-C and HDL-C in Kuopio was significant

HPFS found no association between saturated fat and cholesterol levels, whereas the other three cohorts did.  Yet even when saturated fat intake is associated with cholesterol levels, the difference between the top and bottom quartile/quintiles is a lot smaller than expected considering the large difference in saturated fat (even considering the higher intake of PUFA).  The only other study that reported a relationship between saturated fat intake and cholesterol levels just correlated the two variables instead of reporting the data in quintiles/etc as is typically done in more modern observational studies.  This paper (published in 1981) also discusses that associations between saturated fat and cholesterol in cross-sectional studies are inconsistent.  So clearly there have been more studies that looked at this if you’re interested and that the evidence seems to be inconsistent and seems to have been for some time

There are some explanations for this:

1) Some people have commented that a reason saturated fat may not be associated with cholesterol levels in observational studies is because of the large inter-individual variation in cholesterol levels that makes statistical significance so difficult.  I don’t think this is a good explanation because having a very large sample size, which is necessary for this kind of observational study, means that even a small difference despite large inter-individual variation would still highly significant.  An example of this is the JPHC study in the table above with a weak, yet still significant relationship between saturated fat intake and cholesterol levels because those cholesterol measurements came from 8765 men and 16047 women

2) The feeding studies are confounded by trans fats, dietary cholesterol, phytosterols and other phytochemicals, fibre, etc.  Or perhaps that the effects seen in short term feeding studies aren’t really long term and are compensated for quite slowly by feedback mechanisms

3) It’s interesting that in the NHS and HPFS that saturated fat intake is associated with a lower incidence of ‘hypercholesterolemia’.  Perhaps there is some reverse causality going on whereby people with genetically high cholesterol are informed of this and change their diet accordingly, but still have somewhat cholesterol levels despite a low intake of saturated fat.  This would explain how the relationship between saturated fat and cholesterol levels was absent in the HPFS and less than expected in the other three cohorts.  The NHS and HPFS are nurses and ‘health professionals’ and so presumably quite health conscious, which may explain how there was no relationship between saturated fat and cholesterol levels in the HPFS (more health consciousness >> more reverse causality) whereas the other three cohorts had a weak relationship (less health consciousness >> less reverse causality)

Anyway, the argument that ‘a multivariate analysis that adjusts for cholesterol levels is an over-adjustment’ should be taken with a grain of salt until relevant figures are provided

Finally, whether or not saturated fat intake is associated with higher cholesterol levels in the long term doesn’t necessarily invalidate your response.  A lot of science (clinical trials, observational studies, etc) is based on averages and your response could be quite different to the average

Saturday, November 21, 2015

Is Saturated Fat Associated with Cholesterol Levels in Cohort Studies? Part 1

I previously discussed how in observational studies, when the participants are grouped by something like intake of saturated fat, the baseline characteristics of these groups of participants often differs in some respects.  In the case of saturated fat, those who ate more saturated fat tended to exercise less, smoke more, have lower diet quality (trans fats, whole grains), etc.  Observational studies now often adjust for these confounding variables in what’s called a multivariate analysis, which completely removed the previously highly significant association between saturated fat and coronary heart disease in the study (blog)

However, the reporting of multivariate adjusted risk ratios (RRs) has been criticised by a few people.  They argue that age-adjusted RRs show that saturated fat is associated with coronary heart disease, and that the multivariate adjustment is an over adjustment as it removes the relationship between saturated fat, total-C/LDL-C and coronary heart disease

I think this is a valid concern.  If I were only able to report one figure it would be multivariate without cholesterol levels, because total cholesterol is generally not that affected by other dietary/lifestyle factors or health status (whereas HDL-C and triglycerides tend to be very good markers of insulin resistance).  But why not report both: a multivariate adjustment with and without cholesterol levels?

But these people have made the assumption that saturated fat is associated with higher cholesterol levels, but from what I’ve seen so far there is a conspicuous lack of reporting of this in the observational studies.  This is quite suspicious as conventional nutrition/public health is completely obsessed with cholesterol levels and so if there was a relationship between saturated fat intake and cholesterol I’m sure this would be reported.  Interestingly in the two cohorts I mentioned previously (the Nurse’s Health Study and the Health Professionals Follow-up Study) saturated fat intake was associated with a lower incidence of ‘hypercholesterolemia’ (not sure how this is defined but may be > 240 mg/dl or > 6.3 mmol/l) although average or median cholesterol levels wasn’t reported (blog)

I’ve generally stayed out of the debate of whether saturated fat increases cholesterol.  Many studies suggest saturated fat increases both HDL-C and LDL-C, but doesn’t significantly alter the total-C:HDL-C ratio [1], although this has been debated [2].  But this got me curious: is saturated fat actually associated with higher cholesterol levels in observational studies?

I looked at the observational studies included by the Siri-Tarino (ST) [3] and de Souza (CHD events = DSE, CHD mortality = DSM) [4] meta-analyses.  The following table lists those studies and whether saturated fat intake was associated with cholesterol levels in each of these cohorts

Cohort
ST
DSE
DSM
Associated?
Health Professional’s Follow-up Study (HPFS) [5]
X
X
X
No
Health and Lifestyle Survey [6]
X

X
Not Reported*
Lipid research clinics prevalence follow-up study [7]
X

X
Not Reported
Multinational MONItoring of trends and determinants in CArdiovascular disease (MONICA) [8]
X
X

Not Reported
Honolulu Heart Program [9]
X
X

Not Reported
Strong Heart Study [10]
X
X
X
Not Reported
Framingham Study (1) [11]
X


Not Reported
Ireland–Boston Diet–Heart Study [12]
X

X
Not Reported
Malmö Diet and Cancer Study [13]
X
X

Not Reported
Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC) [14]
X
X
X
Not Reported
Framingham Study (2) [15]
X
X

Not Reported
Western Electric study [16]
X

X
Yes**
Baltimore Longitudinal Study of Aging [17]
X

X
Not Reported
Caerphilly Study [18]
X
X

Not Reported
Health professional follow up study (2) [19]
X


Not Reported
Fat and Protein Intakes and Risk of Intraparenchymal Hemorrhage among Middle-aged Japanese [20]
X


Yes
Nurses’ Health Study (1) (NHS) [21]
X


Not Reported*
Dietary determinants of ischaemic heart disease in health conscious individuals [22]
X

X
Not Reported
Nurse’s Health Study (2) (NHS) [23]
X
X

Not Reported*
Adult health study [24]
X


Not Reported
Israeli Ischemic Heart Disease Study [25]
X

X
No Access
Women’s Health Initiative (WHI) [26]

X

Inappropriate
Japan Public Health Center Study (JPHC) [27]

X

Yes
Zutphen Elderly Study [28]

X

Not Published
Kuopio Ischemic Heart Disease Risk Factor Study [29]


X
Yes
** Only reported a correlation and not by quartile/quintile

In summary most studies didn’t report whether people with a higher intake of saturated fat had level cholesterol levels.  Many of the studies were not specifically looking at saturated fat and coronary heart disease, so this is understandable.  Those marked with an asterisk (*) reported baseline characteristics for saturated fat intake but didn’t include cholesterol levels

I’ll discuss this in part 2