Olive oil and nuts beat a low-fat diet that’s not low-fat

Great-looking results from a big randomized diet study reported today in the New York Times:

About 30 percent of heart attacks, strokes and deaths from heart disease can be prevented in people at high risk if they switch to a Mediterranean diet rich in olive oil, nuts, beans, fish, fruits and vegetables, and even drink wine with meals, a large and rigorous new study has found….

Scientists randomly assigned 7,447 people in Spain who were overweight, were smokers, or had diabetes or other risk factors for heart disease to follow the Mediterranean diet or a low-fat one.

Low-fat diets have not been shown in any rigorous way to be helpful, and they are also very hard for patients to maintain — a reality borne out in the new study, said Dr. Steven E. Nissen, chairman of the department of cardiovascular medicine at the Cleveland Clinic Foundation.

Now, I am not a low-fat dude.  Long ago I dated somebody who was into Dean Ornish and every time she “sauteed” onions in water a little piece of me died.  I pour a lot of olive oil on things, because I like it (especially Frantoia, which the guys at the Italian grocery in the Trenton Farmer’s Market turned me on to when I lived in Princeton) and because mainstream nutritional wisdom has been promoting monounsaturated fats for a long time now.  But I do think low-fat gets kind of a bad rap from the NYT piece.  Even more so in some of the other coverage, like the LA Times, which headlines their story “Mediterranean diet, with olive oil and nuts, beats low-fat diet.”  The Times, at least, points out far down in the piece that the “low-fat” group, while counseled to reduce fat, didn’t actually do so.  To get numbers, you have to go to the supplemental material of the original paper.  There, you find that the Mediterranean eaters were getting 41% of their calories from fat, while the “low-fat” arm got 37%.  A low-fat diet is 22%.  Random googling suggests that most vegans are getting 20%-30% of their calories from fat.

In other words, the study doesn’t really show that the Mediterranean diet is better for you than eating low-fat; it shows that hardly anybody is capable of eating low-fat, which is a different thing entirely.

Tagged , ,

11 thoughts on “Olive oil and nuts beat a low-fat diet that’s not low-fat

  1. First of all, I like Frantoia too!

    Second of all, I think this article does something right that you normally rarely see. Instead of comparing the people who actually maintain their low-fat diets against the people who actually maintain their Mediterranean diets, they compare what happens when people are told to do different things. This is a golden rule in modeling (which again is rarely understood), namely you should only test methodologies that you have true power over.

    Here, the question is, what should doctors tell people who are at risk of various diseases? And the possible answers are: go on a low-fat diet or “go on a Mediterranean diet”. The results of _telling people to do one of those things_ is what we need to compare, not the result of _people actually doing that_.

    The reason? People don’t actually do what you say. If we can find something that people will actually do, then we have a miracle on our hands. I discussed this problem here: http://mathbabe.org/2013/02/13/the-smell-test-for-big-data/.

  2. JSE says:

    I hope it’s clear that I have no problem with the NEJM paper, which I think looks great! I’m just quibbling with the way it got framed in the newspapers. I would say the issue of “intent to treat” is not at all rarely understood, at least in public health, where ease of compliance is understood to be an important part of any intervention. (e.g. people who study contraceptive effectiveness pay close attention to this stuff.)

    I think we need the answers to both questions you mention: what are the results of doing X, and what are the results of telling people to do X? And that popular accounts should be crystal clear about which is being studied in any given case. If you’re a doctor thinking about what advice to give your patients, you need to know what the result of giving the advice is — i.e. what’s under YOUR control. On the other hand, if you’re a person at high cardiovascular risk, like the subjects in this study, you have a lot more control over what you eat and knowledge about your likelihood of compliance than your doctor does. There really are low-fat vegans out there! This study doesn’t tell those people not to be low-fat vegans, it tells them that they are unlikely to succeed in convincing many of their friends to be low-fat vegans.

    And if there were an X such that doing X was very effective, but telling people to do X was not, that would be worth knowing, because it would make “how can we effectively improve compliance with X by e.g. developing versions of X that are easier to do” an important project; while if telling people to do X was ineffective BECAUSE doing X was ineffective, the compliance project would be a waste of time.

  3. I’m glad we’re sorting out the issues, but I’m going to push back that the question of “what are the results of doing X?” is very important.

    The truth is, most people don’t get the subtlety of how darned hard it is to change one’s behavior, or else they don’t think it applies to them, so they end up caring way too much about the results of a diet assuming one sticks to it. In other words, they exclude the probability of they themselves sticking to it from their expectation function. The result is bad results, depression, and shame. And _lots_ of money for the corrupt diet industry.

    So we need to do one of two things. Either de-emphasize the result “if you stick to it” (my opinion) or go for another campaign where people are told that they should try lots of different things before giving up. One reason the second idea is good is that it’s likely (or at least possibly) true, but it’s bad because it makes people even more vulnerable to ridiculous and expensive fad diets (followed by depression). And that’s the last thing we need.

    As you said yourself, “they are unlikely to succeed in convincing many of their friends to be low-fat vegans.” If we can’t convince our own friends of something, chances are our doctors can’t either. So why go around talking about how being a low-fat vegan is a great plan? It’s misleading and ultimately defeating.

  4. Related question: can we use some of these apps that are purported to help you keep track of your diet and exercise to force accurate self-reporting? So we could keep track of people who say they’re going on a low-fat diet and try to follow up a year later and see how that went? Trouble is, there’s still such a survivorship bias for the follow-up. People have been trained to feel personally at fault in these situations.

    But if we could do this, perhaps by inappropriately stalking them, then when they start out a “low-fat diet” we could immediately tell them something like, “great! you have a 0.2% chance of sticking to that, but if you do you’ll lose 65 pounds.”

    And, even if we do _that_, I’m pretty sure people will still consider their odds of success good, because people always think they’re in the top 0.2%. That bias must have a name. I’ll call it the “New Year’s Resolution Bias.”

  5. JSE says:

    Cathy, I think it’s worth knowing what plans are effective, even when those plans are hard to comply with, because when you know this, you know when it’s worth taking steps to aid or even in some contexts enforce compliance. Maybe I’m influenced here by the fact that I’ve thought a lot about tobacco. Quitting smoking is really, really hard, and it’s only a minority who can pull it off by sheer willpower. If you do a trial where you advise a random sample of smokers that they really ought to quit smoking, you’d conclude “telling people to quit smoking is not an effective health intervention,” because almost all of those smokers would keep on puffing.

    On the other hand, the knowledge that we have about tobacco and adverse health outcomes is sufficiently strong that people decided not to settle for “it’s unrealistic to expect more than a handful of people to quit smoking.” Instead, they thought hard about ways to make it easier to quit (or, depending on how you put it, ways to make it harder to smoke.) So: development of Chantix and nicotine patches, smoking bans in restaurants, planes, and offices, increased cigarette taxes, etc. And fewer people do smoke now — a lot fewer!

    I don’t think everybody needs to become vegan or macrobiotic or paleo or whatever, but if we discovered tomorrow clear-cut evidence that (e.g.) a diet high in refined sugars killed you as dead as tobacco, then yeah, I do think that would be really important, and it would impel us to figure out ways to make it easier for people to eat a low-sugar diet. In that scenario I don’t think we would — or should! — just say “screw it, it’s not realistic to expect people to give up sweets, sweets are delicious.”

  6. Rob H. says:

    Here’s a reason why I think it is important to know “what are the results of doing X?”. Doctors need to know; because it can work for some people. While it’s true that there might be some X for which most people will not comply, there’s always some people out there willing/capable/interested in complying. Here’s my issue: doctors seem to be trained to tell you certain things. Those things are not necessarily true/not necessarily the best advice for you. If you try to ask them about facts they kinda fall apart. This is something my wife and I have learned since having our son. You know, your child’s doctor will say “try not to give him juice at night”, and you just want to say: “Look he’s a year and a half old, the AAP recommends not giving any juice, so we don’t give him any juice, it’s really quite easy for us”. Similarly, we have no trouble keeping the TV off when he’s around. I’ve heard that the AAP was going to make a recommendation that children should not watch any TV before the age of 2, but they were told that this would be to difficult for parents, so they lowered the age to 1. They just recently brought it up to age 2. I think the medical community should be forthright about what facts are true, but offer a distinction between what is ideal and what is generally possible. Like Jordan’s idea of saying “hey if you do this, this is great, but we’ve found you only have a 1% chance of sticking with that, so don’t feel too bad.”. This spiel has been a bit incoherent…

  7. JSE says:

    Cathy, just to argue against myself a little: I stand by my claim that it’s important to understand the result of “do X” as well as “tell people to do X,” but I can’t deny that it’s massively harder to design an effective randomized trial of “do X.” You can’t just take the data from a trial like this and pull out the people who actually went low-fat — there may be too few of those to get good power, for one thing. More importantly, the people who follow doctor’s instructions are a biased sample in all sorts of ways, so you’re holding a one-way ticket to Confound City. You can do lab studies where you really do totally control people’s food intake, but then you have real questions about whether you can generalize the results to natural settings, even for people who follow the diet.

    Science is hard!

  8. Science _is_ hard. And science communication is hard too, in a different way. I’d love someone to study the science of science communication.

    So, for example, send out a mass-media message that’s nuanced and plays with our wish-fulfillment biases to half the country (“low-fat diets work but are nearly impossible for most people”) to half the country, and a mass-media message that’s dumbed down and not ridiculous (“eat olive oil instead of junk food, you won’t lose weight but you’ll be healthier”) to the other half. What is the actual effect of each and the relative effect? My prior is that the latter is more beneficial in the long term. It’s also possibly totally ineffective.

  9. JSE says:

    People totally study the science of science communication! What do you think professors of communication do? Well, they do a lot of things — but here at UW for instance we have an ENTIRE DEPARTMENT of life science communication…

    Also: as far as I can tell from the supplemental material, there’s not much evidence that the people on the Mediterranean diet decreased their junk food consumption substantially more than the people in the control group. So maybe a good message would be “pour olive oil all over your junk food, you won’t lose weight but you might protect yourself from stroke.”

  10. valuevar says:

    What about getting this message across: “Science shows that (a) no, neither butter nor potatoes are *intrinsically* fattening, in that fattening and slimming depend on only two numerical variables: if you take in more calories than you burn, you grow fatter; if you take in fewer calories than you burn, you grow thinner; (b) you need to take in some essential nutrients – most importantly vitamins and a variety of proteins (note: if you are a non-vegan in a wealthy country, you are most likely getting all the aminoacids you need in superabundant quantities already) but also certain fats and a some carbohydrates; (c) a good weight-loss diet is one that you can stick to for a year or two, and a good stable-weight diet is one that you can stick for the rest of your life, all while fulfilling (a) and (b). What goes into (c) depends on your preferences and habits, but, if you are not satisfied with current results, you will almost certainly have to be ready to change your habits.”

    Ta-ta! Lots of blah-blah becomes hereby unnecessary.

  11. valuevar says:

    Note: I lost 10kg in the last 14 months, and I just had fromage-blanc cake with berry sauce for dessert (after a medium sandwich for lunch).

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: