Bryan Caplan  

Of Monkeys and Micro

PRINT
Arnold Kling on "Libertarians ... Libertarians and Jim Crow Bleg...
Calorie restricted diets demonstrably increase the lifespans of yeast, fish, rodents, and dogs.  Will they work for humans?  For obvious reasons, controlled human experiments are problematic.  Researchers therefore turned to the next best thing: experiments on non-human primates - rhesus monkeys to be specific.  Two major studies began in the late 1980s - one in 1987, the other in 1989.  The main results are now in: calorie restriction failed.

Results from the 1987 study are openly negative:
It turns out the skinny monkeys did not live any longer than those kept at more normal weights. Some lab test results improved, but only in monkeys put on the diet when they were old. The causes of death -- cancer, heart disease -- were the same in both the underfed and the normally fed monkeys.

Lab test results showed lower levels of cholesterol and blood sugar in the male monkeys that started eating 30 percent fewer calories in old age, but not in the females. Males and females that were put on the diet when they were old had lower levels of triglycerides, which are linked to heart disease risk. Monkeys put on the diet when they were young or middle-aged did not get the same benefits, though they had less cancer. But the bottom line was that the monkeys that ate less did not live any longer than those that ate normally.

Results for the 1989 study are superficially positive but substantively negative.
[A] 2009 study from the University of Wisconsin that concluded that caloric restriction did extend monkeys' life spans.

But even that study had a question mark hanging over it. Its authors had disregarded about half of the deaths among the monkeys they studied, saying they were not related to aging. If they had included all of the deaths, there was no extension of life span in the Wisconsin study, either.

My question for you: How do these results make you feel?  Most people react with joy.  "Yay!  The dietary change I definitely don't want to endure doesn't work anyway!"  On reflection, though, reconciling this elation with basic microeconomics is extremely difficult.

Calorie restriction is a choice - an option.  The better calorie restriction works, the better your options.  The evidence that calorie restriction does not work implies that we are effectively poorer.  Living longer by half-starving ourselves is off the menu.

You could object, "I'd rather live a normal, pleasant lifespan than a long lifespan of constant hunger."  If that's your story, though, you should at least be indifferent to the outcome of the research.  Furthermore, if you care even slightly about people who value longevity more than a full belly, the failure of calorie restriction is still bad news.  You're no worse off, but people you care about are poorer than you thought.  On reflection, only misanthropes have a reason to jump for joy.

What's going on?  This seems like a classic case of cognitive dissonance.  Once you make up your mind to do X, thinking about the advantages of Y makes you unhappy.  The conclusion that X is better than Y all things considered does not put our minds to rest.  We feel best when we conclude that X is better than Y in all respects

Not convinced?  Imagine that researchers discovered that calorie restricted diets made us immortal.  Would any sane person consider this bad news?  No way.  Even if you'd rather die than endure an eternity of hunger, you almost surely care about someone who'd happily make that trade-off.

Of course, there's another sense in which we should be happy about the negative results.  Given that calorie restricted diets don't work, we should be happy to discover the truth so people can stop suffering in vain.  Basic microeconomics doesn't just imply that more options are good.  It also implies that false hope is bad.



COMMENTS (34 to date)
gwern writes:

Cognitive dissonance does not seem like the right silver bullet to use here...

People know they are by and large unable to maintain a regular weight and diet; so they know, a fortiori, that a CR diet (intermittent fasting rarely gets discussed, so we'll ignore that possibility) will be even more unworkable than a regular diet.

So if CR panned out, they know they are *losing out* out on decades of better health in old age. Is it pleasant to know a health loss has been forced on you? Of course not - we are loss-averse! And further, the loss is not forced on you by an outside actor, but your own weakness - so the loss aversion is compounded by further guilt over your weight & diet.

Cognitive dissonance may be a factor for some CR critics (they should be hoping it works, not that it fails), but for most readers who have never heard of it? Loss aversion - the scenario in which they inflicted huge losses on themselves has been avoided by the null results.

AS writes:
"Yay! The dietary change I definitely don't want to endure doesn't work anyway!"

Or maybe you currently have a calorie-restricted diet and are now happy that the study shows that you now have the option of going off your diet.

The general point of this post could be stated more directly: we should be happy when better information leads us to a positive change in our lifestyle. We should not be happy when better information demonstrates that other lifestyles are worse than ours.

Sieben writes:

"Males and females that were put on the diet when they were old had lower levels of triglycerides, which are linked to heart disease risk. Monkeys put on the diet when they were young or middle-aged did not get the same benefits, though they had less cancer."

If there isn't a physiologically plausible explanation for this, then it's evidence that the study is bunk.

"Its authors had disregarded about half of the deaths among the monkeys they studied, saying they were not related to aging. If they had included all of the deaths, there was no extension of life span in the Wisconsin study, either."

Right. Like if you're gonna get hit by a car then the anti-cancer benefits of CR probably don't matter. So what? Isn't it valuable information to know that CR monkeys died fewer age-related deaths?

There are good plausible reasons for CR/IF to work in monkeys. But there are even better reasons for them to work in overweight sedentary americans.

http://en.wikipedia.org/wiki/Intermittent_fasting#Human_studies

"[We] have noticed an improvement in the last few months in the ability of these patients to keep their weight under control by observing one fast per week [water only]. This allows them to be more liberal with their diet on the other days. I cannot overemphasize the fact that they prefer this to perpetual daily denial with no alternative."

Very nice reasoning, Bryan.

Ryan Langrill writes:

Or we could be happy that people who want to restrict caloric intake through public policy have no scientific support?

Alexei Sadeski writes:

Hmmm... I am happy to hear the evidence that calorie restriction doesn't seem to work, but why can't this simply be because I believe alternate dietary regimes superior - and thus am happy to hear that a dietary regime which I always considered difficult to practice has been found useless as well?

louis the briton writes:

People do tend to feel obligated to pursue longevity even if they would get more enjoyment out of de-emphasizing it. The reasons that such an instinct would be in place are obvious.

I am certainly delighted to know I won't have to pursue my breatharian diet.

The Sheep Nazi writes:

Not convinced? Imagine that researchers discovered that calorie restricted diets made us immortal. Would any sane person consider this bad news? No way. Even if you'd rather die than endure an eternity of hunger, you almost surely care about someone who'd happily make that trade-off.

Here is a thought experiment: try substituting the words "accepting Jesus Christ as your personal savior" (not that I do, not that I do ... ) for the words "calorie restricted diets", and then raise your hand if you still give a flying monkey about the microeconomic analysis thereof.

Tracy W writes:

I think the logical reaction is:
"Yay! We have ammunition when someone proposes rationing food to extend life-spans in favour of public health."

There are a lot of people who think it's a good idea to regulate others' lives to improve their health outcomes.

Sieben writes:

A couple of commenters seem relieved that this study will not be used to justify draconian public policy. But the study was only testing life-extension properties of CR. CR already has numerous incontrovertible health benefits for overweight Americans.

They've had all the intellectual ammunition they've needed for decades.

gwern writes:
Right. Like if you're gonna get hit by a car then the anti-cancer benefits of CR probably don't matter. So what? Isn't it valuable information to know that CR monkeys died fewer age-related deaths?

Pure post hoc datamining, Sieben - was that endpoint registered *before* the study started? I think not...

And even at face value, it eliminates the health benefit: the analogy here is not being hit by a car (for monkeys in a research facility?!), it's having blood drawn. If every time you go to the doctor under a CR diet, you're taking your life in your hands, then it's highly questionable it *could* deliver any net longevity gains...

sieben writes:

"Pure post hoc datamining"

Oh, arguing through connotation. My favorite. Is there anything wrong with throwing out the non-age related deaths? They interfere with the hypothesis that CR reduces age-related deaths. And it doesn't matter that it's "post hoc". It has to be post hoc because you don't know in advance each subject's cause of death.

"was that endpoint registered *before* the study started? I think not..."

Does it have to be? They used the same criteria for both CR and control monkeys. I don't see why it would bias the study in favor of CR. Quoting:

"Seven control and 9 CR animals died of non–age-related causes, which included complications of anesthesia, gastric bloat, endometriosis, and injury."

I see absolutely no reason why we shouldn't throw out drug complications and injury. Gastric bloat, as far as I can tell, is a twisted-stomach condition. Yada yada...

"And even at face value, it eliminates the health benefit: the analogy here is not being hit by a car (for monkeys in a research facility?!), it's having blood drawn."

Having blood drawn with a sedative*. I don't know what kind of sedatives they used, nor how dangerous these sedatives are in monkeys. But if you've ever been put under anesthetic you had to sign a whole bunch of waivers in case you couldn't be revived.

And the twisted-stomach thing is kind of like getting hit by a car.

"If every time you go to the doctor under a CR diet, you're taking your life in your hands, then it's highly questionable it *could* deliver any net longevity gains..."

Right. You're just complaining that the benefits might be too small to be applicable. If you're a Cuban Guerrilla warrior, and your life expectancy is 5 years, there's no point in CR. Similarly if you're a monkey and you're only going to live a few decades, the benefit might be really small.

The physiology of CR affects things like oxidative stress, which accumulates over many years. But even though the monkeys only lived into their late 20's, it was still protective. What about humans who are going to live into their 80's? We should expect the benefits to be far greater in humans than short-lived primates.

Also extend that the monkeys are markedly healthier than your overweight (pre)diabetic American. What do you think the results of the study would be if the control group had a 40% obesity rate?

BLM4L writes:

I second all the comments suggesting that the social impact of the data is more relevant than the personal impact. Now we can forward the data to our busybody mothers who are always encouraging us to go on a diet, etc.

Hasrrubal writes:

I'm not even thinking of any social impact or government paternalism.

For me, the fact that calorie restricted diets do not increase life expectancy is a positive benefit regardless of whether or not I would have followed such a diet: If they were to work, I would have to suffer the disutility of my mother nagging me about not starving myself and self righteous lectures from one more group of diet nuts. That's a significant cost to me, one which outweighs the benefit I would receive from knowing that other people would be happier to eat less and live longer.

I don't see any cognitive dissonance.

sieben writes:

Maybe I'm the wrong generation, but the mothers I know do not nag about health or diet.

23/m/tx

Are you all younger or older than me?

Breaking news: "Growing long hair does not increase lifespan"

Me: "Yippie, I can continue my current practice of keeping short hair without worrying that it will kill me. That was a relief!"

Mr Caplan: "Misanthrope, we should welcome long hair and longer lifespan results, if it were true."

Me [dazed and confused]: "ummm... what?"

Zac Gochenour writes:

I think there may be something to the mother's scorn idea - recommending difficult ways of improving health is a common "showing that you care" signal (diets and exercise in general, eating your vegetables, etc) and rejecting that advice causes disutility (guilt, or further nagging). sieben - I would think that kind of nagging would be worse for middle-aged men, does your grandmother nag your father about health? So, folks rejoice that their loved ones no longer have this weapon against them.

Bryan writes,

Of course, there's another sense in which we should be happy about the negative results. Given that calorie restricted diets don't work, we should be happy to discover the truth so people can stop suffering in vain. Basic microeconomics doesn't just imply that more options are good. It also implies that false hope is bad.
On the darker side, folks are unhappy if they've been following calorie-restricted diets because they can less effectively claim that their choices are superior to others. They have until now suffered in vain. They must accept that others willing to sacrifice less will live just as long.

Suppose strong evidence is published that shows recycling does nothing for the environment. Now recyclers can stop suffering in vain, right? I don't think they'd be too happy to discover the evidence.

gwern writes:
Oh, arguing through connotation. My favorite. Is there anything wrong with throwing out the non-age related deaths? They interfere with the hypothesis that CR reduces age-related deaths. And it doesn't matter that it's "post hoc". It has to be post hoc because you don't know in advance each subject's cause of death.

Nothing 'connotation' about it: data mining is a perfectly specific criticism with well known solutions like carefully targeted data collection, multiple correction (Bonferroni or otherwise), pre-registration of analysis, etc.

None of which that study did. As far as we can tell, the researchers sliced the data until finally they were able to show *some* benefit from CR.

They used the same criteria for both CR and control monkeys. I don't see why it would bias the study in favor of CR.

Er... do you really think in data mining that people are applying *different criteria* to the groups? Of course not, the whole point is they're testing who-knows-how-many criteria to find one which finally works. If you look hard enough, you'll always find some way to distinguish two groups. If they couldn't post-hoc split the deaths to reach p<0.05, maybe they would have been able to split by weight; or maybe by cholesterol level; or maybe by gender; or maybe by age of starting; or maybe...

Having blood drawn with a sedative*. I don't know what kind of sedatives they used, nor how dangerous these sedatives are in monkeys. And the twisted-stomach thing is kind of like getting hit by a car.

Yes, as we all know, there's no possible way 'aging' could possibly affect in any way the mortality rate for these things, so it's perfectly legitimate to post-hoc exclude any mortality cause we want!

Right. You're just complaining that the benefits might be too small to be applicable. If you're a Cuban Guerrilla warrior, and your life expectancy is 5 years, there's no point in CR. Similarly if you're a monkey and you're only going to live a few decades, the benefit might be really small.

The effect size complaint is true, but not what I'm getting at here: I'm pointing out that they took the most meaningful endpoint, all-cause mortality, and realized it was not proving their point so they tortured it until they came up with an endpoint that did show their point. All-cause mortality is the most important endpoint! This is why drug trials mandate pre-registered endpoints, because *side effects matter*.

It doesn't matter if CR reduces cancer if it massively jacks up deaths from, say, falls and causes a lifespan decrease. (Or jacks up stomach twisting or anesthesia...)

sieben writes:

"Nothing 'connotation' about it: data mining is a perfectly specific criticism with well known solutions like carefully targeted data collection, multiple correction (Bonferroni or otherwise), pre-registration of analysis, etc."

You know very well that you implied that they just arbitrarily selected data to cook their results. The technical definition of data mining is too non-specific to be a criticism.

"Er... do you really think in data mining that people are applying *different criteria* to the groups? Of course not, the whole point is they're testing who-knows-how-many criteria to find one which finally works."

Oh, so it's a conspiracy theory now. Do you have an actual objection to their filtering criteria? The raw data still even points towards the hypothesis. It isn't like they're reversing a trend. http://www.sciencemag.org/content/325/5937/201/F2.large.jpg

"Yes, as we all know, there's no possible way 'aging' could possibly affect in any way the mortality rate for these things, so it's perfectly legitimate to post-hoc exclude any mortality cause we want!"

As long as they are excluded from both groups, I fail to see how it is necessarily unfair. In fact, they HAVE to throw out these deaths because the sample size doesn't have huge statistical mass. They could just roll with large numbers if they had 500 monkeys, but they had 30. And if there's no reason to blame a twisted stomach on age, you should throw it out.

"I'm pointing out that they took the most meaningful endpoint, all-cause mortality, and realized it was not proving their point so they tortured it until they came up with an endpoint that did show their point."

There's no evidence for this story. You're making it up. For all you know, they could have decided in advance they were going to exclude non-age related deaths. In fact, they probably did unless you think they were ignorant about the non-age related ways monkeys die.

"It doesn't matter if CR reduces cancer if it massively jacks up deaths from, say, falls and causes a lifespan decrease. (Or jacks up stomach twisting or anesthesia...)"

For which I'd really really love to see a physiological explanation. That's the whole point of "data-mining" - you filter data intelligently according to rules you already know.

Finch writes:

I was under the impression that lighter people tend to die disproportionately from injury and infectious disease because they're easier to injure and have less reserves to battle with (hence humans have lower death rates when mildly overweight). It makes perfect sense that the same thing would apply to monkeys, and excluding that data seems deceptive.

sieben writes:

Finch, but these results are from an epidemiological study. Being slightly overweight is correlated with all sorts of things. For example, people eat larger amounts of food at social events, and having a supportive social structure also extends lifespan.

Etc etc...

Finch writes:

The epidemiological study, if I understand correctly, showed no positive effect unless you post hoc exclude exactly the sort of thing you'd expect as a side effect of this treatment. Maybe I don't understand your point...

It's still conceivable that calorie restriction makes sense for the aged, where heart disease and cancer are more important causes of death, but when you're young and disease and injury dominate your risks, it looks like you want to be a little heavy.

sieben writes:

"The epidemiological study, if I understand correctly, showed no positive effect unless you post hoc exclude exactly the sort of thing you'd expect as a side effect of this treatment. Maybe I don't understand your point..."

http://jama.jamanetwork.com/article.aspx?articleid=200731 Is what I'm looking at. I don't see anywhere that they exclude deaths.

"It's still conceivable that calorie restriction makes sense for the aged, where heart disease and cancer are more important causes of death, but when you're young and disease and injury dominate your risks, it looks like you want to be a little heavy."

Conceivable. If I were going to ride a motorcycle, maybe.

Oxidative stress is not directly related to bodyweight. But it is related to overall calorie intake, and is cumulative over a lifetime. Similar with plaque in your cardiovascular system...

Steve Rose writes:

Does microeconomics imply more options are always better? Since we don't have a crystal ball to tell us which is the best option some of our limited resources must be used to search for it, the more options the harder the search. Each new option will have diminishing marginal utility.

Finch writes:

> I don't see anywhere that they exclude deaths.

I'm sorry, I thought you were referring to the study we were discussing. My fault. I meant the monkey death exclusions were deceptive.

> Conceivable. If I were going to ride a
> motorcycle, maybe.

That's not the only source of death for young people. I don't ride a motorcycle, but I'm worried about car accidents, getting sick, medical mistakes, that kind of thing. I'm young enough that cancer and heart disease are worries for the future. Besides, it's hard to be athletic and fit on a calorie restricted diet, and that also affects heart disease risk, so there are other things to consider. Frankly, it's hard to be athletic and not be mildly overweight.

Andy writes:

What if we found that going to college (another very costly activity) actually has no benefit on your life, and that perceived gains are strictly because of selection bias? I imagine those of us that went to college would be unhappy to learn this and deny it, while those that didn't would be pretty happy about it.

So I do agree with Bryan that cognitive dissonance is likely the major explanation here. Of course it's always possible to rationalize it...

Mark Crankshaw writes:

On reflection, only misanthropes have a reason to jump for joy.

As others have alluded, my first reaction to learning that the finding wasn't supported would be: "Whew, now some leftist paternalistic knob doesn't have the justification, right or wrong, to use the government to forcibly starve me 'for my own good'".

Having the ammo taken out of the hands out of the various ideological 'do-gooders' who litter the world is reason enough to jump for joy for me...

Tracy W writes:

Sieben: But CR could increase your chances of dying of non-age-related causes.
For example, good reflexes can help you avoid dying in a car accident, indeed, can help you avoid getting in a car accident in the first place. And good reflexes are helped by good nutrition, plus not being distracted by feelings of hunger.
A good food source in terms of a layer of fat might plausibly help in fighting off an infection, or the fat itself might plausibly help by being natural insulation.

Sieben writes:

Finch,

"I'm sorry, I thought you were referring to the study we were discussing. My fault. I meant the monkey death exclusions were deceptive."

They are deceptive if you read the NYT article. If you read the original study, and think about it, they are not deceptive.

"I don't ride a motorcycle, but I'm worried about car accidents, getting sick, medical mistakes, that kind of thing. I'm young enough that cancer and heart disease are worries for the future."

I'm worried about car accidents. I don't think there's anything you can do about getting sick because the immune system does not appear impaired in normal or slightly-overweight.

Cancer and heart disease should be on your radar *now*. Even though you won't die from them until you're older, you'll get them depending on your young and middle aged lifestyle.

"Besides, it's hard to be athletic and fit on a calorie restricted diet, and that also affects heart disease risk,"

Intermittent fasting has the same benefits as CR, and doesn't involve CR.

"so there are other things to consider. Frankly, it's hard to be athletic and not be mildly overweight."

I assume you by "mildly overweight", you mean carrying fat. Have you seen the crossfit games? Those kids have a borderline mental disability with their diet and training, but consistency allows them to be ripped and put out a relatively decent level of all-round fitness.

Tracy W,

"And good reflexes are helped by good nutrition, plus not being distracted by feelings of hunger."

Okay, so most CR regimens are merely hypocaloric until your bodyweight decreases enough to make them calorie-neutral. I don't know that there's any evidence that people on CR are actually physiologically hungry. Your brain would lose sensitivity to the appetite signaling hormones if they were elevated for the rest of your life.

Also note that plenty of overweight and obese people deal with hunger, and they deal with it much more poorly than the lean. Poorer insulin sensitivity leads to larger blood sugar swings, and you fill in the rest. These people have to be constantly snacking on something - god forbid they miss breakfast! - or they start feeling weak and getting irritated...

At any rate the problem most people have with diets is not hunger per se, but rather an addiction to food. The literature is extensive on how high-density palatable foods mimic and are comparable to opiates and amphetamines.

But even these "withdrawals" will fade out quickly in a month or so.

"A good food source in terms of a layer of fat might plausibly help in fighting off an infection, or the fat itself might plausibly help by being natural insulation."

Blankets come pretty cheap now. I don't know that being slightly overweight helps the immune system, nor that being permanently "underweight" would harm it.

Fasting is also anti-inflammatory, and the remedies for many illnesses are to take a NSAID. So it really depends on the specific illness you're talking about.

thruth writes:

Less impressed by this than other commenters seem to be. The social problem that matters is the largish number of people who are obese not the typical person carrying a few extra pounds. Assume it would not be ethical to feed monkeys multi thousand calorie diets and sit them on a couch in front of a tv all day, but my strong prior is that this would cause major health issues and considerably shorten life spans. (However, perhaps only some monkeys would be susceptible to this and others would self regulate)

sieben writes:

Sleep deprivation and indoor lighting also disrupt appetite regulation.

It's a whole huge interconnected continuum. Testing just one thing is kind of useless. They should use simple physiology to come up with a comprehensive "health and longevity" plan, and then compare that with a control group.

That way you capture the synergy between habits, and synergy is probably the most emphasized thing in the lifestyle literature.

mobile writes:

Bryan, you can call me cognitively dissonant or any other name in the book. You are not going to get me to look at this

http://marginalrevolution.com/marginalrevolution/201
2/09/life-extension-that-works.html

and tell me it should make me happier.

John T. Kennedy writes:

My reaction was "Too bad, a life extension option would be good."

Tracy W writes:

Sieban, the point is that it's plausible that CR could increase the odds of dying from a non-age related cause. You yourself, in your examples of overweight and obese people dealing with hunger, and the possible anti-inflammatory benefits of fasting, illustrate other ways in which diet might affect odds of survival.

Comments for this entry have been closed
Return to top