1Examine.comResearch Digest

Issue 3 January 2015

ERD

2Table of Contents05 Heart benefits of alcohol may not apply to everyone

17 Type 2 diabetes: a preventable disease21 Investigating a progression of carb and saturated fat intakes

33 Whence the hype?43 Running on empty: can we chase the fat away?51 Fitting into your genes: do genetic testing-based dietary recommendations work?

61 Combating obesity through intermittent fasting

70 How does a lifetime of marijuana use affect the brain?

79 A mouses microbiome may cause its brain to leak

87 Ask the Researcher: Stuart M. Phillips, Ph.D., FACN, FACSM

92 INTERVIEW: Ramsey Nijem

3Kamal Patel, Editor-in-Chief

From the EditorHowever, more research is needed ...

Have you ever seen that line in a journal article? Of course you have. Its a part of almost every article that we review for ERD. Is more research ever not needed?

A fellow researcher and I would talk about how ubiq-uitous this phrase was, and whether it really meant anything. He eventually wrote a letter to the editor of an epidemiology journal, including some analysis on how often the phrase was used in major journals. Three years later, I still run across the phrase a dozen times a day. This may never change.

Why is this phrase important? Well, it ties in to one of the most important, yet least talked about issues in health research: when are new trials justified, and what exactly should new trials test? Theres a field of research called value of information analysis, which places a dollar amount on the public health value of each unit of new research on a given topic.

There are only so many research dollars available. Not every topic can get funding for a large randomized trial, and many important topics go unresearched. Id like to know whether taking vitamin D in the morn-ing causes different effects than night-time ingestion. Will we see research on this topic? Probably not.

Major issues that have already been addressed by animal studies and observational trials often are next

in line for randomized trials, such as the impact of [INSERT NUTRIENT OR DRUG HERE] on heart dis-ease biomarkers in [INSERT POPULATION HERE]. Treatment is funded more often than prevention, and multimodal prevention is funded much less often than interventions investigating a single method or pharmaceutical.

Maybe that seems backwards. But its not easy to test the combined impact of getting regular sleep, eating mostly unrefined foods, getting time outside in the sun, and carving out time to relax and get some per-spective. Actually, its pretty difficult to test even one of those interventions. Plus theres much less money to be made on prevention, especially when it comes to free interventions, than there is to be made by sell-ing treatments.

Theres a phrase that refers to the inherent nature of human existence, including choices and difficulties: The Human Condition. Sometimes, I think there is a counterpart in The Research Condition. Health research is complex and shifting, and somewhat inherently flawed. Single trials cant conclusively answer questions. Subtle differences in methods and samples lead to different results. Research doesnt really flip flop very often its just a much more iterative and grueling process than the public knows. And its why more research is always needed.

4ContributorsResearchers

Trevor KasheyPh.D(c)

Alex LeafM.S(c)

Courtney SilverthornPh.D.

Margaret WertheimM.S., RD

Zach BohannanM.S.

Gregory LopezPharm.D.

Pablo Sanchez SoriaPh.D.

Kamal PatelM.B.A., M.P.H., Ph.D(c)

Editors

Arya SharmaPh.D., M.D.

Natalie MuthM.D., M.P.H., RD

Stephan GuyenetPh.D.

Sarah BallantynePh.D.

Katherine RizzoneM.D.

Mark KernPh.D., RD

Gillian MandichPh.D(c)

Reviewers

5Heart benefits of alcohol may not apply to everyoneCETP TaqIB genotype modifies the association between alcohol and coronary heart disease:

The INTERGENE case-control study

IntroductionWith advice coming from everyone from physicians to bartenders, a com-mon message broadcast during the past couple decades has been that moderate consumption of alcohol is not just allowable, but beneficial for heart disease. Indeed, imbibing to the tune of one drink daily for women, or two drinks daily for men, has been associated with lower risk of cardio-vascular disease.

Proposed mechanisms for the protective effect of alcohol on coronary heart disease (CHD) include the potential benefits from the antioxidant effects of polyphenols in wine, and an increase in high density lipoprotein (HDL) levels. HDLs most well known function is to transport cholesterol from arteries throughout the body back to the liver, preventing cholesterol from being deposited in the arteries, which would cause blockages.

6Lipid-containing particles in the blood often gain and lose different types of lipids, such as cholester-ol and triglycerides. The ability of HDL to transfer cholesterol into particles like VLDL is partially reg-ulated by cholesteryl ester transfer protein (CETP). CETP promotes transfer of HDL cholesterol into VLDL, and in exchange HDL receives triglycerides. CETP is hence thought to reduce HDL cholesterol, so less CETP in your blood means HDL particles would balloon up with more cholesterol, and more CETP would mean HDL particles would car-ry less cholesterol.

Hold on, less HDL cho-lesterol isnt that a bad thing? Not necessarily, as HDL is more complex than just the good cho-lesterol moniker it has taken on in public par-lance (and unfortunately physician office parlance as well). HDL also has a lesser known, but important role in the immune system, performing a variety of functions, such as binding toxic substances in the blood. HDL can be anti-inflammatory or inflammatory, depending on the disease state of the body. HDL and LDL are markers of disease, but they each have physiologi-cal functions important to the body, and neither are absolute determiners of or protectors against heart disease.

Back to CETP. There is a known polymorphism in the gene that encodes CETP called CETP TaqIB. A polymorphism is when a particular gene has two or more relatively common possible nucleotide

sequences at a given site in the DNA. Both versions of the DNA sequence would be considered normal, with neither likely to directly cause debilitating dis-ease, like a rare mutation might. However, different polymorphisms may still influence susceptibility to disease.

This study looked at how two polymorphisms in the CETP gene affect the odds of having CHD at

varying levels of alcohol intake. The two different alleles (gene variants) of CETP are called B1 and B2. B2 is associated with decreased CETP mass and increased HDL cho-lesterol. Given that we have two copies each of gene, the three different genotype options in a given subject are B1B1, B1B2, or B2B2.

A previous study showed that men with B2B2 gen-otype who have an ethanol intake of 50 g (about three drinks) or more per day had about a 60% lower risk of heart attacks than men with lower or no alcohol intake. This protective effect of larger amounts of alcohol was not seen in people with the B1B1 or B1B2 genotypes. On the other hand, in a study in a Mediterranean cohort, no interaction between CETP TaqIB, alcohol intake, and CHD was observed.

Why is that? One reason could be simply different populations. As seen in Figure 1, different pop-ulations can have substantially different CETP genotype frequencies. Rodents such as mice have no

HDL can be anti-inflammatory or inflammatory, depending on the disease state of the body.

7CETP gene, and also have lower risk of atheroscle-rosis, though many other factors may be responsible this. Complete CETP deficiency is a rare mutation in humans, although its much more frequent in one area of northern Japan. While the frequency of this mutation is higher in people with heart disease, at least in that area of Japan, recent studies have shown that the extremely cholesterol-rich HDL in these people still maintains its antioxidative function and ability to move cholesterol out of areas of cholesterol buildup. So the impact of CETP on heart disease is still very much up in the air.

The aim of the current study was to re-examine the effect of alcohol intake and its interaction with CETP Taq1B polymorphism on CHD odds.

Moderate alcohol intake is often encouraged to help ward off heart disease. This advice is largely based on HDL effects, but these effects may also be modified by your genotype.

Who and what was studied?Population This case-control study took place in Sweden as part of the INTERGENE research program, which aims to assess the interaction between genetic susceptibil-ity and chronic disease in southwest Sweden. Cases with heart disease were compared against controls who didnt have heart disease, to assess how alcohol and genetic variation impacted disease prevalence.

The CHD cases were patients under age 75, admit-

Figure 1: CETP B2B2 allele frequency in different populations

8ted to three regional hospitals for acute coronary syndrome and diagnosed with myocardial infarc-tion. Of the CHD patients who agreed to participate, 618 patients were included (453 men, 165 women).

Of those, 209 men and 86 women had a first-time myocardial infarction, while the remaining 323 had an exacerbation of previously diagnosed CHD. The controls without CHD were randomly selected individuals aged 25-74 at the time of sampling, and 2,921 of them were included.

Intervention The data collected for analysis in this study was CETP genotype, as well as self-reported information about alcohol intake, including frequency of intake of different types of alcohol (low-alcohol beer, medi-um-strong beer, strong beer, wine, dessert wine, and spirits) with eight response categories ranging from never to three or more times a day.

Alcohol intake information referred to intake over the previous one-year period for controls and for the one-year period prior to the most recent coronary events for study participants. Age and sex-specific standard serving sizes for alcoholic beverages were used to calculate the daily ethanol consumption.

Daily alcohol intake was divided into three levels (low, medium, and high), and the odds ratio (OR) was calculated for having CHD based on genotype and alcohol intake. Abstainers were classified into a fourth group, though high/intermediate intake was compared to the low group, not to the abstainers.

All models were adjusted for age, body mass index (BMI), HDL, sex, and smoking habits. The tertile cut-offs are shown in Table 1. In this study, high

alcohol intake for men was considered about one drink or more daily, while low intake was about less than a drink daily. For women, high alcohol intake was classified as about a drink or more daily, while low alcohol intake was less than of a drink daily.

Table 1: Tertiles of Ethanol IntakeMen (g/day) Women (g/day)

Low < 6.5 < 3.2

Medium 6.5-13.1 3.2-6.3

High > 13.1 >6.3

One drink is 14 grams of ethanol, which is the equivalent of about 12 ounces of beer, five ounces of wine, or 1.5 ounces of 40%-alcohol spirits.

This study examined just over 600 cases of heart disease and almost 3,000 controls, and classified how much alcohol they drank into three catego-ries that differed based on sex.

What were the findings?Characteristics of Case and Controls For both men and women, there was a smaller per-centage of alcohol users in the cases compared to the control groups. For women, 80% of CHD cases and 87% of controls reported using alcohol. For men, 89% of cases, compared to 93% of controls, report-ing drinking alcohol. People with CHD also had lower average ethanol intake compared to controls. There were no significant differences in the distri-bution of CETP genotype (B1B1 versus B1B2 versus B2B2) between cases and controls.

The cases were older than controls (around 62 years, compared to 51) and sicker. Almost 20% of the peo-

9ple with CHD had diabetes, compared to under 5% in the control group. In addition to being heavier, people with CHD were more likely to be smokers.

Alcohol Intake on CHD In the entire cohort, intermediate drinkers had a 35% lower odds of CHD, compared to low drinkers, regardless of genotype. High drink-ers had a non-significant 10% lower odds compared to low drinkers.

Those who abstain from alcohol are often found in observational studies to have a higher risk of heart disease than moderate drink-ers. In this study however, both low drinkers and abstainers had increased odds compared to moderate drinkers, and low drinkers did not have lower odds than abstainers. This suggests that the fac-tors typically attributed to abstainers that may impact heart disease (different social habits, higher previous alcoholism, etc.) may not have had a large impact in this population.

Genotype on CHD There were no significant effects of genotype on CHD odds in the whole cohort, when researchers used B1B1 as a reference. For B2B2, the 10% lower CHD odds was not statistically significant.

When the same logistic regression model was not adjusted for HDL cholesterol, the B2B2 genotype was associated with a 29% lower CHD odds in the whole cohort. The fact that adjustment for HDL level reduced the effect of B2B2 on CHD odds is not surprising, as the CETP gene is known to be involved in the regulation of HDL.

Alcohol Intake and Genotype on CHD B2B2 homozygotes had a remarkable decrease in CHD odds when they were intermediate alcohol drinkers (79%) and high drinkers (52%) as compared to low drinkers. In B1 carriers (B1B1 or B1B2 genotypes), intermediate drinkers had a 20% lower odds of CHD, though it was not statistically significant. B1 carriers who were high drinkers had essentially the same odds as low drinkers.

Why odds reduction instead of risk reduction?

You may have noticed the word odds

popping up a lot in this review. The rea-

son stems back to this study not being

a randomized trial. It didnt actively test

interventions on different groups of peo-

ple, and see what develops over time.

Nor did it observe participants and mea-

sure variables as time progresses, like

a prospective observational trial does.

Rather, at one slice in time it estimated

previous alcohol intake and tested for

CETP alleles in a group with heart dis-

ease and a group without heart disease.

Since the study was a case-control study,

it cant use the simpler and more intui-

tive risk terminology. Randomized trials

happen over time, hence you can be sure

that giving the intervention preceded the

outcome, and estimate the risk of the

outcome based on what intervention

was given. That isnt true of case-con-

trol studies such as this one, and hence

you can only measure the odds of the

outcome in one group versus anoth-

er group. However, when a disease is

rare, happening in around 10% or less

of the population thats studied, the

odds ratio and relative risk

will be approximate-

ly the same, due the

mathematical

formulas for

each con-

verging.

10

B2B2 Genotype in Intermediate Drinkers B2B2 intermediate drinkers had a substantial and significant 59% reduction in CHD odds compared to non-B2B2 intermediate drinkers.

Prevented Fraction Based on the authors calculation of prevented fraction, this population would have had around 6% more cases of CHD if the combination of B2B2 and intermediate/high alcohol consumption had not existed.

While B1B1 and B1B2 genotypes werent associated with lower heart disease risk, B2B2 intermediate drinkers had 79% lower risk than low drinkers, and B2B2 high drinkers had a 52% low-er risk. These numbers equate to an estimated 6% reduction in CHD for the overall population.

What does the study really tell us?Based on the results of the current study, intermediate to high alco-hol intake does not significantly reduce CHD odds in people with B1B1 or B1B2 genotypes. In B2B2 genotypes, intermediate alcohol intake was associated with a 79% reduction in CHD odds, while high alcohol intake was associated with a 52% odds reduction.

These results also held up to a variety of sensitivity analyses, such as measuring alcohol intake in four cutoffs rather than three, including or excluding adjustment for HDL and various other potential con-founders, or when analysis was restricted to those age 60 or older or those who were enrolled at their first cardiac event.

One strength of this study was that different cut-offs of alcohol intake were taken into account, rather than just comparing low and high intake. The models were adjusted for age, BMI, HDL, sex, and smoking habits, to correct for common confounding factors. The authors also tested additional factors, like leisure time physical activ-ity, financial security, education levels, marital status, and diabetes status, but these had no effect on the results. It could be surmised that intermediate drinkers have more healthy behaviors than the

11

high alcohol group, but at least for the factors men-tioned, this was not the case. Thus, the protective effect of B2B2 at intermediate and higher alcohol intakes could not be explained by HDL cholesterol or other lifestyle and socioeconomic variables.

That being said, the cases and controls differed widely on a variety of characteristics associated with disease, such as age, weight, and diabetic status. It is possible that there were other important confound-ers that were not controlled for.

The study also didnt discuss potential mechanisms that may explain the results. Previous research in Norwegians showed that HDL may not be so important for the protective effect of alcohol on heart disease. However this Swedish study looked

specifically at CETP, a gene that appears to be only involved in transfer of cholesterol from HDL to other lipoproteins. Yet it found that the additional protective effect of CETP in intermediate and high drinkers (on top of just the alcohol intake) was not explained by HDL levels. This could be due to a vari-ety of factors perhaps a simple measurement of HDL cholesterol is less important than the number and type of HDL particles. As was referenced before, HDL can be anti-inflammatory or pro-inflammatory depending on physiological context, so simply stick-ing HDL into a regression may not fully describe the role of HDL in the relationship between CETP gen-otype and heart disease odds.

The study results didnt change when sensitivi-ty analysis was performed with different alcohol intake cutoffs and different confounders. However, the cases and controls differed in a variety of characteristics, and its possible that important potential confounders werent controlled for.

The big pictureHaving the B2B2 genotype didnt have a strong pro-tective effect on its own, and neither did drinking intermediate or high amounts of alcohol on its own. But combining these two factors was associated with a substantial reduction in the odds of heart disease. The authors focused mostly on intermediate intakes in their discussion, but high intakes also had a sub-stantial reduction in odds, at 52% (compared to 79% in intermediate drinkers). This may be because high intakes come with much higher risks.

Earlier studies didnt take into account CETP gen-otype, and likely showed a less substantial but still

Earlier studies didnt take into account CETP genotype, and likely showed a less substantial but still protective effect of alcohol intake due to a dilutional effect

12

protective effect of alcohol intake due to a dilu-tional effect meaning that the substantial odds reduction in people with B2B2 likely may have been diluted by the lack of CHD odds reduction in people with B1B1 or B1B2 genotypes.

These results confirm a previous study, which showed that men who were B2B2 homozygotes with an alcohol intake of 50 grams a day or more had lower myocardial infarction risk, and the risk reduction was the strongest when the participants drank 75 grams a day or more. In the current study, however, the greatest risk reduction was seen at an alcohol intake of 6.5-13.1 grams a day, significantly lower daily intake than seen previously.

It is surprisingly easy to derive different conclusions based on something as simple as cutoff points the same data can be sliced into two parts with high versus low intakes, or several different intakes. And the reference group can also differ between studies. In this study, the reference group was made up of low alcohol drinkers, rather than those who totally abstain, as abstainers can be quite a diverse group that includes anybody from former alcoholics to those who dont drink for religious reasons. Some

large and well known previous studies, such as the Harvard-run and U.S.-based Nurses Health Study and Health Professional Follow-Up Study, suggest a protective effect of the B2 allele. The reference group in that analysis, however, was abstainers rather than those with a low alcohol intake. Women in those studies were found to have stronger benefit from the B2 allele than did men, which was not found in this Swedish study. Because study designs and pop-ulations differ, its hard to directly compare different CETP studies.

This study also had some important methodologi-cal limitations. Subjects were queried on frequency of alcohol intake, but were not asked about portion size. Standard portion sizes were used to calculate daily alcohol intake. This could lead to inaccuracies in daily intake data. In addition, under-reporting of alcohol intake is common during self-reporting, which could skew the intermediate and high tertiles of intake. Furthermore, CHD cases could also have reduced their alcohol intake in response to the diag-nosis or under-report intake if they think they are supposed to limit intake, but this effect is likely to be same regardless of CETP genotype. This is a weak-ness of the case-control design, as a prospective study

13

that collects data before CHD develops may be less subject to this kind of under-reporting. Its also pos-sible that intermediate alcohol users could also have generally healthier eating and lifestyle habits that were not captured in the logistic regression model.

This is also just one study among several on the topic, some of which show conflicting results. This paper was done on a geographically limited sample in Sweden, so the results may not apply to those in another region, like East Asia or Central America. The small sample size also limits the conclusions that can be made from this paper. Headlines read-ing Heart benefits unlikely from alcohol likely wont mention that this study only included 13 cases who had the B2 allele and were intermediate alco-hol drinkers. With comparison groups this small, this study is just one more step in the progression of studies on the topic, rather than being the final word on alcohol and heart disease.

Its important to remember that a variety of factors could influence the effect of alcohol on heart dis-ease, other than just genetics, such as age, sex, and insulin resistance. Observational studies cannot attribute causation or lack of causation to HDL or

LDL, no matter how strong the associations appear. While HDL = good, CETP = bad is a simplistic and inaccurate way of thinking, it is surprisingly pervasive. CETP may promote heart disease in some situations, and have no effect in others.

Thus meta-analyses of CETPs overall effect on lip-ids and heart disease risk may inadvertently gloss over interaction effects from factors like alcohol intake levels or other variables that may moderate CETPs effects. The topic of heart disease, alcohol, and HDL is a great example of how focusing on a single article abstract without context, even if that abstract describes a well-conducted meta-analysis, can be quite misleading. A meta-analysis is only as good as the studies it contains, and the more complex the interactions get and the more hetero-geneous the study designs are, the higher the risk of a meta-analysis coming to erroneous conclusions. A meta-analysis of seven studies found that alcohol did not interact with the B2B2 genotype, but it com-pared current drinkers versus nondrinkers, which is likely to be too crude of a comparison to uncover the more complex relationship found in this study.

With comparison groups this small, this study is just one more step in the progression of studies on the topic, rather than being the final word on alcohol and heart disease.

14

This study confirms some previous evidence while conflicting with other evidence, likely due to dividing alcohol intakes into different levels while using low drinkers as the reference group rather than abstainers. The study is another part of the CETP and heart disease puzzle, which is yet to be fully solved.

Frequently asked questionsDoes frequency of alcohol consumption matter? Would 49 grams of alcohol once weekly (average of seven grams/day) be just as beneficial for CHD risk in a B2B2 homozygote as daily alcohol intake of 7 grams? Its unclear from these study results how frequency of alcohol intake affects CHD risk reduction. Since binge drinking is not advised, the smaller amount would be more consistent with current health guide-lines for daily consumption. Heavy drinking increases the risk of some types of stroke and atrial fibrillation, which highlights the variety of other cardiovascular outcomes that are related to alcohol consumption.

Is B2B2 protective for CHD when combined with intermediate alcohol intake in both men and women? Its unclear at this point whether the B2B2 geno-type with intermediate alcohol intake is protective against CHD in women. The study under review and Nurses Health Study may have not had a large enough number of heart disease cases to detect these effects. For women who are non-drinkers or low drinkers, increasing alcohol intake to reduce CHD risk wouldnt necessarily be advised, giv-en other data that suggests a higher risk of other chronic disease, including breast cancer, linked with alcohol intake.

Context is also very important: the additional effect of alcohol on heart dis-ease wont be nearly as important for a young person without many risk factors as it is for someone who has already had heart disease. The combined risks of alcohol side effects, plus potential risk of alcoholism, may very well out-weigh alcohol benefits for heart health even if one is a B2B2 carrier.

Why are studies on cardiovascular effects of alcohol and CETP so con-flicting? Its not really possible to do a ran-domized trial of different alcohol intakes, and see what the cardio-vascular effects are. Without RCTs, observational studies in different populations couple with mechanis-tic and animal studies to form the evidence base.

Analyses in observational stud-ies can use a variety of statistical methods and control for different possible confounders, which could lead to different conclusions even using the same data. So, even though the largest meta-analysis on CETP to date shows that the B2 allele has a statistically sig-nificant but weak protective effect, the result is heavi-ly dependant on the methods used by the studies it included.

15

Additionally, the mechanisms by which CETP may help prevent or promote heart disease are also not clear. In other words, this is a research area that is still progressing, and disagreements exist within the academic community. We will keep our collective eyes out for new studies on this topic.

Does my CETP allele mean that I have higher risk of heart disease? This is the million dollar question, for which there is only a five cent answer: we dont know. Although this particular study had compelling results due to studying a variety of alcohol intake levels and adjusting for a variety of variables, CETP study results in general are really all over the place.

For example, one review found that the effect of B2B2 differed depending on the population that was looked at. In participants with a high risk of heart disease it was protective, while in general popula-tions it promoted heart disease! The frequency of B2B2 also differed, being much less frequent among those with high risk. B2B2 sometimes could predict whether a lipid-lowering drug would prevent heart disease, and sometimes couldnt.

Can I take a drug to modify my CETP activity and prevent heart disease? Because increased CETP activity decreases HDL levels, this became a research target for new medica-tions in the 2000s. One promising drug, torcetrapib, reliably raised HDL levels by inhibiting CETP activ-ity, as well as lowering LDL. However, the trial was terminated early due to torcetrapib causing a 25% increase in cardiovascular deaths alongside a 60% increase in deaths from any cause.

So to repeat: we dont know quite how CETP affects

heart disease. The effect of your genotype may be modified by your diet, habits, medications taken (especially statins) and even other genes. HDL and LDL by themselves dont mean that much in isola-tion, and neither does your CETP genotype. Some people are able to get a portion of their genomes sequences through services such as 23andme, and that may help inform the effect of alcohol on a par-ticular individuals heart health. That being said, the evidence is nowhere near concrete, and the uncer-tainty about alcohol benefits on heart health is one of the major takeaways on this topic.

What should I know?

[...] there's only a small portion of the population for whom alcohol intake is protective against CHD, and most all of them are unaware that they have a potentially protective gene.

16

In short, moderate alcohol consumption may not protect everyone equally from heart attacks. Protective effects likely depend on genetics. The results of this study raise the question of whether the recommendations regarding alcohol intake for the prevention of CHD are too overarching. Substantial CHD odds reduction was only seen in people who were B2B2 homozygotes, with intermediate to high alcohol intake. For someone giving advice about how to prevent heart disease (like a physician, or someone advising an older parent), keep in mind that the evidence is still quite mixed on this topic. In the context of public policy, the authors estimat-ed that 6% of heart disease was prevented by the combination of B2B2 and intermediate/high alcohol intake. This is not a huge amount for something that can have several important detriments like drinking alcohol does.

Its important to note that only 19% of the entire cohort in this study had the B2B2 genotype. While the frequency of this genotype in the general pop-ulation is unknown, the beneficial effect of alcohol intake on CHD odds would only apply to the small segment of the population who are B2B2 homo-zygotes with intermediate to high alcohol intake. Perhaps in the future, genetic testing will help us determine our behaviors around alcohol. But for now it seems theres only a small portion of the pop-ulation for whom alcohol intake is protective against CHD, and most all of them are unaware that they have a potentially protective gene.

Well discuss the potentially complex relationship between alcohol and heart disease in the private ERD readers Facebook group. Join us!

17

Type 2 diabetes: a preventable disease

By Stephan Guyenet, Ph.D.

Three thousand and five hundred years ago, ancient Egyptian physicians reported

excessive urination in some of their patientsa key diagnostic sign of diabetes.

The mummy of Queen Hatshepsut, a powerful pharaoh who ruled ancient Egypt

during this time period, suggests that she was obese and likely suffered from type

2 diabetes. Throughout history, other royals have been posthumously diagnosed

with probable type 2 diabetes, including the portly King Henry VIII of England. Dia-

betes has been a scourge of the affluent for thousands of years.

Diabetes is defined as a failure of blood glucose con-trol, leading to excessively elevated blood glucose. This failure of blood glucose control results from insufficient action of the pancreatic hormone insulin, which normally constrains blood glucose concen-trations, both in the fasting state and after meals. During type 1 diabetes (formerly called juvenile-on-set diabetes), the bodys immune system attacks and destroys insulin-secreting beta cells in the pancreas,

leading to a near-total disappearance of circulating insulin. In type 2 diabetes (formerly called adult-on-set diabetes), the bodys tissues lose their sensitivity to the insulin signal. The pancreas compensates by secreting more insulin, but eventually the beta cells are unable to maintain this excessive level of insulin secretion, insulin levels decline, and blood glucose levels rise.

18

This failure of blood glucose control, and accompa-nying metabolic disturbances, leads to the familiar signs and symptoms of diabetes: excessive thirst and urination, glucose in the urine, excessive hun-ger, weight loss, fatigue, slow healing, and eventually, vascular disease, kidney failure, as well as nerve and retinal damage.

The reason type 2 diabetes is no longer called adult-onset diabetes is that it now occurs in chil-dren as well as adults. This trend is part of an increase in global diabetes risk that affects people of nearly all age groups in all affluent nations. Diabetes is extremely rare in cultures that maintain a lifestyle similar to our (non-royal) distant ancestors, yet more than a third of modern Americans are project-ed to develop diabetes at some point in life. Nearly all of these cases will be type 2 diabetes. Fortunately, the causes of diabetes are well known, so much so that we know how to prevent the large majority of cases. Lets have a look.

Obesity Over the last century, but particularly the last three decades, Americans have bought progressively lon-ger belts. In 1971, 15 percent of Americans were obese, yet by 2009, that number had more than dou-

bled to 36 percent. The rest of the affluent world is following closely behind. Excess body fat is likely the single largest contributor to the modern epidemic of diabetes.

The following graph illustrates the relationship between body mass index (BMI; a measure of body fatness) and diabetes incidence over a five-year peri-od in American men:

Diabetes Risk According to BMI

A BMI between 18.5 and 25 is considered lean, 25 to 30 is considered overweight, and 30 or great-er is considered obese. As you can see, the risk of

Diabetes is extremely rare in cultures that maintain a lifestyle similar to our (non-royal) distant ancestors, yet more than a third of modern Americans are projected to develop diabetes at some point in life.

19

developing diabetes increases rapidly with increas-ing BMI, and the relationship is extremely strong. A man with a BMI greater than 35 (obese) has a 42-fold greater risk of developing diabetes than a man with a BMI below 23 (lean). If we zoom in on the lower end of the graph, we can see that diabetes risk increases by 50 percent even before we leave the lean BMI range, and more than doubles for people who are only slightly overweight!

Diabetes Risk According to BMI

Countless experiments show that this is more than just an association: excess body fat contributes to the metabolic disturbances that lead to type 2 diabetes. This appears particularly true of the vis-ceral fat that surrounds the organs underneath the abdominal wall.

Age Nearly all lifestyle-related disorders are strongly linked to age, and type 2 diabetes is no exception. Among the elderly, the yearly likelihood of being diagnosed with diabetes is more than 30 times great-er than among young adults. Part of this excess risk isnt linked to age directly, but to the fact that most

people gain fat, lose muscle, and become more sed-entary with age.

Physical activity Muscle tissue is the single largest user of glucose in the body, and when its fuel needs are high, it increases its sensitivity to insulin to accelerate glu-cose uptake. Because of this, physical activity causes a rapid and profound increase in muscle insulin sen-sitivity, leading to an increase in whole-body insulin sensitivity. This increase in insulin sensitivity only lasts a few days, so regular physical activity is essen-tial to maintain it.

Not surprisingly, people who are more physically active have a lower risk of developing diabetes, and the association is substantial. People who engage in regular vigorous exercise, or even walk regularly, have just over half the diabetes risk of people who are the most sedentary.

Genetics One of the most effective ways to avoid type 2 diabe-tes is to choose your parents wisely. All of the most common forms of diabetes, including type 2 dia-betes, have a strong genetic component. Like most lifestyle-related disorders, diabetes is not usually caused by a single gene variant. Rather, its caused by complex interactions between many different gene variants and the environment in which a person lives.

Possibly for genetic reasons, certain racial groups are at a higher risk of diabetes than others. For exam-ple, Asians, including people of Indian descent, are at a higher risk of developing type 2 diabetes at any given BMI. In other words, a modestly overweight Indian or Chinese person may have the same diabe-tes risk as an obese Caucasian person.

20

The genes that influence type 2 diabetes risk tend to be involved in the development and function of the insu-lin-secreting pancreas, and to a lesser extent, body fatness. Some of these genes may determine how well beta cells are able to cope with the metabolic battering that accom-panies obesity and insulin resistance.

Preventing type 2 diabetes Some risk factors arent modifiable: we simply have to live with them. We cant change the genetic cards weve been dealt, nor can we roll back the years of our lives that have elapsed. Still, the risk factors we can control are so pow-erful that they can eliminate the large majority of type 2 diabetes risk. Several randomized controlled trials have clearly demonstrated this, including the massive Diabetes Prevention Program (DPP) trial. This trial reported that a combination of dietary weight loss and regular exercise reduced the risk of developing diabetes by an astounding 58 percent over a 2.8-year period in pre-diabetic volun-teers. Several similar trials conducted in other countries and other racial/ethnic groups reported almost identical results. This is one of the greatest triumphs of modern biomedical science.

Keep in mind that these trials started with people who were already nearly diabetic, and who didnt lose much weight or adhere particularly closely to the intervention. Imagine what a lifetime of healthy living could do.

diabetes risk increases by 50 percent even before we leave the lean BMI range, and more than doubles for people who are only slightly overweight!

Stephan is an obesity researcher, neurobiologist, and author. In addition to his research, he enjoys synthesizing and com-municating science for a general audience. He has a B.S. in biochemistry (University of Virginia) and a Ph.D. in neurobiology (University of Washington). His blog Whole Health Source is a free resource for anyone who loves the science of health.

21

Investigating a progression

of carb and saturated fat

intakesEffects of step-wise increases

in dietary carbohydrate on circulating saturated fatty acids

and palmitoleic acid in adults with metabolic syndrome

IntroductionSaturated fat reduction has long been a major target of dietary guidelines, although recent meta-analyses have failed to show an association with heart disease. Current recommendations in the U.S. include limiting saturated fat intake to less than 10% of total energy intake. However, a reduction in fat intake typically leads to an increase in carbohydrate intake. A conse-quence of overconsumption of carbohydrates is increased de novo lipogenesis (DNL). DNL is a process which involves the synthesis of fatty acids from non-lipid sources, such as carbo-hydrates or amino acids. Interestingly, even energy-balanced diets, and single-meal consumption of carbohydrates above the normal oxidative capacity of the body have been shown to

22

increase DNL. The percentage of ingested carbohy-drate contributing to DNL is however quite minor in people who arent insulin resistant and overfeeding on refined carbohydrate.

The major end-product of DNL is the saturated fat palmitic acid (denoted 16:0, referring to 16 carbons and zero double bonds), which can be desaturated within the body to form the monounsaturated fat palmitoleic acid (16:1). Higher blood levels of pal-mitoleic acid have been associated with an increased risk of metabolic syndrome and greater amount of inflammatory markers. Palmitoleic has mixed evi-dence however, also being associated with some positive biomarkers such as higher HDL and greater insulin sensitivity. Divergent impacts could be due to the effects of different lifestyle factors and differ-ent physiological conditions (such as how much of DNL is from adipose tissue versus from the liver).

This study sought to assess how incremental chang-es in dietary carbohydrate intake and decreases in saturated fat intake affect plasma saturated fatty acid and palmitoleic acid levels. The study was con-ducted in adults with metabolic syndrome under hypocaloric conditions.

Saturated fat is commonly targeted for reduction by dietary guidelines. This typically leads to an increase in carbohydrate intake, which at high levels may cause the body to create fats through de novo lipogenesis. This study investigated sev-eral levels of saturated fat and carb intake to see how they affected plasma saturated fats and pal-mitoleic acid.

Who and what was studied?The study was an 18-week controlled dietary intervention in which the participants were ini-tially fed a low-carbohydrate diet that gradually shifted to a high-carbohydrate diet over six con-secutive phases (from lowest carb to highest carb: C1C2C3C4C5C6).

Prior to beginning the six feeding interventions, the participants were instructed to follow a low-carbo-hydrate run-in diet for three weeks that mimicked the first low-carbohydrate phase, in order to initiate metabolic adaptations to carbohydrate restriction. Baseline and run-in nutrient intakes were deter-mined with the help of three-day food logs.

The percentage of ingested carbohydrate contributing to DNL is however quite minor in those who arent insulin resistant and overfeeding on refined carbohydrate.

23

All food was provided for the subjects during the 18-week intervention. Participants picked up their meals three to four times per week, and if the sub-jects could not travel to pick up their food, the researchers arranged for delivery in order to ensure that every subject received their food as planned. Blood testing was done at baseline, after the run-in diet, and after each phase (before transition to the next diet) to determine fatty acid composition and other blood markers.

Over the entire 21-week period (intervention and run-in), the subjects diets were designed to produce a 300 kcal deficit per day. Resting energy expendi-ture (REE) was estimated at baseline with indirect calorimetry and multiplied by an activity factor to estimate the total daily energy expenditure (TDEE) of the subjects. Protein was held constant at 1.8 grams per kilogram of ideal bodyweight. As carbo-hydrates were increased every three weeks over the six feeding phases, total fat was decreased to main-tain energy intake. Thus, across the entire study, protein and calorie intake was similar. Saturated fat was also maintained, at 40% of total fat intake. In comparison, Americans only derive around 34% of their calories from any kind of fat, with around 13% coming from saturated fat.

Only very-low and non-caloric products such as coffee, tea, water, and diet soda were allowed to be consumed by the participants in addition to the provided foods. Beef, eggs, and dairy were the pri-mary protein and fat sources, with higher and lower fat versions used depending on the study phase. Low-glycemic carbohydrates were emphasized throughout.

The subjects were 12 overweight and obese men and four women with metabolic syndrome, between 30 and 66 years old (average 44.9) with BMI ranging from 27-50 kg/m2 (average 37.9). Exclusion criteria included having diabetes, liver, kidney, or other met-abolic or endocrine dysfunction. Participants who were physically active were asked to maintain their activity levels while sedentary people were asked not to begin an exercise program.

This study investigated the effects of various car-bohydrate diets on a group of overweight and obese participants. Study participants initial-ly ate a low-carbohydrate diet that turned into a high-carbohydrate diet over 18 weeks, in six phases.

What were the findings?Energy intake (EI) across the feeding interventions averaged about 2,500 kcal per day and protein intake averaged about 125g per day (20% EI). As designed, protein and energy intake remained constant over the 18-week intervention. As seen in Figure 1, car-bohydrate intake started at an average of 47 grams per day (7% EI) and rose to an average of 346 grams per day (55% EI). Total fat intake started at an aver-age of 209 grams per day (73% EI) and dropped

Indirect calorimetry

Indirect calorimetry measures the production of

carbon dioxide and consumption of oxygen to esti-

mate heat production. This is then entered into an

equation to estimate resting energy expenditure.

Although not without error, indirect calorimetry

remains the gold standard for measuring energy

expenditure in laboratory settings.

24

to an average of 80 grams per day (28% EI). The authors claim that compliance was high, based on verbal communication and inspection of returned food containers. There were no dropouts.

Both body weight and fat mass (measured by DXA) showed a significant decline from baseline to C1 (about seven kilograms and four kilograms, respec-tively), and continued to decline throughout the entire intervention, ultimately resulting in an aver-age loss of about 10 kilograms of bodyweight and eight kilograms of body fat. Neither weight loss nor fat mass were significantly different between C4 and C6, suggesting that most of the change occurred in the first 12 weeks (run-in, C1, C2, & C3).

Total, LDL, and HDL cholesterol values were not significantly altered across any of the feeding phases. Triglycerides levels dropped about 22% from base-line to C1. These levels stayed constant through C5 and had a significant return to baseline values at C6.

Compared to baseline, fasting glucose & insulin, HOMA-IR (measure of insulin resistance), and systolic and diastolic blood pressure significantly decreased at C1, but were not significantly altered throughout the six feeding phases.

Despite saturated fat intake starting at 84 grams per day and decreasing to 32 grams per day, the propor-tion of total saturated fatty acids in blood lipids was not significantly affected. Palmitic acid (16:0), the predominant saturated fatty acid within blood lipids, significantly increased from baseline to C1 and sig-nificantly decreased from C1 to C2, C3, C4, and C5. C6 was not significantly different from C1.

Stearic acid (18:0, which is commonly found in ani-mal fats and cocoa) was not significantly changed in cholesterol esters. But from baseline to C1, it was significantly reduced in phospholipids and also decreased in triglycerides through the intervention, ending with a significant reduction in C6 relative to C1. Phospholipid concentrations showed an oppo-

Figure 1: Carb and saturated fat intake by study period

25

site pattern, increasing throughout the intervention and ending with a significant increase in C6 relative to C1.

There was a significant reduction in total monounsaturated fatty acid concentrations from baseline to C1 only. Similar to 18:0, as carbohy-drate increased, plasma oleic acid (18:1) decreased in triglycerides, but increased in phospholipids.

Palmitoleic acid (16:1) was significantly reduced from baseline to C1 in triglycerides and cholesterol esters, and trended for an increase in phospholipid concentrations. All these markers showed increasing concentrations with increasing carbohydrate intake and ended the intervention with significantly greater concentrations of palmitoleic acid at C6 relative to C1.

There was great individual variation in palmitoleic acid concentra-tions during each diet phase with notable outliers. For instance, one subject had triglyceride concentrations of palmitoleic acid rise by nearly three-fold from C1 to C4 (2% to about 5.8%) and further rise from C4 to C6 (about 5.8% to 7%). However, another subject showed no changes across the entire intervention, and another showed reductions as carbohydrate intake increased.

Study participants lost body weight and fat over the 18-week intervention, with most of the changes occurring in the first 12 weeks. The blood samples researchers analyzed suggested that carbohydrate intake can influence blood levels of compounds like palmitoleic, stearic, and palmitic acid.

What does the study really tell us?There are numerous studies showing associations between high-er proportions of palmitoleic acid in blood and tissue, and adverse health outcomes such as metabolic syndrome in adults and ado-lescents, hypertriglyceridemia, type-2 diabetes, coronary heart disease, and prostate cancer. However, since none of these studies establish causality, it is possible that these conditions lead to high-

Lipoproteins and lipid fractions

This study looked at how much pal-

mitoleic acid was contained in three

different locations in blood plas-

ma: triglycerides, phospholipids,

and cholesterol esters. Lipoproteins

shuttle lipids (such as fatty acids

and cholesterol) around the body.

Phospholipids form the outer shell

of lipoproteins, while cholesterol

esters and triglycerides make up the

majority of the core.

So the phospholipid fraction refers

to the fats that are contained in the

phospholipids, with the same rea-

soning for triglyceride fraction

and cholesterol ester fraction.

Sometimes these different fractions

respond the same way to diet, and

sometimes they dont. Hence its

important to measure all of them.

26

er proportions of palmitoleic acid (for example, palmitoleic acid may be the bodys attempt at a pro-tective response to what is being eaten) rather than vice-versa. With the mixed associations shown in studies, it is hard to know for sure what the exact health effects of palmitoleic acid are.

It is also difficult to quantify the amount of pal-mitoleic acid needed to increase the risk of these endpoints, as few studies have done so. In the Physicians Health Study, one standard deviation increase in plasma phospholipid palmitoleic acid concentrations was associated with a significant 17% higher risk of heart failure even after adjustment for BMI, alcohol con-sumption, smoking, exercise, and plasma omega-3 levels.

In the study under review, baseline daily intake of car-bohydrate and fat averaged 333 grams and 130 grams, respectively. During the first phase of the intervention, carbo-hydrate intake dropped to an average of 47 grams, while fat intake rose to an average of 209 grams. It was during this time that the most significant changes in blood lipid fatty acid concentrations occurred, including a major reductions in palmitole-ic acid levels. Additionally, this was when significant improvements in insulin sensitivity, blood pres-sure, and plasma triglyceride levels were observed. However, this was also when the most significant

reductions in weight and fat mass were observed, making the causative factor difficult to isolate. And there was no weight loss matched control group to account for weight loss effects. Between the lower palmitoleic acid concentrations, the weight and fat loss, and the reduction in carbohydrate intake, we cannot say which came first and which led to which.

On the other hand, by the end of the intervention, when carbohydrate intake was similar to baseline intake (346 grams vs. 333 grams) plasma palmi-toleic acid levels returned to levels similar to those

observed at baseline despite significantly lower weight and fat mass, strongly sug-gesting that it was carbohydrate intake that influenced plas-ma palmitoleic acid levels.

The authors also repeated the entire experiment back-wards in five additional subjects

(from high to low carbohydrate intake) and found that plasma palmitoleic acid responded in the exact opposite pattern as the main study group, which supports the idea that carbohydrate intake influ-ences palmitoleic acid concentrations. Even so, the overall diets were hypocaloric, and we cannot con-clude how carbohydrate intake would influence palmitoleic acid levels under eucaloric or hyperca-loric contexts.

With the mixed associations shown in studies, it is hard to know for sure what the exact health effects of palmitoleic acid are.

27

This study provides evidence to suggest carbo-hydrate intake influences palmitoleic acid levels. Although evidence is mixed, high levels of pal-mitoleic acid in the blood are associated with metabolic syndrome, type 2 diabetes, coronary heart disease, and other health problems. In this study, participants experienced a drop in palmi-toleic acid levels when they were eating low-carb meals in the first phase of the study.

The big pictureWith 18 full weeks worth of food provided for the participants, this study provided a well-controlled environment in which to study the effects of diet on palmitoleic acid. Yet despite the findings from this study, the relative risk from various palmitole-ic acid concentrations in the blood remains to be determined. In the previously mentioned Physicians Health Study, the highest quartile had an average palmitoleic acid level of only 0.50%, whereas in the current study, even when phospholipid palmitoleic acid concentrations were at their lowest during the low carbohydrate phase, absolute concentrations averaged 0.61%, putting these participants above the vast majority of the Physicians Health Study subjects.

Other blood lipid changes add further complex-ity to the implications of this study. For instance, increasing carbohydrate intake led to greater phos-pholipid oleic acid concentration, which in contrast to palmitoleic acid, has been shown to attenuate the pro-inflammatory and cytotoxic effects of excessive saturated fatty acid incorporation. Myristic acid, which showed a reduction with carbohydrate restric-tion, plays a physiologically critical role in de novo ceramide synthesis (necessary for regulating cell dif-

fer-entiation, prolifer-ation, and apoptosis) and has been shown to increase delta-6 desat-urase activity (first step in creating long-chained polyunsaturated fatty acids such as EPA, DHA, and arachidonic acid from their short-chained precursors).

The applicability of this study to real-life situa-tions is uncertain. There were only 16 participants, with widely varying BMIs, each using a particular dietary composition for a limited period of time. The effect of carbs on blood lipids was confounded by the weight loss that was designed into the study, without a weight loss control group that would help to isolate the effects of carbs. Also, a variety of differ-ent outcomes were measured. So while palmitoleic acid was emphasized in the title and study discus-sion, other important outcomes had different results.

For example, outside of C1, cholesterol and blood pressure didnt change regardless of diet. The sub-jects in this study already had metabolic syndrome,

28

so changes in things like blood pressure and tri-glycerides may be more important than changes in bound plasma fatty acids, since some of these fatty acids are linked to metabolic syndrome (which they already have) while blood pres-sure may have a more direct impact on their health. Also, circulating free fatty acids, which are linked to metabolic and heart health, were not assessed.

While the total proportion of plasma saturated fats didnt differ in any of the diet phases, different indi-vidual plasma fatty acids can have different effects. Palmitic acid, the predominant saturated fatty acid which was noted in the paper to be a predictor of metabolic syndrome and heart disease, was actually lower in phospholipids (but not the other two lipid fractions) from C2-C5 than it was during the low carb C1 or high-carb C6 periods. This finding was not explained, nor were changes in stearic acid and oleic acid. So while a variety of fatty acids were mea-sured and reported, palmitoleic acid was the only one focused on in the discussion. Unfortunately it was also the only focused on in many news stories with inaccurate headlines such as Heart disease and diabetes risk linked to carbs, not fat, study finds.

It must be noted that this study was funded by the Dairy Research Institute, The Beef Checkoff pro-

gram, the Egg Nutrition Center, and the Robert C. And Veronica Atkins Foundation. The funding sources did not have a say in designing the study or writing the man-uscript. However, these organizations are quite clearly interested in the research on saturat-ed fatty acids, thus the variety of stud-

ies funded by them. The primary investigators are also noted low-carb researchers. This also doesnt mean the study is biased, but it is one thing to keep in mind when interpreting the study findings. A given topic (here, the effect of carbohydrate intake on plasma saturated fatty acids), can be explored in a variety of different ways, and the results can be interpreted by the study authors in different ways as well. Its important to look at the broader context of literature and the nitty-gritty study details rather than just take the authors word for it.

Other plasma fatty acids, such as palmitic, myristic, and oleic acid, may be important for evaluating the health effects of different carbohy-drate and fat intakes. Although measured, these were not a focus of the study. Nor were more direct predictors of heart and metabolic health, such as blood pressure. The study was funded by dairy, beef, and low-carbohydrate organizations.

A given topic [...] can be explored in a variety of different ways, and the results can be interpreted by the study authors in different ways as well.

29

Frequently Asked QuestionsWhat else influences plasma palmitoleic acid levels? The current study lends support to the idea that palmitoleic acid con-centration in the plasma is more reliant on carbohydrate intake than fat intake. However, the study was conducted under hypocaloric con-ditions, and previous research has suggested that dietary intake of palmitoleic acid (which is rich in a few select foods such as macadamia nuts) does significantly influence plasma concentrations during weight maintenance. Alcohol has also been suggested to reduce palmitoleic acid concentrations, with one study reporting significantly lower levels in people consuming more than 100mL of ethanol consumption per week (seven regular 12-ounce beers) compared to people consuming less. This study also found palmitoleic acid concentrations to be inde-pendent of smoking status.

How do various biomarkers of fatty acids in the body differ? Biomarkers of fatty acid composition differ from dietary intake, in that biomarkers reflect both the intake and the utilization of the fat-ty acids. Because not everyone is similar in how we absorb, transport, and metabolize nutrients, biomarkers allow us to look beyond simple dietary intake and focus on the physiological consequences of consum-ing certain substances. Moreover, biomarkers can provide a long-term picture of dietary intake.

Due to the essential nature of fatty acids in cell structure, assessment can involve numerous body tissues in addition to blood and urine (e.g. hair, nails, skin, breath, saliva, feces). However, measuring blood plas-ma is the most common method. Serum triglycerides reflect dietary intakes over the past hours to days, whereas cholesterol esters and phospholipids reflect daily intakes. Only body fat stores (adipose tissue) tend to reflect long-term dietary fat consumption (e.g. years), and even this measure can be inaccurate in people who have experienced cycles of fat loss and gain.

How strongly is palmitoleic acid associated with heart disease, when compared to other biomarkers? Although statistically significant, the strength of the relationships

30

between palmitoleic acid and health parameters is low to moderate. For instance, in one study of over 3200 Chinese adults, palmitoleic acid concentrations could only explain about 37% of the variance in triglyceride levels and 14% of the variance in HDL-cholesterol levels.

It should also be kept in mind that fatty acid levels in any biomarker represent a proportion and not an absolute measure. Thus, greater integration of cer-tain fatty acids into the biomarker can reduce the percentage of other fatty acids without their absolute amount changing. All of the aforementioned studies demonstrating associations between fatty acids and health outcomes were based on percentages, mak-ing it difficult to draw conclusions as these are not quantifiable values. One person could have double the amount of palmitoleic acid in serum as another person and still have similar percentages if they also have double the amount of blood lipid.

There is also evidence of seasonal variations in fat-ty acid profiles. One early study showed greater proportions of saturated fatty acids in the adipose tissue of the legs and arms during summer com-pared to winter. This difference was attributable to a reduction of palmitoleic and oleic acid levels, with

a simultaneous increase in palmitic, myristic, and stearic acid levels. Although these changes were in adipose tissue and not serum biomarkers, it rais-es the question of whether the current study could have been influenced by seasonal changes as its six month duration, by necessity, spanned more than one season. Since subtle changes in plasma fatty acid levels were tracked over increments of time, it would be difficult to differentiate what changes were at least partly a result of the season.

What dietary sources have a lot of palmitoleic acid in them? According to the USDA nutrient database, roasted chicken skin from the leg and thigh contains the greatest amount of palmitoleic acid with 2.8 grams per 100 grams of food. Beef fat follows with about 1.9 grams, then turkey skin with 1.34-1.5 grams, and finally butter at 0.96 grams. Poultry skins contain the most palmitoleic acid on average, followed by beef fat and butter. Macadamia oil is a rich source, containing 19% palmitoleic acid.

Keep in mind that palmitoleic acid is different than trans-palmitoleic acid. The latter comes from very limited sources, mostly red meat and dairy from grass-fed cows, and is not synthesized by the body.

Only body fat stores tend to reflect long-term dietary fat consumption (e.g. years), and even this measure can be inaccurate in people who have experienced cycles of fat loss and gain.

31

Trans-palmitoleic acid in plasma lipids and adi-pose tissue has been repeatedly associated with better metabolic outcomes, as shown in this paper by ERD reviewer Stephan Guyenet, Ph.D.

Are there benefits to palmitoleic acid from diet? In plasma? Elsewhere? A very recent study published in December of 2014 found that feeding mice 300 milligrams of pure palmitoleic acid per kilogram of bodyweight daily, in addition to their normal diets for ten days signifi-cantly increased glucose uptake in fat tissue through increased expression of glucose-uptake transporter 4 (GLUT4; necessary for insulin-stimulated glucose uptake into tissues). This was despite no changes in plasma fatty acid levels.

Earlier studies have also found palmitoleic acid to enhance glucose uptake and insulin sensitivity of skeletal muscle, and reduce liver fat buildup. The authors of this study suggest that palmitoleic acid may act as a major signaling lipid produced from fat tissue for communication with distant organs. In obese sheep, infusion of palmitoleic acid twice dai-ly for 28 days preserved insulin sensitivity before beginning an obesogenic diet, possibly through a reduction of intramuscular fat.

It appears that the benefits of palmitoleic acid

revolve around insu-lin-mediated glucose disposal into both muscle and fat tissue. This raises an interesting contradiction, with the studies demonstrating asso-ciations between palmitoleic fatty acid levels in the blood and some adverse health outcomes such as diabetes. Like certain cholesterol markers, palmitole-ic acid may be more of an indicator that something might be physiologically wrong rather than a cause. DNL is one possible cause of increased palmitoleic acid levels, and very high levels may be a marker that something is increasing DNL to dangerous amounts (such as prolonged overeating of carbohydrate, or worsening glucose tolerance from uncontrolled diabetes, both of which can disrupt carbohydrate metabolism). Suggesting that palmitoleic acid is 100% detrimental does not seem accurate given the complexity of evidence on the topic.

What should I know?This study suggests that the presence of certain fat-ty acids in blood lipids appears to depend more on carbohydrate than fat intake under hypocaloric conditions in overweight and obese people with metabolic syndrome. There were minor but uni-form changes in a few select fatty acids, such as

32

myristic acid, oleic acid, and palmitoleic acid, but no significant changes in total saturated and monoun-saturated fatty acid concentrations.

There was also inter-individual variance in the pal-mitoleic concentration response to carbohydrate intake, which is important given the small sample size. While most subjects showed a positive asso-ciation, others stayed relatively unchanged and some showed an inverse association. Moreover, there was greater variance as carbohydrate intake increased. The absolute palmitoleic concentration varied between about 2-4% in plasma triglycerides when carbohydrate intake was lowest during C1, but varied between about 2-7% during the high-carbo-hydrate C6 phase.

Still, the implications of changes in plasma palmitole-ic acid levels have yet to be determined. Many studies demonstrate associations between adverse health outcomes and increased palmitoleic acid levels, but reverse causality cannot be ruled out, nor differing impacts of palmitoleic acid in different contexts. We also do not know what influence many other dietary, lifestyle, and environmental factors have.

Rather than having obvious health implications for differing carb levels, this study serves as additional evidence for those eating low-carb higher saturated fat diets (and losing weight) who are apprehen-sive about impacts on their plasma fatty acids. As is the case with cholesterol, what you eat does not translate directly to what is floating around in your blood. However, the lack of correlation between dietary saturated fat and plasma saturated fat was already shown by a previous paper from the same research group (albeit only the triglyceride fraction was studied).

Its also important to know what this study does not show: it doesnt show that DNL happens at major or dangerous rates when eating moderate carb levels, it doesnt show that increasing levels of carb intake increased overall plasma saturated fat, and it doesnt prove that low-carb diets are superior to moder-ate carb diets for heart or metabolic health. While weight loss decreased as carbs were added, that may very well be due to increased water weight or chang-es in compliance.

The authors conclude that the increased propor-tions of palmitoleic acid concentrations may signal impaired carbohydrate metabolism, yet in vitro and animal studies have suggested that palmitoleic acid is insulin-sensitizing. It seems prudent not to draw health-based conclusions from this study. Rather, the conclusion appears to be that consumption of carbohydrates can have an impact on plasma fatty acid proportions in overweight and obese individu-als under hypocaloric conditions. Whatever health implications this may lead to will require further testing to illuminate.

The health implications of this study are unclear. The lack of impact of dietary saturated fat on plas-ma saturated fatty acids was already shown in previous studies. This study did show an effect of carbohydrate on palmitoleic acid levels, but the relative importance of that is unknown.

Low carb diets are nothing if not controversial. For some evidence-based discussion on their potential health effects, check out the ERD private Facebook group.

33

Whence the hype?The association between exaggeration in health

related science news and academic press releases: retrospective observational study

IntroductionWhen it comes to health news, even though we know not to believe the hype, hype still happens and it has an impact. Not only is the publics use of health care services influenced by the media, but even professionals arent immune. Press coverage of medical research findings is associated with those findings being cited more by other scientists.

Even doctors in the ER test more for certain infec-tions that have been getting heavy press coverage. Since the press is so influential, its important that the media reports medical findings accurately. But it doesnt seem like thats happening: past research has shown media coverage of medical and nutri-tional research is often distorted.

But all of this doesnt imply that the blame lies with the science journalists. They are often under

immense pressure to write more, fast, which encourages reliance on press releases and summa-ries from news agencies, universities, and other public relations outlets. This is why its quite pos-sible that journalists are reporting the information they are receiving (fairly) accurately, and instead it is the information sources they rely upon which lead to media hype.

Indeed, a previous study of press releases from med-ical centers found that many provided exaggerated claims, while few provided caveats and precautions about their claims. Similar results were found in cancer genetics research, where press releases often exaggerated causal claims which were then repeat-ed by the media. But the origin of the hype may go back even further than press releases. One study found that exaggerated claims often could be traced back to the abstract of the original journal article.

34

The purpose of the study under review was to expand upon the research above and trace the source of the hype in health science news.

Hype is ubiquitous in health news reporting. But this hype may come from places other than journalists exaggerating findings. Health news impacts not only the general public, but also phy-sicians and other researchers.

Who and what was studied?The researchers began by searching for public-ly-accessible press releases from Russell Group universities (the top 20 research universities in the UK) that covered research related to human health and were based on peer-reviewed research published in scientific journals in 2011. For each press release based on published scientific research, associated print and online news stories were then located. Broadcast news wasnt examined in this study.

With all of this information in hand, the researchers rigorously defined how hype was created at each of the three stages: original journal article, press release, and news report.

This was accomplished by creating a detailed coding system, which reviewers could use to grade each source for the kinds of claims they were making in order to compare the hype level and notice any dif-ferences between the research, press releases, and news reports. To do this rigorously, the researchers focused on three specific areas:

Advice-giving (e.g. Eating chocolate may be beneficial for or Doctors should advise patients to). This was coded at four lev-els depending on how implicit or explicit the advice was.

Causal statements from correlational research (e.g. drinking wine might increase cancer risk... from a study that only observed cor-relations between these two things). This was coded at seven different levels, on a continu-um from statements that explicitly mentioned correlation, to those that were ambiguous (e.g. wine linked to cancer risk) to those that were explicitly implying causality (e.g. drinking wine increases cancer risk).

Conclusions phrased in human terms when research was done on animals, cells, or simula-tions (e.g. a pregnant womans stress levels concerning studies that were only done in rats). These were also coded at different levels depending on how implicitly or explicitly the conclusions were stated.

Even doctors in the ER test more for certain infections that have been getting heavy press coverage.

35

When coding for advice, the entire journal article, press release, or news story was examined. There was a total of 213 press releases (116 of these had news reports related to them), and 360 total news stories included.

Furthermore, for press releases and news stories, only the title and first two sentences were coded, since news writing is formulaic and often fol-lows an inverted pyramid structure, where the main claims are stated first. A sample of 182 press releases, 95 with news, and 261 news stories were used here. Only the abstract and discussion were coded for the original journal articles. Finally, when examining human conclusions from non-human studies, the main statements of 105 press releases (48 with news) and 115 news articles were coded, while only the abstract and discussion sections of journal articles were coded.

Two other areas were also examined to get a measure of how well-jus-tified the claims made in press releases and news articles actually were. This was done by noting which press releases and news articles had explicit caveats to their causal claims, advice, and inference to humans (e.g. The scientists who carried out the study emphasized that they could not say for certain...) and explicit justification for any of these three types of claims (e.g. even after taking into account the effect of extra body weight on blood pressure, there was still a significant link with sweetened drinks). In addition to these two areas, some other facts about the studies being reported were collected as well, such as duration, sample size, and sources of quotes.

The researchers explicitly took the peer-reviewed journal article as the baseline for the claims being made in press releases and news stories concerning the research. The original journal articles themselves were not fact-checked or examined to see if they were over-hyping anything. Which is not surprising, given that the authors of this study arent like-ly to be experts in dozens of biomedical and health research areas. So hype was measured by whether press releases and news articles were exaggerated compared to the original journal article.

If the original journal article itself contained hype, this study would not be able to detect it. But, if hype does exist in the original peer-reviewed research (and the authors of this study think its likely), then any hype

36

found in this study is likely an underestimate of overall hype, since hype originating in the peer-re-viewed scientific literature is not being taken into account. The researchers were also quite careful to make sure that their coding scheme was reproduc-ible. They did this by double-coding 27% of press releases and journal articles, and 21% of news sto-ries. They found that there was a 91% concordance rate in coding. The researchers then ran simulations to make sure that a 10% discrepancy in coding wouldnt affect their main conclusions, and it didnt.

Researchers examined press releases from the top 20 research universities in the UK to determine the origin of hype, or exaggeration, in media reports on new scientific findings. Exaggeration was determined by the presence of advice unsup-ported by scientific evidence and inappropriate extrapolation of evidence.

What were the findings?The researchers found that 40% of press releases con-tain more direct or explicit advice than the journal articles upon which they were based did. Similarly,

33% of press releases contained more strongly-word-ed claims of causation than the associated journal article warranted, and 36% of press releases inflated relevance to humans from non-human studies. So, it seems that press releases tend to add quite a bit of hype in all three areas studied.

It was also found that 36% of news reports contained more direct or explicit advice than the corre-sponding journal article. However, this does not necessarily imply that the journalists were the ones inflating the advice. The odds of exaggerated advice in news was 6.5 times higher when the press release contained exaggerated advice than when it didnt.

A similar pattern held for the two other areas of hype examined. While 39% of news articles were more strongly deterministic than what was warrant-ed by the associated journal article, the odds that the news had distorted causal statements was 19.7 times higher if the press release also contained distortions. Similarly, 47% of news articles reporting on non-hu-man studies contained exaggerations, with the odds of these exaggerations being 56% higher if the press releases contained similar distortions. As seen in

But the origin of the hype may go back even further than press releases. One study found that exaggerated claims often could be traced back to the abstract of the original journal article.

37

Figure 1, hype occurs in both press releases and news articles, and its much more likely to be present in news articles if the press releases also contain hype.

So there is hype but why the hype? The authors hypothesized that one possible motivation for exag-gerating claims in press releases could be to increase the chance that the press release will be picked up and reported by the news. But when the researchers looked at the data, they found that there was no statistically significant association between the percentage of press releases that had at least one news story published on their topic and whether or not the press release was hyped in any of the three ways this study examined. Also, the average number of news stories per press release did not vary between any of the three types of hype. So, whatever the motivation, hype in press releas-es is not actually correlated with more press coverage.

Finally, the researchers found that caveats about and jus-tifications for the claims being made were quite rare in both press releases and news stories, with at most only 17% of these claims having some sort of caveat or jus-tification (depending on the type of claim and source). There was no association observed between caveats and justification in press releases and news uptake, however. But there was a strong association between press releas-es having caveats and justifications about their claims and news sources having them as well.

The results of this study show that about 40% of press releases generated by the scientists contain the seeds of hype: exaggeration. Moreover, news reports based on hyped-up press releases tended to contain more hype and exaggeration than news reports based on press releases with cautionary statements.

Figure 1: How press release hype correlates with news hype

38

What does the study really tell us?The study tells us that both biomedical press releases and news reports contain exagger-ations that go beyond the peer-reviewed journal articles upon which they are based. Specifically, this study looked at exaggera-tions of three kinds. They found that 33% to 40% of press releases and 36% to 47% of news reports contain stronger inferences than were warranted by peer-reviewed journal articles, depending on the type of exaggeration.

This study also tells us that news reports were much more likely to contain these types of exaggerations if the associated press releases also had them. The odds of that news reports would contain exaggerations was 6.5 to 56 times higher if the associated press release also had such exaggerations.

While the population under scrutiny was press releases from Russell Group univer-sities in the UK, the authors explicitly state that they have no reason to suspect that this group of universities differs from other sources of press releases in any significant way, although this claim was not supported or argued for in the paper. If the authors are correct, these results should be generalizable to press reports outside of Russell Group universities and the news based on those reports.

Overall, these results are at least consistent with the hypothesis that a lot of the hype

Why correlation doesnt necessarily equal causation

Correlation just means that when you see one thing occur a lot,

another thing occurs a lot along with it. For instance, in this study,

there is a strong correlation between news reports that have hype

and press releases that have hype. Assuming the observed correla-

tion is actually true, then there are generally three explanations for

why it could occur:

A causes B: For example, it could be that press releases con-

taining exaggerations are indeed picked up by the media

and repeated.

B causes A: This would be that exaggerated news stories

about a piece of research lead to exaggerated press releases.

Assuming that press releases are written before the news

stories, though, this possibility is unlikely here, since causes

dont work backward in time.

Some third factor causes both A and B: perhaps journalists

are ignoring the press releases and working directly from

the journal articles and interviews with the researchers. And

perhaps the press releases are doing the same. Thus, the

source of the hype in this case would be the the original

researchers.

There is no way to differentiate between these three possibili-

ties from a correlation alone. However, one can narrow down the

possibilities through independent reasoning, as we did by using

temporal reasoning above. However, if possible, the best way to

establish causation is not through observational studies like this

one but through carefully controlled experiments where research-

ers actively intervene by changing only one variable and then

seeing what happens when compared to a control group. This is

part of the reason why randomized double-blinded, placebo con-

trolled trials are the gold standard in the biomedical sciences.

39

found in medical reporting originates not with the journalists reporting the news, but with the press releases written by universities. But before jumping to conclusions about causality ourselves, an import-ant caveat must be mentioned, one with which the authors of this study were also well-aware: this study was observational in nature, which means that although it can provide information on correla-tions, causality cannot be directly inferred.

However, there are several lines of reasoning to sug-gest that press releases are indeed a major reference for news articles. First, other retrospective and prospective studies have found that press releases influence news. Second, the researchers of this study took a look at the dates, quotes, and areas of focus in press releases and news reports, and found that these three areas seem to point to reliance on press releases by the media. Specifically, news stories were only selected if they were published within 30 days of the press release date.

Furthermore, the authors found that 87% of news articles selected were released within one day of the publication of the associated press release, leaving very little time for the journalists to do any addi-tional independent research. Also, 72% of quotes

found in news stories were also found in the press release, which also points to the medias reliance on press releases. Finally, study details such as sample size and study duration were very rarely reported in the news if the press release did not include similar details, but was usually reported in the majority of

news articles only if the associated press release had similar details.

So, while causation cannot be definitively established in this kind of obser-vational study, there is additional evidence that at least points in the direc-tion that exaggeration in press releases leads to exaggeration in associated news articles.

Finally, the authors found no statistically signifi-cant correlation between whether press releases had any of the three types of hype examined and whether and how much they were picked up by the

media. They also found that press releases including caveats and justifications didnt seem to affect news coverage. So, cautious, carefully-crafted press releas-es do not seem to be correlated with lower press coverage, and over-hyped press releases dont seem to get more press, either.

72% of quotes found in news stories were also found in the press release, which also points to the medias reliance on press releases.

40

Though the study was observational in nature and did not attempt to determine if the original journal article contained hype, it provides evi-dence to suggest press releases can significantly influence the way news is presented to the public. The research also suggests that hyped-up press releases get the same amount of coverage as press releases with cautionary statements, due to the news medias reliance on press releases.

The big pictureMedia is often blamed for hyping medical findings, but this study adds to a growing body of research which suggests that the fault does not lie solely with journalists. Many of the exaggerations found in the news were also found in the press releases on the same topic which preceded the news reports. Since press releases are often crafted in collaboration with scientists, both non-scientist writers at universities

and scientists themselves can take responsibility for more accurate biomedical reporting by crafting more careful press releases.

Journalists could in theory take more time to independently check facts and read the pertinent background literature, but the current journalis-tic culture has put a lot of pressure on journalists to produce more material in less time than ever, and so journalists may be forced to rely on easier and quicker sources of information, such as press releases and information from news agencies. And an entire journalistic culture can be very hard to change, particularly when its encouraged by a changing industry.

It may seem that there are a lot of troubling find-ings in this study. But because the authors found no incentive to hype up press releases (since more hype doesnt lead to more press), they end with a hopeful message: a relatively small handful of people in uni-

the current journalistic culture has put a lot of pressure on journalists to produce