Trivial Disputes

December 17th, 2005 by Ben Goldacre in bad science, chocolate, mail, references, statistics | 26 Comments »

Ben Goldacre
Saturday December 17, 2005
The Guardian

Christmas is a time for pedantic family disputes, and newspaper stories about how alcohol and chocolate are good for you. Imagine, if you can, how Christmassy life would be if we could combine these two elements.

This week, the Daily Mail reported on a scientific study which reported an observation from survey data: that people who drink alcohol in moderation have lower levels of obesity than people who drink heavily. No surprise there. But the same study also found, more interestingly, that people who drink moderately have lower levels of obesity than teetotallers, people who don’t drink at all.

“Scientists said” they began, as is traditional: “drinking a few times a week can cut the risk of obesity by 27% compared to teetotallers… Researchers are unsure quite why small, regular amounts of alcohol help to keep drinkers weight down…” And so on.

Now look. I freely confess, I am possibly one of the pickiest people on the planet. But read those sentences again: this Daily Mail article is suggesting that there is a causal relationship at play here, between moderate drinking and lower rates of obesity.

And that’s not quite what the research showed. It looked at survey data, and found that people who said they were teetotallers had higher rates of obesity than people who said they were moderate drinkers. That’s what it found. It’s perfectly possible, and you might even say, likely, that there is a third, unrelated factor at play here, one that causes people both to be teetotal and have higher rates of obesity.

Let’s be honest here: teetotallers are abnormal. They’re different. They’re not like everyone else. I’m sure they have their reasons for not drinking. And in fact, those reasons might even be our missing link, the factor that causes you to be both a teetotaller and obese: they might be moral, or cultural, perhaps people from ethnic groups who drink less are more likely to be obese; perhaps people who deny themselves the indulgence of alcohol are more likely to indulge in chocolate and chips; perhaps pre-existing ill health will force you to give up alcohol, and make you more likely to be obese, and that’s what’s skewing the figures.

Perhaps these teetotallers are recovering alcoholics: because among the people I know, they’re the ones who are most likely to be absolute teetotallers, and they’re also more likely to be fat, from all those years of heavy alcohol abuse.

And you know what people are like: you can’t trust them as far as you can throw them, especially when you ask them about vice. Perhaps, in the survey, some of the people who said they were teetotallers were just lying about how much they drank, perhaps because they were, in fact, big drinkers, and they were ashamed of that, or in denial about it. Perhaps, even, and I don’t think this is too much of a long shot, people who lie to themselves, and others, are more likely to be obese.

That’s why we do scientific experiments, where we can control as many of the variables as possible. This study is a useful bit of preliminary survey data, but let’s say we care enough about the question that we want to discount these extra possible causal factors we’ve listed above: we’d have to take a group of people, and randomly assign them to a moderate-alcohol or no-alcohol group, and follow up what happened to their weight later. We might see a weight change over time – another hint at a causal relationship – and we’d know that ex-alcoholics, and big chocolate fans, were equally represented in each group.

But it’s also why designing scientific experiments can be difficult sometimes, or even impossible. I can’t take a group of people, with strong religious or cultural reasons for not drinking, and persuade them to be randomised into a group where they would have to drink alcohol.

Outside of the black and white, mechanistic, simple cause and effect universe of the press, finding an association between two things is not enough to say that there is a causal relationship between them. Black people might commit more crime, but it’s very unlikely that this is because the colour of somebody’s skin causes criminal behaviour: in just the same way that drinking moderately doesn’t necessarily make you trim.


++++++++++++++++++++++++++++++++++++++++++
If you like what I do, and you want me to do more, you can: buy my books Bad Science and Bad Pharma, give them to your friends, put them on your reading list, employ me to do a talk, or tweet this article to your friends. Thanks! ++++++++++++++++++++++++++++++++++++++++++

26 Responses



  1. CaptainSensible said,

    December 17, 2005 at 12:09 pm

    I’m not teetotal, but I rarely drink alcohol (last drink was a couple of months ago). I don’t dislike the stuff (often treat myself to some decent red wine at Xmas), but a somewhat solitary lifestyle and lack of surplus cash to spend on alcoholic treats means that I just don’t drink a lot. I used to be a moderate drinker, but I sort of lapsed into being virtually teetotal without realising it. I’m not religious.

    My height is 186 centimetres and my weight varies between 78-80 kilograms, which puts me near the middle of the normal range on the BMI chart (I could be fitter though). I have been slightly overweight in the past (85+ kilograms), but I managed to get back to normal weight via dietary changes (cutting out sugar in/on tea/coffee/weetabix etc.) and a bit more exercise.

    If I ever start drinking regularly again, I will probably gain weight if I don’t compensate for the extra calories by doing more exercise (unless alcohol affects your digestive system’s ability to handle food in the normal way or something).

  2. rich13 said,

    December 17, 2005 at 12:52 pm

    Top comment, CaptainSensible. Humanities, was it?

  3. CaptainSensible said,

    December 17, 2005 at 1:26 pm

    I do have a humanities degree (to my shame); are you implying that comment is out-of-order?

    My experience might be anecdotal etc., but I still find it hard to believe that increasing my alcohol intake to moderate levels (i.e. 4-5 units per week instead of 4-5 units every six months) would make me lose any more weight.

  4. BSM said,

    December 17, 2005 at 1:56 pm

    I think the Captain is making a valid observation. He is not drawing a global conclusion from his personal experience, but he is correctly bemused at the global conclusion being drawn by the Daily Merde.

    He is correct to wonder how, if there really was a causal connection between alcohol consumption and being less obese than a teetotaller, adding 4-5 units of alcohol might cause him to lose weight.

  5. rich13 said,

    December 17, 2005 at 1:57 pm

    1. Exactly. Nail on head.

    2. See above.

  6. katie said,

    December 17, 2005 at 2:37 pm

    I live in an area with a very high concentration of very religious people who are all teetotalers (it’s a small town). And the leading two health problems in the town are high blood pressure and Type II diabetes (everyone is very overweight).

    Now, I’m definitely not saying that I think drinking would make these people skinnier. But I think it’s the lifestyle that these lovely people lead–instead of having a couple of beer to relax, they down massive bowls of potato chips and eat all kinds of delicious fatty, high-calorie homebaking.

    I would like to see a study where they measured the caloric input (and/or maybe physical activity) of moderate drinkers versus teetotalers and then looked at overweightedness. But then again, would you really trust self-reported results about caloric intake?

  7. Coobeastie said,

    December 17, 2005 at 3:53 pm

    Very anecdotal, but everyone I know could give up chocolate or alcohol, but not both. As far as booze is concerned I can take it or leave it, but deny me my chocolate and I get grouchy.

  8. Tessa K said,

    December 17, 2005 at 8:29 pm

    If you switched from being teetotal to drinking alcohol for the purpose of the experiment, of course you would put on weight. More calories going in, the same number being burnt up as before. However, that would not be a valid experiment as the figures in the article are comparing current lifestyles. Similarly, a drinker who gave up would lose weight. (In both cases, all other things being equal, of course).

    The word ‘teetotaller’ in the article doesn’t reveal how many have given up and how many never drank, as Ben points out. It’s often said that when you give up one addiction, you’re likely to replace it with another. If the replacement is pies, you’ll pork up.

    For the record, I don’t drink and I am thin. Yes, I know, I am a freak of nature (the not drinking, not the thin part).

  9. RS said,

    December 19, 2005 at 8:37 am

    Well Type II DM and drinking alcohol aren’t the best combination, which is one potential skew.

    But don’t obese people underestimate what they eat, so it wouldn’t be a stretch of the imagination to think they underestimate what they drink too, particularly as alcohol is recognised as a calorie source.

    Can we make a silly argument about otaining a proportion of your calories via alcohol rather than fatty foods, and the impact on weight?

  10. Teek said,

    December 19, 2005 at 9:20 am

    “drinking a few times a week can cut the risk of obesity by 27% compared to teetotallers”

    this appears to arise from the “cum hoc ergo propter hoc” fallacy (no i’m not a flaky humanities grad, i just spend too much time on Wikipedia – en.wikipedia.org/wiki/Correlation_implies_causation_%28logical_fallacy%29)

    just because non-drinkers seem to be more obese than moderate drinkers, doesn’t mean drinking moderate amounts make you thinner. as Ben said, there could be a common reason or missing link that cause folks not to drink and to be fat at the same time.

    oh, and folks, re: the above “i’m think and don’t drink” type stories – nice testemonials, but let’s not forget – the plural of anecdote is not data…!” www.badscience.net/?a=xdforum&xdforum_action=viewthread&xf_id=1&xt_id=8

    p.s. when can we buy those T-shirts already…?!

  11. GWO said,

    December 19, 2005 at 10:00 am

    I’m more interested in the alternative causation.

    Isn’t it just as likely that being thin makes people psychologically more likely to drink alcohol

  12. RS said,

    December 19, 2005 at 10:47 am

    “Isn’t it just as likely that being thin makes people psychologically more likely to drink alcohol”

    Maybe the fat people are all hiding at home, friendless, while the beautiful young things are out socialising in the fashionable bars and clubs?

  13. Delster said,

    December 19, 2005 at 11:03 am

    I used to drink quite a bit. Out most nights and maybe 5-8 pints a night, every night!

    However i was also very fit. 5′ 10″ in height and about 170lb without a single scrap of spare flesh on me, the reason i was able to maintain that? i was in the military and doing a lot of hard physical work. I was on my feet for the whole day, I also swam 2.5km every afternoon after work.

    These days i am a more moderat drinker but am also quite a bit heavier, the reason? i took a desk job.

    I dare say that if you took a bunch of heavy drinkers and put them through basic training they would lose more weight than a control group of heavy drinkers who just cut out the booze.

    my point? i hear you cry…. well why do people keep looking for “causes” for obesity etc when it’s a simple matter of what goes in has to be used or removed or else it accumulates.

  14. Jeremy Miles said,

    December 19, 2005 at 2:17 pm

    Let’s not forget age as a possible third factor. In my experience, older people are more likely to be non-drinkers than younger people. As I get older, I find I drink considerably less (having young children makes hangovers really, really grim) and I also find I’m less sprightly than I once was (children don’t both fit on the back of my bike, so I can’t cycle as much).

    However, it’s not really for us to provide potential reasons. If they want to convince us of a causal argument, they need to present it.

  15. John Coffin said,

    December 19, 2005 at 7:33 pm

    Calorie intake in the form of alcohol doesn’t float in a separate univerise from other sources. Most drinking alcoholics (in my personal sample of several hundred acquaintances) eat poorly. Nothing quite dampens the appeal of a hearty breakfast like dry heaves and a splitting headache; the appeal of fresh fruit and robust vegetables is attenuated by a constantly sour and turbulent stomach.

    It is also worth noting that I’ve never heard of wet drunk who didn’t consider THEIR drinking a model of moderation. As the old joke goes: ‘How do you know when an alcoholic is lying?…..their lips move.”

    The impossibility of extracting valid data from surveys and limited clinical observation (‘gerbils and neutrons don’t lie, people do’) is nicely put in the following citation:

    Addiction
    Volume 91 Issue 6 Page 779 – June 1996
    doi:10.1046/j.1360-0443.1996.9167795.x
    How to have a high success rate in treatment: advice for evaluators of alcoholism programs
    William R. Miller, Martha Sanchez-Craig
    Two seasoned alcohol treatment researchers offer tongue-in-cheek advice to novice program evaluators faced with increasing pressure to show high success rates. Based on published examples, they advise: (1) choose only good prognosis cases to evaluate; (2) keep follow-up periods as short as possible; (3) avoid control and comparison groups; (4) choose measures carefully; (5) focus only on alcohol outcomes; (6) use liberal definitions of success; (7) rely solely upon self-report and (8) always declare victory regardless of findings.

  16. Teek said,

    December 20, 2005 at 8:37 am

    (1) choose only good prognosis cases to evaluate; (2) keep follow-up periods as short as possible; (3) avoid control and comparison groups; (4) choose measures carefully; (5) focus only on alcohol outcomes; (6) use liberal definitions of success; (7) rely solely upon self-report and (8) always declare victory regardless of findings.

    hey, maybe the Bristol homeopaths have read this advice, they seem to apply it word for word in their “research”!!!

  17. Tessa K said,

    December 20, 2005 at 11:40 pm

    Lies, damn lies … and statistics.

  18. Mathew said,

    December 22, 2005 at 2:29 pm

    Surely the conclusion reached by the study in the Daily Mail is just a variation of what is known as the ‘alcohol paradox’: that drinking alcohol in moderation does not lead to the expected weight gain (expected, that is, from the naïve viewpoint that more calories = more weight).

    Several studies have shown that moderate consumption of alcohol does not lead to weight gain and some have actually show that, in women, those who drink alcohol actually weigh less than those who don’t drink.

    For example, Mark F McCarty, writing in The American Journal of Clinical Nutrition, states that ‘large cross-sectional epidemiologic surveys show that women who are moderate drinkers tend to be much lighter than women who do not drink.’

    The British Health Survey, published in 1995, found that women who were moderate drinkers were about half as likely to be obese as non-drinkers.

    A study of 138,000 men and women by Harvard University, published in 1991, found that men who drank moderate amounts of wine or beer gained no more weight over the years than men who did not and that women had a body mass index about 15 per cent lower than non-drinkers.

    And finally, in one very recent study (published in The American Journal of Clinical Nutrition in 2003) by the Department of Primary Care and Population Science, Royal Free and University College Medical School, London, the authors concluded that ‘there is no evidence that light-to-moderate drinking is associated with weight gain’.

    Several short-term diet studies have echoed these findings: people who temporarily added alcohol to their diets usually lost weight.

    Those scientists who specialise in the field of alcohol metabolism have known about this relationship between alcohol and weight for over twenty years (and refer to it as ‘the alcohol paradox’, as they have been unable to explain why alcohol does not put on weight). For example, Eric Jéquier of the Institute of Physiology, University of Lausanne, writing in The American Journal of Clinical Nutrition, stated that ‘the relation between alcohol consumption and body weight remains an enigma for nutritionists’. And John Crouse, medical researcher at Wake Forest University in Winston-Salem, North Carolina, states: ‘It comes as a shock to see that people who drink more don’t weigh more’.

    Sorry about the somewhat ambiguous nature of the references – the above stuff comes from a note I wrote for a friend of mine months ago in which I excluded all unnecessary detail to avoid boring her and I have since lost the original list of references I had.

  19. Tessa K said,

    December 22, 2005 at 4:21 pm

    According to The Sceptic in today’s Guardian, there is a correlation between drinking and living longer. He says: ‘The only gloom is that the evidence is far less promising for women: it may be that the health benefits of alcohol only apply to men.’

    At last, the excuse men need for skiving off down the pub.

  20. Martin said,

    December 23, 2005 at 7:53 am

    Does this mean that there’s some truth in the old joke that giving up drinking doesn’t make you live longer – it just seems that way.

  21. peter augustus said,

    December 29, 2005 at 6:13 pm

    Surely the key factor is in the word moderate. The fact that someone is capable of moderating their alcohol intake makes it likely that they can also moderate their food intake. Heavy eaters get fat, heavy drinkers get fat and people that both eat and drink more calories than they can work off get fat. I know, I’ve tried it all and shown that even running 50 miles a week won’t make you thin if you eat and drink to excess.

  22. Hatter said,

    January 4, 2006 at 10:46 am

    Expecting weight gain from moderate alcohol consumption assumes all other things remain equal. Consider that when socialising you could be eating snacks, smoking a cigarette or sipping a drink. Any one of those can give you something to do with your hands and provide an oral fix. Chances are that snacking would give the highest caloric intake, smoking the lowest.

    What would be absolutely hilarious is if they did this same survey amongst users of other recreational drugs. The headline could be ‘Study finds LSD consumption reduces the risk of obesity by 80%’

    John Coffin there is good reason why the * Anonymous groups are violently opposed to anyone assessing the efficacy of their method. It wouldn’t do for the general public to become aware of their monumental failure rate.

  23. Anne the mad scientist said,

    January 10, 2006 at 9:28 am

    There is no doubt in my mind why this survey revealed that women who drink stay thin. It was conducted on scientists. I have to work four times as hard as my male colleges to receive a quarter of the recognition. Little wonder that I weigh less than eight stone and drink like a fish…

  24. Julie Stahlhut said,

    January 24, 2006 at 8:04 pm

    Different individuals also have different tastes in food and drink, as well as different reasons for either drinking or avoiding alcohol. I know quite a few people who have no moral or health objections to alcohol but can’t stand the taste, so one thing I wondered is if some people who dislike the taste of alcohol might have a higher-than-average attraction to sugar, fats, or other high-calorie foods. It might also be interesting to break down the “drinking” group by which alcoholic beverages they tend to choose — in other words, does it matter if you prefer to drink dry white wine rather than liqueur-and-cream dessert drinks?

    Not that it seems to matter to me; I drink in moderation and like just about every kind of food you can imagine, from tofu and bean sprouts to fettucine Alfredo. Then again, I don’t eat a lot of desserts. Maybe I can tell myself that if I didn’t drink wine and the occasional cocktail, I’d be 50 pounds overweight instead of only 15. :-)

  25. wayscj said,

    November 21, 2009 at 7:32 am

    ed hardy ed hardy
    ed hardy clothing ed hardy clothing
    ed hardy shop ed hardy shop
    christian audigier christian audigier
    ed hardy cheap ed hardy cheap
    ed hardy outlet ed hardy outlet
    ed hardy sale ed hardy sale
    ed hardy store ed hardy store
    ed hardy mens ed hardy mens
    ed hardy womens ed hardy womens
    ed hardy kids ed hardy kids ed hardy kids

  26. iphone revolution said,

    December 30, 2009 at 8:31 am

    iphone wireless

    iphone

    Apple iphone