Sunday, July 23, 2017

New Paper on Glucose Dose and Suppression of Endogenous Glucose Production

Our research group recently published a paper in Diabetes titled ‘The Effect of Ingested Glucose Dose on the Suppression of Endogenous Glucose Production in Humans’.  The paper has been uploaded as an electronic form and will be properly published in an issue in a few weeks.  The paper requires a subscription to read so I’ll summarise it below:

The aim of the study was to get a better understanding of the regulation of endogenous glucose production (EGP) under physiological conditions, as most of the research on EGP has come from studies using the euglycemic-hyperinsulinemic clamp (a method that involves infusing glucose and insulin to maintain normal glucose levels and high insulin levels over time)

We used the dual tracer technique to accurately measure EGP before and after different glucose doses (25g, 50g, 75g) in healthy young males.  The dual tracer involves two stable (non-radioactive) labelled glucose tracers* – one that is included in the glucose drink, and the other is infused at a variable rate** to mimic the fall in endogenous glucose.  The dual tracer technique can also estimate the rate of glucose entering the bloodstream from the drink (Ra) and the rate of glucose exiting the bloodstream into tissues (Rd)***

The main findings are:

  • Consistent with previous research, different glucose doses produced a dose dependent insulin response but near identical glucose responses****
  • The 25g, 50g and 75g glucose doses all resulted in a near identical suppression of EGP*****
  • Whereas Ra and Rd were dose dependent, such that higher glucose doses results in a greater rate of glucose absorption and glucose disposal 

In the discussion we mention:

  • These results indicate that the suppression of EGP is likely the first line of defence (so to speak) to deal with oral glucose loads, which makes sense as inhibiting EGP is faster and requires less insulin than stimulating significant glucose disposal in muscle.  Minimising insulin secretion may also be the reason why we have a glucose response at all to low glucose doses (as healthy people clearly have the capacity to have almost no glucose response if their bodies were wired up to do that)
  • The fact that larger glucose doses result in greater insulin responses but near identical glucose responses means that something else besides the in blood glucose must be responsible for the difference in the insulin response, probably incretins 

So what are the practical implications of this paper?  This paper is mainly aiming to address textbook physiology type questions so, like many scientific papers, isn’t very actionable on an individual level.  Although it should go towards easing some of the anxiety of those who think that getting a glucose response from a banana or potato (~25g carbs) means that they are 'metabolically broken’, as this is normal and doesn’t necessarily mean that double or triple the carbohydrate intake will double or triple their glucose response.

* These are 1 and 2 atomic mass units heavier than regular glucose and this difference can be detected using gas chromatography-mass spectrometry (GC-MS)

** This avoids potential issues in calculating EGP such as paradoxical increases in EGP shortly after a glucose load and negative EGP during the euglycemic-hyperinsulinemic clamp.  There is a slightly amusing example of this in the following paper, where they find that one group of mice has negative EGP during the clamp, while the other group of mice has ‘impaired’ EGP because theirs is only ~0

*** Although this is more accurately done using the triple tracer technique, which involves infusing a second tracer at a variable rate to mimic the Ra

**** Previous studies have consistently found that different glucose loads (33g vs. 66g vs. 100g, 50g vs. 100g, etc*) produce a dose dependent insulin response but near identical glucose responses in healthy people.  In contrast, people with borderline or overt impaired glucose tolerance (2h glucose ≥ 7.8 mmol/l) have greater glucose responses from larger doses of glucose.  This begs the question, how is it that in healthy people, larger glucose doses don’t produce a greater glucose response?

***** EGP was suppressed by ~55% unlike the near complete suppression that is common from when using the euglycemic-hyperinsulinemic clamp.  This is because under physiological conditions insulin inhibits glycogenolysis but not gluconeogenesis, whereas the prolonged supraphysiological hyperinsulinemia from the clamp is necessary to persistently inhibit gluconeogenesis

Sunday, July 2, 2017

The AHA's Presidential Advisory on Dietary Fats and Cardiovascular Disease

The American Heart Association (AHA) recently released a presidential advisory on dietary fats and cardiovascular disease (CVD) [1].  As you would expect from the AHA, they claim that “…randomized controlled trials that lowered intake of dietary saturated fat and replaced it with polyunsaturated vegetable oil reduced CVD by ≈30%, similar to the reduction achieved by statin treatment” and conclude that “lowering intake of saturated fat and replacing it with unsaturated fats, especially polyunsaturated fats, will lower the incidence of CVD”.

I recently published a meta-analysis on the randomised controlled trials (RCTs) that replaced saturated fat (SFA) with polyunsaturated fat (PUFA) to see if this intervention reduces the risk of coronary heart disease (CHD) (blog) [2].  I concluded that “available evidence from adequately controlled randomised controlled trials suggest replacing SFA with mostly n-6 PUFA is unlikely to reduce CHD events, CHD mortality or total mortality”.  So not surprisingly I disagree with the AHA’s presidential advisory and will explain why below.

Speaking of which, as my meta-analysis was published about a month ago some people asked if I thought the presidential advisory was a response to that.  I doubt it.  The authors would have to have been made aware of my meta-analysis (without there being significant media coverage), organise themselves, then write this paper and get it past review in less than month.  They also didn’t cite it my meta-analysis, but one could argue that could be because of other reasons.  I think it’s more likely that this is a response to the updated meta-analysis by Ramsden et al in April 2016 after recovering data from the Minnesota Coronary Survey [3].

In this post I’m going to focus on the presidential advisory, mostly the RCT evidence, and not the media coverage.  From what I’ve seen, the media coverage badly misrepresents the evidence.  They claim SFA is bad and draw special attention to coconut oil, claiming that coconut oil is worse than butter because it has more SFA.  This is a position you can only take if you ignore that SFA also increases HDL-C (as a result SFA doesn’t significantly affect the total-C:HDL-C ratio) and HDL-C is associated with a lower risk of CHD/CVD, ignore meta-analyses of observational studies, and ignore the almost total absence of RCTs for anything related to SFA and CHD/CVD other than replacing SFA with PUFA [1].  The conventional position is not that SFA is bad, but that replacing SFA with PUFA will reduce the risk of CHD/CVD.  Replacing SFA with MUFA is not a particularly defensible position unless you only look at blood lipids and ignore meta-analyses of observational studies.

The AHA’s Selection Criteria

The claim that RCTs where SFA was replaced with PUFA reduced the risk of CVD by ~30% doesn’t come from an earlier meta-analysis, but is from a meta-analysis the AHA did in the presidential advisory.  The AHA looked at previous systematic reviews and meta-analyses for RCTs (Mozaffarian et al 2010, Hooper et al 2015, and Chowdhury et al 2015) and applied the following inclusion criteria to select their core trials:

  1. Compared high SFA with high PUFA
  2. Didn’t include trans fats (TFA) as a major component
  3. Controlled the dietary intake of the intervention and control groups
  4. Had at least 2 years of sustained intake of the assigned diets
  5. Proved adherence by objective biomarkers such as serum cholesterol or blood or tissue levels of polyunsaturated fatty acids
  6. Collected and validated information on cardiovascular or coronary disease events 

These criteria are very reasonable.  (1) and (6) are actually essential.  In my meta-analysis I tried to account for (2) and (3) with my ‘adequately controlled’ and ‘inadequately controlled’ categories, and I reported changes in serum cholesterol for every trial (5).  The rationale for (4) makes some sense and the AHA say “trials of serum cholesterol–lowering agents show that a reduction in coronary heart disease (CHD) incidence occurs with a lag of 1 to 2 years”.  However, the primary reason given is that “the 2-year minimum duration is that changes in polyunsaturated fatty acids very slowly equilibrate with tissue fatty acid levels; it takes ≈2 years to achieve 60% to 70% of the full effect”.  This isn’t necessary for adherence and isn’t relevant for the proposed mechanism.  The proposed mechanism for PUFA reducing the risk of CHD is by lowering LDL-C, which short term feeding studies and the National Diet Heart Study found maximally occurs within days or a couple of weeks [4]

After applying their criteria, they included the following trials as part of their core trials:

  • Los Angeles Veterans Administration Trial (LAVAT)
  • Oslo Diet Heart Study (ODHS)
  • Finnish Mental Hospital Study (FMHS)
  • Medical Research Council Trial (MRCT) 

The Core Trials

When briefly summarising each trial, the AHA mentions key facts like the number of participants, the basics of the interventions and the number of events and deaths in each group.  However, the AHA doesn’t address the major issues in most of those trials which are discussed in my recent meta-analysis and I’ll also mention below:

Los Angeles Veterans Administration Trial

I see there being two key issues in LAVAT.  Firstly, that the researchers mostly omitted conventional margarines and hydrogenated shortenings (major sources of TFA) from the high PUFA diet [5].  Secondly, the α-tocopherol (vitamin E) intake in the high SFA group was 9.4-fold lower than the experimental group (22.6 mg vs. 2.4 mg) [6] and was deficient, being only 16.0% of the current RDA (15 mg) [7].  These issues were not reported in the monograph [8], which the AHA cites, but in the other papers that I cited here, which can be easily found by looking at the Cochrane meta-analyses (the AHA cites the one by Hooper et al 2015). 

There’s the argument that smoking is a confounder in LAVAT.  The high PUFA and high SFA groups had similar number of low or non-smokers, but despite randomisation and a large number of participants, the experimental group had more moderate smokers and fewer heavy smokers [9].  My guess at a reasonably appropriate way to account for this is to take the incidence of endpoints in person years stratified by smoking status, weight them according to the total number of participants in each smoking stratification and sum that all together (see table below).  This results in the effect of higher moderate smokers and lower heavy smokers probably cancelling each other out to a large extent.  Interestingly, there was an interaction between smoking and the high SFA, vitamin E deficient control diet such that much of the increased risk of CHD/CVD in the high SFA group was in the moderate and heavy smokers (supporting an oxidative stress model of CHD)

Total incidence in person years
Average weighted incidence in person years

Hard EP

Oslo Diet Heart Study

The high PUFA group received a multifactorial dietary intervention that included advice to increase fish, shellfish and whole plant food consumption, advice to moderate sugar consumption and restrict shortening (a major source of TFA, and hydrogenated marine oils were a major source of fat in Norway at the time of trial), and the high PUFA group received sardines canned in cod liver oil [10].

This isn’t mentioned in the main publications of the trial [11] [12] ([11] is the one cited by the AHA), but is mentioned in the monograph [10], in an online version of the relevant chapter from the monograph [13], by Ramsden et al in the 2010 and 2016 versions of their meta-analysis [14] [15], and by Hoenselaar is his review [16] (and no doubt others have done the same in peer-reviewed journals).

ODHS should not have been included with the core trials for failing to meet the TFA (2) and controlled diet (3) criteria.  The AHA excluded DART and STARS because SFA was replaced with PUFA and carbohydrate, but included ODHS despite all the other dietary variables that are likely more meaningful than replacing SFA with carbohydrate, which the AHA doesn’t even think affects the risk of CHD/CVD.

Finnish Mental Hospital Study

The AHA includes FMHS as part of their core trials and refers to the core and non-core trials as ‘randomised controlled trials’.  However, FMHS isn’t a randomised trial.  Some have suggested that it’s a cluster randomised trial (in this case that the hospitals, rather than the patients, were randomly allocated to go on the high SFA or high PUFA diet first).  However, there is no mention of this in any of the publications from the trial.  Even if the researchers did flip a coin to randomly allocate the hospitals, a cluster randomisation with 2 clusters is probably quite inadequate.  Hooper et al 2012 [17] and 2015 [18] in their meta-analyses required at least 6 clusters and excluded FMHS for that reason.

A couple of issues in FMHS demonstrates that how 2 clusters can be inappropriate.  The control group in FMHS received more of a cardiotoxic antipsychotic drug called thioridazine in hospital N (0.82 vs. 1.79) and slightly less in hospital K (0.43 vs. 0.14), which averaged to an overall greater use in the control group (0.63 vs. 0.97) [14] [19].  Also, the participants in the control group remained in the hospitals longer than those in the experimental group, which led to an overestimation of the effect size also points to inadequate randomisation.  Fortunately the AHA correctly used the RR from incidence by age-adjusted person years to account for this.

In addition, due to the more detailed dietary information provided in FMHS [19], Ramsden et al [14] was able to estimate TFA intake in both of the groups and found TFA intake to be lower in the experimental group in both hospital K (0.0 vs. 2.0% of total energy intake) and hospital N (0.2 vs. 0.6% of total energy intake).

Medical Research Council Trial

I don’t see there being any major issues in MRCT and is the only one of the four core trials that I categorised as adequately controlled in my meta-analysis.  Some minor issues include: that the methods used to reduce SFA intake in the high PUFA group included forbidding “butter, other margarines, cooking-fat, other oils, fat meat, whole milk, cheese, egg yolk, and most biscuits and cakes”.  This was very would be expected to reduce TFA intake in the high PUFA group to some degree and these methods were very common in the diet heart trials.  And also that the participants were instructed to consume at least half of the soybean oil unheated and most of the participants achieve this by drinking the oil produced [20].  This doesn’t represent the way in which oils are usually used, which is for cooking, and cooking causes heat damage to oils.

The Non-Core Trials

The AHA considered the following to be non-core trials:

  • St Thomas Atherosclerosis Regression Study (STARS)
  • Diet And Reinfarction Trial (DART)
  • Houtsmuller et al (HDAT)
  • Rose Corn Oil Trial (RCOT)
  • Minnesota Coronary Survey (MCS)
  • Sydney Diet Heart Study (SDHS) 

Reason for exclusion
My comment
Replaced SFA with PUFA and carbohydrate
Carbohydrate was 10% lower (234.2 vs 267.1 g/d).  Doesn’t address what are likely to be far more meaningful dietary changes besides SFA, PUFA and carbohydrate (similar to ODHS, see my paper)
Replaced SFA with PUFA and carbohydrate
Carbohydrate only increased from 44% to 46% of total energy intake
Researchers were not blinded
This wasn’t one of their criteria, but it was a very badly reported study with lots of issues and unknowns
Small number of participants (N =54) and short duration
Small number of participants wasn’t one of their criteria and wasn’t mentioned when discussing STARS.  Debatable whether it should have been included with the AHA criteria as it did have mostly 2 year follow up
Average duration 384 days, withdrawals, intermittent treatment
Could have used data from participants who remained in the study for ≥ 1 year as they had an average of 2.9 years on the diet
Replaced SFA with PUFA and TFA
See below

Sometimes the exclusion of these trials was justified, sometimes it was not.  On balance the non-core trials are unfavourable for the diet heart hypothesis.  RCOT, MCS and SDHS are unfavourable, DART was pretty neutral, and while STARS and HDAT were favourable, they are both quite small.  So the exclusion of the non-core trials helped the AHA get an impressive RR when conducting a meta-analysis on their core trials, which is part of what makes me wonder whether the criteria were designed to get such a favourable result.

Also, in relation to MCS, the AHA also said “another concern is the use of lightly hydrogenated corn oil margarine in the polyunsaturated fat diet. This type of margarine contains trans linoleic acid, the type of trans fatty acid most strongly associated with CHD”.  However, this ignores several points made by Ramsden et al about TFA intake in MCS (see below):

“Because the trans fatty acid contents of MCE study diets are not available, one could speculate that the lack of benefit in the intervention group was because of increased consumption of trans fat. Indeed, in addition to liquid corn oil the intervention diet also contained a serum cholesterol lowering soft corn oil polyunsaturated margarine, which likely contained some trans fat. The MCE principal investigator (Ivan Frantz) and co-principal investigator (Ancel Keys), however, were well aware of the cholesterol raising effects of trans fat prior to initiating the MCE. Moreover, Frantz and Keys previously devised the diets used in the institutional arm of the National Diet Heart Feasibility Study (NDHS), which achieved the greatest reductions in serum cholesterol of all NDHS study sites. Hence, it is highly likely that this experienced MCE team selected products containing as little trans fat as possible to maximize the achieved degree of cholesterol lowering. Perhaps more importantly, it is clear from the MCE grant proposal that common margarines and shortenings (major sources of trans fat) were important components of the baseline hospital diets and the control diet (but not the intervention diet). Thus, confounding by dietary trans fat is an exceedingly unlikely explanation for the lack of benefit of the intervention diet.” [3]

It seems that the AHA uses TFA and multifactorial dietary interventions as justification to exclude trials when it’s convenient and ignores these issues when it’s not.

Should the Sydney Diet Heart Study be dismissed so easily?

Some have suggested that the high PUFA group in SDHS had a higher intake of TFA due to the use of Miracle Margarine, which has been suggested to have been rich in TFA at the time of the trial [21].  However, Ramsden et al [22] [23] has provided some arguments suggesting that TFA is likely to be a major factor in SDHS.  The AHA ignores this debate and uncertainty and confidently states that the study was comparing a high SFA diet with a high PUFA and TFA diet because the high PUFA group was given a margarine high in TFA.

“The Sydney Diet Heart Study showed that using a margarine rich in trans unsaturated fat to replace saturated fat increased CHD events, confirming similar adverse results in epidemiological studies.”

I’m going to play the AHA’s game and assume for this section that the high PUFA group in SDHS did in fact have a higher intake of TFA, but *spoiler alert*, they’re not going to like to outcome

To get a rough indication of the TFA intake in the high PUFA in SDHS let make a few assumptions and rough calculations.  (1) Assume that the Miracle Margarine used in SDHS was composed of 25-40% TFA, which based on study looking at the TFA content of safflower margarines of that time that was cited by Gutierrez* in her rapid response [21], and (2) assume that the high PUFA group replaced roughly half their original fat intake (which I think is a reasonable estimate given the change in SFA intake) with similar amount of miracle margarine and safflower oil.  Therefore Miracle Margarine provided about 9% of total energy intake and TFA intake from Miracle Margarine would be 2.25-3.60%.  Ramsden et al estimated that TFA intake in the high SFA group was 1.6% [14].  The high PUFA group were advised to restrict common margarines and shortenings (which are major sources of TFA), but let’s assume they were 50-75% compliant.  Therefore, a generous estimate of TFA intake would be 1.60% in the high SFA group and 2.65-4.40% in the high PUFA group.

High SFA group
High PUFA group
Follow up
Follow up
Total fat (%)
SFA (%)
MUFA (%)
PUFA (%)

The RR in SDHS was 1.57 for CHD mortality and 1.49 for total mortality.  The AHA cites two analyses of observational studies which found using a substation analysis that replacing 2% of energy from SFA with the same energy from TFA increased the risk of CVD mortality by 5% [24] and 16% [25], and total mortality by 16% [25].  Even assuming a TFA intake of 4.40% in the high PUFA group, using the 5% increased risk of CVD mortality with 2% SFA > TFA substitution results in an amended RR of 1.50 for CVD mortality.  So I’ll continue to be generous and use the 16% increased risk to calculate an altered RR in the table below.  The RR is still comfortably > 1.00 whether you assume a TFA intake in the high PUFA group of 2.65% or 4.40%.

TFA intake in high PUFA group
Altered RR for CVD mortality
Altered RR for total mortality
Assume 2.65%
Assume 4.40%

In summary, in the bottom row in the table above I’ve granted that Miracle Margarine is 40% TFA (upper-end), assumed it contributed about 9% of total energy intake, granted that the high PUFA group were only 50% compliant in restricting common margarines and shortenings, and finally used the most favourable of the AHA’s cited data for SFA > TFA substitution.  After doing all this to alter the RR to account for potential differences in TFA intake, SDHS is still an unfavourable study for the diet heart hypothesis!!!  This should be a serious wakeup call to the AHA and other diet heart advocates.

* In her rapid response, Gutierrez says “the PUFA-supplemented (intervention) group may have been provided with atherogenic trans fat, and the investigators cannot prove otherwise” [21].  However, this argument cuts both ways – ‘the high PUFA group may have had a lower or similar intake of TFA and critics cannot prove otherwise’ – and is similar to the stupid argument of ‘you can’t prove that God doesn’t exist’

Total Mortality

The AHA only conducted one meta-analysis that pooled the primary end-points from each of their core trials, which includes:

  • LAVAT: CVD mortality
  • ODHS: the number of participants with total CHD events (doesn’t count multiple events in the same person more than once, and includes ‘soft’ events like angina)
  • MRCT: the number of participants with total CHD events
  • FMHS: CHD mortality 

The figures here are all correct, even appropriately using the age-adjusted person years in FMHS, which is actually surprising given the previous meta-analyses

A major issue here is that the AHA doesn’t conduct a meta-analysis for total mortality.  This is important because whole point of trying to reduce your risk of CHD or CVD is to reduce your overall risk of morbidity and mortality.  One can argue that the AHA is justified to focus on CHD/CVD since that is the purpose of the association, but what good does it do to reduce your risk of CHD/CVD if doing so increases your risk of non-CHD/CVD morbidity and mortality, such that you would be no better off?

This is particularly relevant for the AHA’s selection of trials.  While the high PUFA group in LAVAT and FMHS had a lower risk of CHD and CVD mortality (RR = 0.80 and 0.59), they had a near identical risk of total mortality (RR = 0.98 and 1.01), because non-CHD/CVD mortality was higher [8] [26].  As those trials were both many times larger than ODHS and MRCT, and therefore have a much larger weighting, the pooled RR of the AHA’s core trials for total mortality is 0.98 (CI = 0.90-1.07, P = 0.65).

So will replacing SFA with PUFA reduce your risk of dying?  Even with the AHA’s selection of trials and even while ignoring the major issues in LAVAT, ODHS and FMHS, the answer is still no.

Sunday, May 21, 2017

My Meta-Analysis on Saturated Fat, Polyunsaturated Fat and Coronary Heart Disease

I just published a paper in Nutrition Journal titled ‘The effect of replacing saturated fat with mostly n-6 polyunsaturated fat on coronary heart disease: a meta-analysis of randomised controlled trials’.  Nutrition Journal is an open access journal and the paper shouldn’t be too technical so I encourage you to read it but I’ll also summarise it below:

I begin in the introduction by discussing the effect that saturated fat (SFA), monounsaturated fat (MUFA) and polyunsaturated fat (PUFA) have on cholesterol levels and how this formed the basis of the diet heart hypothesis, which is that replacing SFA with PUFA would be expected to reduce the risk of coronary heart disease (since MUFA didn’t affect total cholesterol it was largely ignored at this time).  Several clinical trials were conducted, mostly in the 1960s and 1970s, to test the diet heart hypothesis.

More recently, several meta-analyses have pulled the results of the diet heart trials and most came to the conclusion that replacing SFA with PUFA would reduce the risk of coronary heart disease.  However, there is a lot of controversy surrounding the diet heart hypothesis, and it is an important topic since the advice to reduce saturated fat is very influential in conventional dietary advice.  So I thought it would be useful to do my own research.

I researched the clinical trials included in the previous meta-analyses and found that the dietary advice or the foods given to the participants in the trials would be expected to result in the high PUFA group having a lower intake of trans fats (to varying degrees) compared to the high SFA group.  I also found other confounding variables such as two trials where the high PUFA group received a multifactorial dietary intervention, one trial where the high SFA group was given a vitamin E deficient diet, and another trial where the high SFA group received more antipsychotic medication that was found to be cardiotoxic.  These issues were rarely discussed or taken into account by the previous meta-analyses.

Therefore I conducted my own meta-analysis by including the trials in previous meta-analyses and then categorising them as ‘adequately controlled’ or ‘inadequately controlled’ based on the degree of confounding variables in each trial.  I considered the adequately controlled trials to be those that most accurately tested the effect of replacing SFA with PUFA, while the inadequately controlled trials to have too much of a difference between the groups to be a valid test of replacing SFA with PUFA.

The main findings are:

  • The results from the adequately controlled trials suggest that replacing SFA with PUFA is unlikely to increase or decrease the risk of coronary heart disease or all-cause mortality
  • This is the case regardless of whether the Sydney Diet Heart Study is included or excluded
  • The suggestion of benefits reported in earlier meta-analyses is due to the inclusion of inadequately controlled trials

In the discussion, I mention that there are issues in drawing conclusions from risk factors (such as cholesterol) and observational studies, and that the issues inherent in these types of studies are likely due to them having discordant results with this meta-analysis.  I also discuss that similarly there are issues in drawing conclusions from nutrients (like SFA) to foods, as foods are made up of many nutrients and other substances that may affect outcomes you’re interested in, such as the risk of coronary heart disease.

So what are the implications of this paper?

  • Saturated fat is unlikely to increase the risk of coronary heart disease
  • Replacing saturated fat with polyunsaturated fat (such as butter with margarine or vegetable oil) is unlikely to increase or decrease the risk of coronary heart disease
  • Results from observational studies and enthusiasm from discovering a new risk factor or discovering a target for an existing risk factor should be tempered with a degree of caution and scepticism until they are tested
  • Dietary guidelines and public health policy should focus on other things that are more likely to be effective at preventing coronary heart disease

If you like talks, the content of the meta-analysis is similar to a talk I have in 2015 at the Ancestral Health Society of New Zealand’s (AHSNZ) international symposium in Queenstown.  You can view the talk here and the other talks here.  There were a lot of great speakers there talking about a diverse range of topics such as nutrition, exercise, behaviour change, sustainability, active transport and mental health.  So it shouldn't be too hard to find something that interests you.

AHSNZ is holding another international symposium in October at Queenstown which you can find out more about the symposium here.  I’ll be speaking again, this time about insulin resistance, which is the focus of my PhD (so this paper is very unlikely to count towards my PhD and has mostly been a weekend project) and will be a focus on the blog over the next several months.

Tuesday, April 18, 2017

Public Health Strategies Part 4C: Bans

Bans on Unhealthy Foods

In earlier blog posts I discussed taxes and subsidies as public health strategies, mainly related to the current more popular calls to tax sugar sweetened beverages (SSBs) and subsidise fruit and vegetables.  In researching those posts I came to the conclusion that both strategies have some potential to influence consumer choices and therefore population health when implemented broadly (e.g. tax ‘extra’ foods instead of just SSBs) and strongly (e.g. 50-100% tax rather than 10-20% tax).  However, the narrow range of targets combined with the weak tax/subsidy is unlikely to substantially improve population health.  In addition, assuming consumption wouldn’t be affected, a 100% tax on SSBs would only increase the average household budget of low income earners by 1% [1], which is something that even most low income earners in an affluent country like Australia could shrug off (not to mention those on higher incomes).  This raises the question of whether more aggressive market controls such as bans on the marketing, sales and/or possession (the latter has actually been proposed [2]) of foods like SSBs are required to truly reduce the prevalence of obesity and metabolic disease

Bans can apply to the ability of producers to market a product or sell a product and the ability of consumers to possess a product.  Bans are not necessarily universal, it may only apply to certain times (such as advertising on TV when children are most likely to be watching and bands on selling alcohol after a certain hour of the night), places (such as bans on selling SSBs in schools) and very specific products (such as the proposed New York City ban on big gulps).  Non-universal bans on selling products are intended to lower consumption by reducing convenience, while universal bans on selling products or being in possession of them (such as illicit drugs) are intended to eliminate consumption

Like any other policy, putting a ban on unhealthy foods is likely to have unintended consequences and these will depend on what kind of ban is implemented.  A universal ban on unhealthy foods with little to no redeeming qualities like SSBs is unlikely to happen, and it could result in unintended consequences similar to the war on drugs and the American prohibition on alcohol, in addition to losing an opportunity for tax revenue.  Alternatively, a non-universal ban that reduces convenience can be easily circumvented by a determined consumer who can plan in advance to take extra drinks with them to consume late at night, to take SSBs with them to a school/university campus and to simply order to small size drinks instead of the big gulp.  Of course, not everyone will be the determined consumer and the point is that small changes in convenience can have large impacts in behaviour.

The rationale of banning something like SSBs (as opposed to TFA, asbestos, etc) is to save people from themselves, but is it the government’s job to do this?  Some people would object to the nanny-statism and demand the freedom to eat/drink what they choose, although if you’re asking the government to pay for your healthcare one could argue that you are implicitly trading some liberty for security.

Bans to reduce the consumption of unhealthy foods by children are another story.  Unlike adults, we do not hold children completely legally responsible for their actions (which scales with age of course), and so there is an argument to be made to protect children from themselves, or from their parents*.  With health issues like obesity and tooth decay being an issue among some young children, perhaps banning the sale of SSBs for children or to children (SSBs would be an 18+ substance like cigarettes and alcohol) would be help to reduce these issues.  Unfortunately, my look at the research at the moment suggests that just banning sodas at high schools results in an increased consumption of other SSBs [3], while banning the sale of all SSBs at middle schools only seems to reduce consumption at school and doesn’t appear to reduce overall consumption [4].  Perhaps there would be more success at primary schools, but also it seems people will get their sugar fix no matter what, and so the narrower the ban, the less likely it is to be effective at all

* Bit of a rant: sorry Guardian author, if a 2 year old needs 20 teeth removed due to tooth decay, that’s not an issue with oral health prevention [5], that’s child abuse from parents who don’t sufficiently care.  This is similar to issue that sometimes comes up where children fed a vegan diet are malnourished, leading to an Italian proposal to jail parents feeding young children a vegan diet [6].  I don’t want to necessarily jail vegan parents or vilify parents who give their children SSBs, just when there is evidence of harm, because it’s about the outcome (in the absence of losing the genetic lottery), not the methods to get there.  Vilifying ‘wrong’ methods regardless of any feedback from outcomes could lead to a hideously broad application of that Italian proposal (‘oh, you’re feeding your child a low carb diet?  That’s against the dietary guidelines.  It doesn’t matter that there’s nothing wrong with your child, you’re going to jail’) and leads to dogmatically sticking with the ‘right’ method, such as the dietary guidelines, rather than updating your methods based on the feedback from outcomes in research and clinical practice (outcome based medicine > evidence based medicine).  If you think I’m being harsh, consider how you would feel if you had your teeth removed, got type 2 diabetes or suffered developmental issues before you had the chance to make your own choices or because your mother smoked and drank heavily while pregnant with you.  The right as a parent to bring your child up their own way should not trump their responsibilities to bring them up well.

** The Torba province of Vanuatu is aiming to impose strong restrictions on junk food while promoting locally grown, organic food [7].  It will be interesting to see how that goes.  They have a fairly special advantage from being isolated and a small community which might lead to it working out very well (but good luck trying to implement the same in Australia, etc)

*** Alternatively if you see obesity and lifestyle diseases as an product of market failure, you could just ban capitalism and adopt socialism or communism, which is proving to be a really effective policy at reducing obesity in Venezuela at the moment :p

Ban on Trans Fats

This leads me to the FDA ban on partially hydrogenated oils (PHOs).  In the US, the FDA has ruled that PHOs are not generally recognised as safe for use in human food.  The response to this seems to generally be positive, with a small number of doubts and concerns coming from some libertarians.  I agree that banning PHOs won’t have much effect as consumption of PHOs is quite low and they aren’t actually that bad.  In Australia in 2009, the average intake of TFA is 0.5% of total energy intake, with 60-75% coming from animal foods, so only 0.125-0.2% from PHOs, though intakes in the US seem to be quite a bit higher [8].  A recent study using the Nurses’ Health Study and the Health Professionals Follow-up Study found that each 1% increment of total energy intake from TFA was associated with just a 10% increase in total mortality (the effect would be diluted as TFAs from animal foods are pretty neutral, but still) [9]

Even though at current intakes, PHOs aren’t that bad, they are still a great example of something that should be banned.  (1) PHOs are really only convenient for the food industry and consumers don’t seek them out (copha is disgusting) (no black market).  (2) They will likely just be replaced with SFA rich fats/oils (no negative unintended consequence for consumers, just a drop in sanity from the Heart Foundation).  (3) It’s unlikely that this will translate to regulation on ‘unhealthy’ foods like red meat because PHOs don’t have any nutritionally redeeming qualities, whereas red meat certainly does (particularly when most people could use more protein, iron and zinc, etc) (so the libertarian concern of increasing government regulation is unlikely)

Some objections to a ban on PHOs is the false dichotomy that SFA is a larger issue [10], concerns that the PHOs would be replaced with SFA rich fats/oils, and concerns that that enforcing a ban would be too hard [11].  The ban would be hard to enforce 100%, but would be just as hard to enforce accurate labelling [12] which seems to be the Australian Heart Foundation’s preferred method, at least initially.  Mandatory labelling hasn’t been implemented because TFA intake in Australia is less than the WHO target of 1% [13], but I still think consumers should still have a right to know

Down the Conspiracy Theory Rabbit Hole

Given that the FDA is banning PHOs due to their adverse health effects and having no redeeming qualities, why aren’t cigarettes also going to be banned?  Cigarette smoking is associated with far worse health outcomes than PHOs and also results in second-hand smoke and more litter.  About 15% of Australians still smoke despite decades of health messaging stating the adverse effects of smoking, cigarettes been heavily taxed, advertisements for cigarettes been banned, plain packaging laws, and graphic imagery depicting some potential consequences of smoking.  Surely banning the sale and possession of cigarettes entirely would be one of the most productive policies for public health (and even poverty), and potentially a popular policy for the vast majority who don’t smoke (cleaner air and environment) and especially for ex-smokers.  One could speculate that the reason why governments haven’t banned the sale and possession of cigarettes is that the tax revenue they get from them exceeds the healthcare costs from smoking.  When looking at the costs of smoking, you see stats along the line of ‘smoking costs $X’, but these stats are probably irrelevant if not compared with the costs of not smoking.  I came across a study from the Netherlands (so could be different elsewhere) that predicted that the healthcare costs of smokers and of people with obesity are actually lower per person because they live shorter lives [14], and haven’t yet seen another study with a similar type of analysis.  Another consequence of smokers and people with obesity or diet and lifestyle diseases tending to die younger is that on average you would expect them to cost the government less money on pensions.  I wonder if these reason are a factor why many governments aren’t aggressively addressing diet and lifestyle diseases.  After all, even if the average life expectancy suddenly increased by 10 years because the government implemented rigourous public health policies, people will still want their pension at 65 and moving the pension age up to 75 might be more politically unpalatable than getting people to stop having pizza and coke for dinner.  (That might be a bit of a stretch, but I think you could make a reasonable case for pensions being a factor (alongside more major ones like debt) behind the constant drive for economic and therefore population growth and for the migrant crisis in Europe, but that’s getting quite off topic).  But this is hardly a flawless conspiracy theory given that government had a major role in the reduction in smoking, the investment governments make in medical research, and the demographics that receive most of the healthcare expenditure (i.e. not working)

Sunday, February 26, 2017

Public Health Strategies Part 4B: Subsidies

In an earlier blog post I discussed taxation as a public health strategy, particularly related to the proposed tax on sugar sweetened beverages (SSBs) in Australia.  In this post I’m going to look at the opposite of taxation: subsidies.

Putting a tax on unhealthy foods would generate extra tax revenue and so the question becomes whether the government should reduce taxes in other areas (or use it to help pay off national debt) or put that extra revenue into something, and if so, what?  Generally I have seen calls to tax unhealthy foods being coupled to calls to subsidise health foods, like fruit and vegetables (F&V).  There are a few rationales for subsidising healthy foods like fruit and vegetables:

  • Reducing the cost of fruit and vegetables would increase the consumption of them and displace unhealthy foods, which will ultimately improve population health and reduce healthcare costs
  • Coupling a fruit and vegetable subsidy to a tax on unhealthy foods (like sugar sweetened beverages) is also a means to reduce the increase to cost of living as a result of the tax, provided people purchase fruit and vegetables 

However, there may be a few problems if a health food subsidy was put in practice

  • A recently published Australian modelling study estimated that a F&V subsidy ($0.14 per 100g of fresh and preserved F&V*) would increase F&V consumption by 42g (a serving of fruit and vegetables is considered 150g and 75g respectively).  However, it was estimated that the subsidy would also increase sodium consumption by 48mg and energy consumption by 236kJ (56.6 calories), because “however, using price subsidies or discounts as an incentive to purchase more fruits and vegetables may have the effect of increasing real income available to buy food, including unhealthy products, and could therefore lead to an overall increase in dietary measures such as saturated fat, sodium, or total energy intake”**.  As a result, their model predicted that a F&V subsidy would actually have adverse health outcomes [1].  The major benefit of food taxes is that they generate revenue [2].  This revenue should go towards initiatives that are at least cost effective, but with a F&V subsidy there’s this study says there’s a 89% chance that it wouldn’t be.  Not a great policy
  • A subsidy on F&V isn’t likely to offset the increase in cost of living from a tax on unhealthy foods such as SSBs.  The estimates show that there isn’t going to be much change in behaviour.  So the people who are already low SSB consumers and high F&V consumers are the ones who will benefit.  This got me thinking if the promotion of taxes + subsidies in some people (not all) may at least partly driven by financial self-interest, but you can defend this motivation in countries with socialised healthcare.  (By the way, my diet is very rich in F&V, with no SSBs and low added sugar, so I would benefit a lot from such policies) 

In my opinion as a stingy student currently on an unflattering income, many F&V are already very cheap as there’s a lot you purchase for < $4-5 per kg or much less.  I think the reason why so many people don’t consume the recommended intake of F&V [3] is because other foods simply taste better, the structure of their habitual meals is not conducive to eating many F&V (cereal for breakfast, sandwiches for lunch, etc), and they don’t value/are empowered about their health enough to change.  When people say cost is important, they are comparing apples with apples, and not apples with muffins.  The apple wins easily on cost, but the café bought muffin wins on palatability and reward, and because most Australians have that money to spend, that’s what most people choose

A tax on unhealthy foods should be coupled with a subsidy or health initiative that is actually cost effective in itself.  An idea circulating around AHSNZ is that a tax on SSBs could be coupled to subsidy on dental health or free dental for children.  This would disproportionately benefit lower income families who are less likely to have private health insurance, see the dentist less often and more likely to have worse diets.  It is also likely to be more cost effective as healthcare spent in younger people has a greater return on investment, and dental health is one of the major health issues for children, and one (rampant tooth decay) that is potentially not reversible unlike obesity and type 2 diabetes.  Some people may be against the government using taxes and subsidies to save people from themselves, but may concede that something should be done as tooth decay is so common in children [4].  I would still like to see an estimate of the cost effectiveness of any policy, as good intentions are not a guarantee of good policies

* For example, if a fruit or vegetable was priced at $4 per kg, this subsidy would cover 35% of the costs.  This method of subsidising has a greater effect on cheaper F&Vs such that it wouldn’t be practical as very cheap F&Vs like carrots would be almost free.  In fact, at the time of writing this Coles has a special on carrots at $1.20 per kg, so they would be paying the customer to purchase them, pretty crazy! (but don’t forget that F&V are expensive and cost of healthy foods is a limiting factor in population health…)

** I think this point is debatable.  Paying less for F&V would result in consuming more F&V and this may have the opposite effect on calorie intake as F&V are more satiating than most foods per calorie.  In addition, the sodium > blood pressure data they used was based on a large effect from observational studies [5] rather than the small effect in RCTs [6], although sodium could be a surrogate marker for highly processed foods and such foods are unhealthy for other reasons besides sodium.  That being said, if the estimates on calorie and sodium intake were ignored, increasing F&V intake by 42g alone isn’t going to have that much impact on population health

*** The study also modelled the effect of taxes on SSBs, sugar, saturated fat and sodium.  The study estimated that all these taxes combined, plus the F&V subsidy, would reduce 470,000 disability adjusted life years (DALYs, or years with chronic disease) and would reduce health healthcare expenditure by $3.4 billion.  These figures seem impressive, but need to be put in context.  The study used a population of 22 million, so this works out to average reduction of 0.0214 DALYs per person (7.8 days) and average reduction in healthcare costs of $155.55 per person across their lifetime (or a few dollars per year, depending on how long you think the average person will live for (e.g. 40 years = $3.86 per person per year)).  This magnitude of response is consistent with another Australian study I discussed previously.  Modest taxes on unhealthy foods are somewhat useful at generating revenue for the diseases they increase the risk of, and will very marginally improve population health, but they won’t come close to solving the obesity/chronic disease epidemic