Friday, March 7, 2014

The WHO and recommendations on sugar consumption: an ill-posed problem

The search for magic health bullets, or the Single Evil goes on unabated, despite a steady record of essential failure.  Is that fair?  Well, we've got decades of very extensive, expensive, and expansively technical studies of some questions of major public health relevance -- and little to show for it.

As a new example, after a two-year effort to connect the dots between between sugar and disease, the World Health Organization, whose large professional research staff should know how to do that,  believes it has done so, and now recommends we all reduce our sugar intake to 10% of our diet, or better yet, 5%.  That's 12 or fewer tablespoons of sugar a day.  
There is increasing concern that consumption of free sugars, particularly in the form of sugar-sweetened beverages, may result in both reduced intake of foods containing more nutritionally adequate calories and an increase in total caloric intake, leading to an unhealthy diet, weight gain and increased risk of noncommunicable diseases (NCDs).
Also of great concern is the role free sugars play in the development of dental diseases, particularly dental caries. Dental diseases are the most prevalent NCDs globally and though great improvements in prevention and treatment have occurred in the last decades, dental diseases continue to cause pain, anxiety, functional limitation and social handicap through tooth loss, for large numbers of people worldwide.
So, sugar is bad for our teeth and bad for our health.

But how good is the evidence?

To answer that question, the WHO commissioned a meta-analysis, that is, a single combined analysis of all pertinent previous cohort studies of the effects of sugar on obesity, which was published last month in the British Medical Journal by Morenga et al.  The authors note that they included studies that controlled for lifestyle factors and medical interventions.  They looked at two groups of studies; cohort studies and trials.  Of the trials, they chose two groups.
One group included studies in which participants in the intervention arm were advised to decrease or increase sugars, or foods and drinks containing sugars. Although such advice was generally accompanied by the recommendation to increase or decrease other forms of carbohydrate, there was no strict attempt at weight control. These trials are referred to as ad libitum studies. The other group of trials attempted to achieve isoenergetic replacement of sugars with other forms of carbohydrate. Interventions designed to achieve weight loss were excluded because the ultimate aim of the review was to facilitate the development of population based recommendations rather than nutritional recommendations for the management of obesity.
That is, people in these studies were either asked to add or subtract sugary foods and drinks from their diet, but change nothing else, or they were asked to subtract sugar and substitute it with a different carbohydrate.  Presumably the latter was to control for the effect of simply adding or subtracting calories of any sort, though this isn't clear in the paper.

To identify studies that might clarify the issue, a literature search identified 7895 potentially relevant studies, but of these, 19 met their criteria for ad libitum studies, and 11 others met their criteria for isoenergetic replacement studies.  And they also identified 9445 potentially relevant cohort studies, and of these, 38 were deemed to be appropriate for inclusion.

From these 30 trials and 38 prospective cohort studies, what do they conclude?  Well, of the 30 trials, five studies measured the effect of reducing dietary sugars.  To quote the authors of the WHO commissioned BMJ paper, "Reduction in dietary sugars intake was associated with significantly reduced weight (-0.80 kg (95% confidence interval -1.21 to -0.39); P lt 0.001) at the end of the intervention period by comparison with no reduction or an increase in sugars intake."

So, five studies report statistically 'significant' weight loss with reduction in sugar intake.  But what are we talking about here?  Well, an average decrease of less than 2 pounds, or at most 2 1/2 pounds, with variation around that.  This may be 'statistically significant', but all I can say is that if I lost 2 pounds I wouldn't think it important enough to tell my mother about, never mind publish it.  And, only one of the studies, participants were asked to substitute low sugar foods for the high sugar foods they were eliminating.  So, one study tried to test the effect of eliminating sugar, but not changing caloric intake, but the others did not.

And what about the study duration differences?  Comparing 10 weeks to 8 months, the durations of some of the studies, is an apples to oranges kind of comparison -- what about the effects of differential seasonal differences in diet, exercise, their correlations, and so forth?  And, three of the five studies reported results only for those who completed the study, which is a possible source of bias.  Who was determined enough to finish, and who not, and why?  Indeed, the authors write that excluding two of these studies weakened the results somewhat, meaning that there was indeed a difference between completers and non-completers, at least in those two studies.

The authors identified 12 studies that asked participants to substitute sugars with other macronutrients.
Interventions ranged from two weeks to six months, and sugars were in the form of either sucrose or fructose used to sweeten foods or liquids. We saw no evidence of difference in weight change as a result of differences in sugars intakes when energy intakes were equivalent (0.04 kg (95% confidence interval −0.04 to 0.13)).
While Morenga et al. emphasize the results that showed weight gain to be associated with sugar intake,  a number of these studies reported no change in weight, reporting "Sucrose intake not significantly associated with weight gain at follow-up", or "No relation was seen between sweet foods consumption and WC [weight change]".  See Table 5 in the paper.  

Data from studies on children were 'equivocal', and the authors discuss possible reasons for this, including poor compliance with dietary advice.

So, after all of this effort, on a subject that has been widely trumpeted as well-known, is sugar bad for us?  Maybe, but we don't know it from this study.  All the studies included in the meta-analysis are based on methods of assessing food intake that are themselves questionable.  Dietary recall, the method used in the trials, and food frequency questionnaires used in the cohort studies, or indeed just about any other method of dietary assessment, are notoriously inaccurate measures of what we eat.  And, the amount of weight gain reported to be due to sugar consumption is perhaps statistically significant, but with respect to the dangers associated with obesity, it's trivial.

In addition, do we actually know whether study participants cut down on total caloric intake when they reduced sugar intake?  If they didn't substitute sugar with something else, it's impossible to know whether any subsequent weight loss is due to the reduced sugar in their diet, or simply fewer calories.  The fact that the subjects in the isoenergetic studies lost no weight when they replaced sugar with other carbohydrates would suggest that this is the case.

Morenga et al. say that possible confounders -- lifestyle or dietary factors -- that could influence an association between sugar consumption and weight gain, were controlled for in the studies they include in their analysis.  Epidemiological studies can never anticipate or control for all factors that might influence a causal variable, because causal pathways are complex and all components of those pathways can not be known completely. The problem of confounding, that is, causation by an unmeasured factor whose exposure is correlated with a measured one (like sugar intake?) is notoriously tough to identify.

All of this is predicated on the assumption that weight gain is unhealthy.  Or leads to ill health.  The WHO paper didn't address the strength of the evidence for this. Nor did they show any evidence to suggest that sugar contributes to ill health, or even weight gain, any more than any other food.  When speculating on how sugar could be associated with weight gain, the authors write, "The most obvious mechanism by which increasing sugars might promote weight gain is by increasing energy consumption to an extent that exceeds energy output and distorts energy balance."  That is, people who gain weight consume more of anything than they burn, as the isoenergetic studies suggest.  By this logic, eliminating milk or eggs or meat and not replacing them would cause weight loss, too.

Morenga et al. do write that mechanisms by which sugar itself might be a cause of weight gain have been suggested, but that is beyond the scope of their paper.

Tooth decay
Sugar consumption is much less unambiguously associated with tooth decay.  Or at least I thought so, and it seemed something 'everybody knows'.....until I read the review commissioned by the WHO, a paper in the Journal of Dental Research, the foundation for the WHO recommendation that sugar consumption should be less than 5% of our diet.
From 5,990 papers identified, 55 studies were eligible – 3 intervention, 8 cohort, 20 population, and 24 cross-sectional. Data variability limited meta-analysis. Of the studies, 42 out of 50 of those in children and 5 out of 5 in adults reported at least one positive association between sugars and caries. There is evidence of moderate quality showing that caries is lower when free-sugars intake is < 10% E. With the < 5% E cut-off, a significant relationship was observed, but the evidence was judged to be of very low quality. The findings are relevant to minimizing caries risk throughout the life course.
The evidence for an association is judged to be 'moderate' or 'of very low quality.'  This largely reflects the fact that most studies didn't ask the question the WHO was interested in -- is there a threshold amount of sugar that is highly associated with cavities?  Still, the studies do show an association, even if confounders like socioeconomic status might be problems.

Based on these studies, the WHO now recommends that we cut sugar down to 10 or even better, 5% of our diet.  That's 6 - 12 tablespoons of sugar a day.  According to the Guardian, average intake among adults in the UK is 11.6%.  Surely reducing it to 10% isn't going to eliminate diabetes or heart disease. If we could even measure our intake accurately enough, and had the patience to tinker with our every mouthful.

Further, the WHO has asked a question that can't be answered:  Is there a threshold over which sugar consumption is unhealthy?  We wrote not long ago about the well-posed problem problem and this is a good example.  There isn't a single answer to the question.  Indeed, it's not even possible to understand, never mind answer. Why should there be a threshold?  What kind of sugar consumption?  Does the vehicle for sugar consumption (cake vs soda, e.g.) make a difference? What does unhealthy mean anyway?  The question is really not a question at all from a scientific point of view.

Is this the best we can do?  Is it more than today's example of hand-waving dressed in the language of science?  After all, we've been studying these subjects for decades.  And if it is the best we can do, why bother and why spend the resources when there are more soluble problems that deserve attention? 


Holly Dunsworth said...

This attack on sugar reminds me of the other kind of attacking that our world leaders like to do with little evidence.

Kirk Maxey said...

There are risks to this kind of nonsense. I watched as a well educated, extremely health conscious mother transitioned her 9 month-old son from a diet of 100% breast milk (which has a lot of sugar) to an essentially sugar-free diet rich in fiber, vegetables and little meat. He dropped from the 60th percentile in weight to the 4th, and began to accrue language and social developmental delays. He was rescued by a scandalous intervention - ad libitum access 2 days each week to unlimited popsicles, ice cream and fruit juices. He's now back at the 55th percentile, and his delays have reversed.

Ken Weiss said...

We are 'fed' similar kinds of stories by the media and (sadly) the journals every day. The flurry of definitively stated but easy to critique and conflicting claims means we have no scientific way to determine what to do. But we know that moderation works, at least as well as anything else in most of these situations.

JayMan said...

Wow, have you been reading my stuff?

(My posts on Health and Medicine)

This was brilliantly written! Please don't take this the wrong way, but I wish it didn't come from you. This is because the people in my circles will be apt to not take this seriously. But I think you're quite correct. The evidence that sugar (or for that matter, most dietary components) is harmful or leads to weight gain is lacking. Indeed, I made much the same point, even down to questioning tooth decay. I think in the zeitgeist to blame health outcomes on lifestyle, people have sacrificed sound scientific practice. Indeed, whole disciplines in this space are effectively compromised, because slipshod research methods are held up to be good practice, which is sad.

Thanks a lot for this post.

Geoff Dougherty said...

One thing to consider is that, while a two-pound weight loss may not seem like much on an individual level, small changes multiplied across a large number of people actually yield big gains in population health, especially when the health issue is as prevalent as obesity.

Another: While it would be great if we could design experiments that looked at replacing sugar with some other food, and then measure weight loss, there is some evidence that the relationship between sugar and weight is complicated by the fact that eating sugar encourages one to eat more of everything else, too. Isoenergetic studies don't really capture this dynamic, which means that they may underestimate the effect of sugar.

Ken Weiss said...

These are good points, but the 2-pound issue depends on the variance not just the mean. It is possible that it's just a typical common response level (very small). It is also possible that some people are way more sensitive and raise the average; if this were the case, then they are the ones who should be identified for health-care reasons.

The correlation of causes in your last sentence is a huge problem for any sort of public health research; it would mean that it is not, in fact, the sugar but something one might call the 'eating habit'. That to me seems far more likely than that sugar, per se, is the villain in this particular story.

Anne Buchanan said...

Thanks for these comments. Perhaps a more relevant point about the 2 lb weight loss is that the studies were only 10 weeks to 8 months in duration, so it's possible that given more time, people who reduce sugar consumption would lose more weight. But the problem is that we don't know that from the studies that the WHO based their recommendations on. And, we certainly don't know that people can maintain low sugar diets or weight loss long term.

And then there's the confounded difference between statistically significant and biologically significant findings, as well.

I am not arguing that sugar is harmless. I don't think we know that one way or the other; I certainly don't. I am arguing that the studies that the WHO based its recommendation on are, to my mind, not sufficient evidence.

Jim Wood said...

Perhaps, if the people in your circles won't take this argument seriously because it comes from Anne Buchanan, they need to open their minds a bit bit?

Geoff Dougherty said...

The WHO's actions would look more sensible if they did a better job of communicating the context in which they're developed. The WHO isn't making a clinical recommendation with the hope that the most obese individuals will eat less sugar. It's making a policy recommendation that member nations can use to develop population-level health interventions that will shift the entire distribution of BMI to the left, regardless of its variance. Ideally those interventions (like, say, a tax on high-fructose corn syrup) would work regardless of whether individuals can stick with a low-sugar diet for several months.

There's a classic article on the difference between clinical and population health interventions here:

Right now, many developing nations are undergoing an epidemiologic shift. In the past, their mortality was determined largely by infectious disease, and in the future it will be determined mostly by chronic disease. Without intervention, they will undoubtably develop the same kinds of norms and policies that led obesity, diabetes and cardiovascular disease to be huge problems in the U.S.

So the impetus to act, even in the absence of perfect science, is strong. Although honestly I think the science on sugar is on par with the science on many other potential determinants of health. It would be great if it were better. Some authors should have used intention-to-treat study designs and didn't. The effect sizes are on the small side, and not all of the studies point in the same direction.

But taken together, they do seem to suggest that sugar makes you fat. And although observational studies, especially poorly designed ones, may be confounded, they have a pretty good track record of getting the direction of effect right, even if they overstate the effect size.

Anne Buchanan said...

You bring up a very important point about the WHO's goal being to shift population level obesity and non-communicable disease rates. And, by the way, they are still taking comments on their recommendation, as far as I know, so it is not yet writ in stone.

But as I read it, they aren't recommending population level action -- they are recommending that individuals reduce their sugar consumption. Our sugar consumption. Based on studies that are less than convincing. Indicting a single nutrient is always suspect, to my mind, because of confounding and because of heterogeneity.

You are doing a much better job making their case than they have done! It is good that NCDs are getting the attention of the WHO. But we don't know what makes us fat. Today it's antibiotics! I think if sugar was the major contributor, epidemiological methods would be able to determine it because they are best at finding risk factors with large effects.

The other side of Rose's point is the ecological fallacy, of course. Pity epidemiology.

Geoff Dougherty said...

I think we can blame the media for a lot of the confusion.

Here's part of the text from the WHO webpages: "When finalized, the recommendations in this guideline can be used by programme managers and policy planners to assess current intake of free sugars relative to a benchmark and develop measures to decrease intake of free sugars, where necessary, through public health interventions."

And here's the lead on the CNN story: "The World Health Organization wants you to stop eating so much sugar. Seriously."

Anne Buchanan said...

True, they haven't specified the public health interventions they have in mind -- recommending a tax on high-fructose corn syrup, as you suggest, or the removal of soda from schools or the like -- and you're right that that is different from recommending that individuals cut down their sugar intake.

But both types of recommendations would be based on the same evidence, and the same epidemiological methods that are really good at identifying strong point-cause risk factors, and not so good when the effect is weak, long-term, confounded, contextually dependent, and so on.