Monday, February 6, 2012

Sweet tooth or sweet talk? The truth about the truth about sugar

You have to suspect any story with a title that begins "The truth about....", and the story in last week's Nature is no exception: "Public Health: The Toxic Truth about Sugar". Just the latest in a number of stories indicting sugar as the cause of all that ails us (almost literally), the piece's own summary is this:

  • Sugar consumption is linked to a rise in non-communicable disease
  • Sugar's effects on the body can be similar to those of alcohol
  • Regulation could include tax, limiting sales during school hours and placing age limits on purchase
The link is largely based on epidemiological evidence -- everywhere that people have switched to a "western diet", meaning lots of processed foods, including sugar, obesity has skyrocketed, as have hypertension, type II diabetes, cardiovascular disease, and cancers.  Therefore, sugar is the cause.

The evidence is clear:  even at Starbuck's it's hard to find some actual coffee (not to mention the impossibility of a 'small' coffee) in amongst the choco-banana-raspberry flattes.  And then there are the Scots' deep-fried Mars bars.  Who could doubt that our commercially whetted sweet tooth is the tooth that bites with poisoned fangs?  But is the evidence actually so clear?

The conclusion looks suspiciously in need of the "correlation is not causation" reminder.  Indeed, a reader who posted a comment on the paper in Nature, Geoff Russell, pointed this out with cogent examples. We'll reproduce his entire comment here, as it makes the point well.
Geoff Russell said:
Australia provides a natural test of the sugar-is-the-evil-bullet theory. We don't produce much corn here, so continue to use cane sugar for most of our sweetening. In the 1960s we didn't have an obesity epidemic. How much sugar did we consume? According to the FAO, 52.3 kg per person per year in 1965 (of 55kg total sweeteners). 
What about now, in the midst of our own obesity and type 2 diabetes epidemics?
We are down to 39.6 kg of cane sugar per person per year, with an additional 8kg of non-sugar sweeteners. Overall there has been a modest decline in all sugars despite a rise in obesity and diabetes. How has our food supply has changed over the past 4 decades? We have more Calories. If may be tempting to attribute the US obesity crisis to sugars, but obesity increases elsewhere demonstrate that more Calories and less exercise are a sufficient explanation.
Similarly, compare Cuba and Italy. Cuba consumes 500 kCal per day of sugar and Italy just 300 kCal, Italy has an obesity/type 2 diabetes problem while Cuba's rates are very low. Historically, Cuba has eaten even more sugar than she does now ... without the evil consequences that this article portends.
There is no denying, of course, that obesity and what are usually thought of as its sequelae are public health problems in much of the world.  Whether or not said sequelae are indeed sequelae of obesity, or whether obesity and the rest are, individually, consequences of fat consumption, or sugar consumption, or processed food in general, or simply of excess calories relative to energy usage (too much munching in front of the telly) has still not been determined, though many have their favorite candidates.  Cholesterol, saturated fat, red meat, the non-Mediterranean diet, not enough exercise, and others.  

There's a fundamental problem with this simple sugar analysis when it doesn't reliably predict on either the individual or population basis (as Geoff Russell's comment points out).  Sugar, and/or what it's usually allied with, may well have detrimental effects on health, but clearly it's not as simple as is being said.  There's a well-known issue in epidemiology called the ecological fallacy, whereby we paint individuals with a brush dipped in a population-based paint.  That is, when we attribute generalizations about a group to causation at the individual level -- stereotyping is an example, but so is the (erroneous) assumption that because, say, risk of heart disease is higher among people who smoke, everyone who smokes will have a heart attack. 

It simply can't be possible that sugar is the single or even primary cause of the obesity etc. epidemic.  Too many healthy people consume a lot of sugar for this to be true.  And surely too many unhealthy people don't.  Trying to attribute this vast epidemic to a single dietary substance is denying the complexity of these diseases and of causation.  Unfortunately (in our view), people are beginning to feel so fiercely about sugar as the root of all evil that they are (in our view) no longer able to assess the science.  Instead, it's become such a strongly held belief that it's in danger of becoming a dogma that no longer needs to be tested, or the supporting studies examined with a skeptical eye.  That's never a good thing.

This is relevant for an MT post not because we don't want it to be true because we both OD on sugar all the time (we don't), or that we only eat celery and carrots (we don't), but because it relates to the general problem of inferring causation, especially when we can't do definitive experiments.  That is the common situation in human and evolutionary genetics.   So, it's worth sitting contemplatively over a cup (small) of coffee discussing whether there are better ways to know about weak, gradual,  or complex causation.

One lump, or  two?

23 comments:

  1. Bravo! The anti-sugar thing, along with the anti-(anything else that's yummy) thing comes, I am convinced, from the same well that produces other sorts of "western" self-loathing. I blame Lennon and McCartney's trip to India in the 1960s!

    The answer must be overconsumption. Sugar is a calorie-rich food, as is beef, as is bacon and beer and all the other supposedly naughties. As my fat cat who died a premature death can attest, we just aren't designed to cope with times being too good.

    ReplyDelete
    Replies
    1. You may be surprised to find that sugar has less calories than beef or beer or bacon. Fat has more than twice the calories as carbohydrates and alcohol nearly twice as much. Just saying.

      Delete
    2. Yep, it's all the stuff that sugar is added to that adds the calories.

      Delete
  2. I think it's a puritan thing. Did did you ever see a study about toxic broccoli? We're not supposed to enjoy ourselves!

    ReplyDelete
    Replies
    1. In my lectures on these epidemiological issues over the years, I refer to it as the "science by the seven deadly sins".

      Delete
  3. I agree with Anne about our culture. So indulging on food is bad, on porn is bad, on prostitution is bad, on drugs that make you feel good is bad, but Brussels sprouts and cauliflower are not bad-- because they taste bad.

    It's a complex of lifestyle components that vary among people, and are responsible. This is obvious because the disease prevalances have clearly changed but our genomes haven't.

    On the other hand, we're talking mainly about events in later life, which at least we get to enjoy, even if we've been too hefty for most of our lives. Indeed being stocky was for most people a sign of good health and usefulness as a spouse. And in the past there wasn't much opportunity just to veg out, since people had to do real work in those 'halcyon' days. And, all winter, had to eat awful stuff they stored in their root cellar.

    ReplyDelete
  4. A moral imperative for regulating sugar consumption. Amazing. That used to be what parents were for.

    ReplyDelete
  5. You do get a sense of religious zeal from these guys. I'd call it the missionary position, but that would suggest enjoyment, and there's none of that here!

    ReplyDelete
  6. And our Internet time needs gov't regulation too: http://news.discovery.com/tech/social-media-addictive-120204.html#mkcpgn=fbsci1

    ReplyDelete
  7. Since parents are apparently Twinkie-ing out with their kids, while vegging with them in front of the telly, the Mom effect seems to be nil.

    ReplyDelete
  8. Funny about the social media addiction thing -- wanting to check email is a craving, and _therefore_ should be resisted. I think we're back to Puritanism here.

    ReplyDelete
    Replies
    1. Exactly. It is just assumed that there is something *wrong* with wanting to do things, especially when one wants to do them to the exclusion of doing other things. Those other things, of course, are almost always behaviours some religious group thinks should be done, unless, of course, it is some religious fanaticism to begin with.

      It also is related to the "war on drugs". The original reasons (to the extent that any were discussed at all, many of which were just racist) for the laws against the consumption and distribution of what are now "controlled substances" was that the use of them violated Protestant Puritanism: feeling good.

      Now, it is almost always cashed out as some form of harm reduction---as remarkable an untruth as any born of religious moralism.

      Delete
    2. And some religions, like in some ways Calvinism, revel in luxuriating pleasure because it shows you've been saved. We just heard an installment of BBC Radio4's In Our Time, on the Kama Sutra, which made some points about the appropriate role that pleasure has in human affairs.

      But the bite of Puritanism in this country has been deep, and we're not alone about that.

      So, let's start doing some studies on the harmfulness of cauliflower!

      Delete
  9. It's more than a tad strange to see TV commercials telling children to play --I just saw one during the Superbowl last night.

    ReplyDelete
  10. I think people who separate their senses of self from their bodies are weirded out when cravings and desires chip away at "self" control.

    ReplyDelete
  11. Maybe we need a branch of epidemiology by and for hedonists. It would turn the field completely on its head. How dangerous are cauliflower and leafy greens? Or exercise? (My sister, an OT, warns against exercise all the time, so there's a start!) Self control? Why?

    One of the sugar-is-poison people spoke at Penn State a while ago. He's young now but he said he has this vision of dying at age 75 at his grandchild's birthday party after eating ice cream and cake. That's almost as sad as having to make commercials to tell children to play. And a lot of years without sugar.

    ReplyDelete
  12. On the ecological fallacy, you write:
    ``There's a well-known issue in epidemiology called the ecological fallacy, whereby we paint individuals with a brush dipped in a population-based paint. That is, when we attribute generalizations about a group to causation at the individual level -- stereotyping is an example, but so is the (erroneous) assumption that because, say, risk of heart disease is higher among people who smoke, everyone who smokes will have a heart attack.''

    That description captures a nuance of the ecological fallacy, but really misses how devastating the fallacy is. The point is this: it is possible to have a strong positive (say) correlation between (as in the famous graph from a Canadian professor), say, fat consumption and breast cancer rates when measured by countries, but to have no or the reverse correlation at the level of individuals within countries. Of course, the individualistic fallacy exists as well (generalizing from within-category to between category), and, or course, it all culminates in Simpson's paradox (an actual reversal over levels). The problem is that there are three correlations: average within-group, between-goup, and overall, ignoring group. If you only know one of them, you are extremely-limited on what you can say about the other two (see Cornfield in his response to Fisher and smoking and lung cancer on how knowing even one can put limits on just how reversed one of the others could be). Simpson's paradox---a topic for another time---really does severely limit causal conclusions from almost all epidemiological reasoning, especially as the confounds are rarely known. It is why, in his book _Causality_, Judea Pearl devotes so much of his effort on it.

    ReplyDelete
  13. Thanks, as usual,for this thoughtful comment. It is related to confounding, in which causation is inferred falsely from correlation because the correlate was itself correlated with some other risk factors (e.g., it was quipped 30 years ago that breast cancer was associated with owning telephones or indoor plumbing).

    There's no easy answer in observational research, as in much of evolutionary (esp. selection) scenarios as well as in genetic epidemiology, because causation is itself difficult to show definitively, and because variation is so inherently a part of life.

    George Davey Smith has suggested a method called 'Mendelian randomisation' (spelled that way because he's a Brit!), that uses genetically demonstrated causal effects to confirm suspected environmental causes,and hence to reduce problems of confounding.

    Generally, however, it's a real problem as you clearly pointed out!

    ReplyDelete
  14. Thanks, John. The many issues that make observational studies so difficult to interpret (to be charitable) are why it's primarily when effects are large that they are identifiable -- when they are harder to miss than to find! Mendelian randomisation is one trick, but unfortunately it's only applicable in rather unusual instances.

    ReplyDelete
  15. Here is the paper on Cornfield's challenge to Fisher:
    http://dl.dropbox.com/u/2793590/99asa.pdf

    ReplyDelete
  16. You might also find Judea Pearl's take on Simpson's Paradox enlightening:
    http://ftp.cs.ucla.edu/pub/stat_ser/R264.pdf

    ReplyDelete