[ Question ]

How can I find trustworthy dietary advice?

by Daniel Kokotajlo1 min read17th Jan 202122 comments

30

NutritionWorld ModelingPractical
Frontpage

I currently think that the official dietary advice is sometimes untrustworthy, and I don't know when it's trustworthy and when it isn't, so I'm in a state of epistemic learned helplessness. Some reasons why I think this:

--I get the impression that much of the official advice of the past was actively harmful; the official advice has flip-flopped on many things over the past decades; things which were touted as healthy were later shown to be unhealthy, and vice versa. So presumably some of the current official advice is also actively harmful, and some of it is merely useless. But I don't know which.

--I get the impression that there are studies available supporting pretty much any conclusion; if I want to believe that red meat is bad for me, I can find mountains of studies showing that it is bad for me, and if I want to believe that it is good for me, I can go talk to carnivore diet people and they'll give me convincing evidence it's actually really good for me.

I'm optimistic that if I spent a few weeks of effort looking into the literature and reading the studies myself I could come to good opinions on this topic. But maybe someone here has already done this? Or maybe I'm just wrong about the official dietary advice?

New Answer
Ask Related Question
New Comment

10 Answers

I'm studying nutrition at a tertiary level for pretty much this exact reason and what it has taught me from speaking to dietitians who teach my course and from doing the course is:

a. Advice for the average person with a typical western diet boils down to "would it kill you to eat a damn vegetable?"

b.  There's a lot of organic chemistry that despite me being 3 years into a 4 year degree hasn't paid off (I am doing the degree that feeds into a masters of dietetics, so I'm sure that's where it was going to pay off. Alas, I'm not going to do that masters any time soon because my government engineering job 8 years in pays more than and end of career dietitian despite the higher level of education, and the masters would require me to work 9-5 for 6 months for free)

But most importantly, and most relevantly for the lesswrong sort of person, is that it is basically impossible to study diets in anything resembling double blind randomised placebo controlled. Diets are strongly linked to culture and personality and strongly influenced by those around you and virtually impossible to double blind. Like, imagine you wanted to study whether eating red meat is good for your health. Imagine getting 10,000 people and randomising them into three groups (high red meat, moderate red meat, no red meat). Would a steer farmer from Texas really stop eating red meat just because he was randomised into the no group? Would I, a vegan for 5 years, start eating 4 serves of red meat because I was randomised into the high group? (no, but I wouldn't sign up for the study for this reason, which is a confounder because people like me then wouldn't sign up). And if the Texan did stop eating red meat, would he keep it up for 20 years? Would someone randomised into the high group who got diagnosed with heart disease and was told by their doctor to cut down on red meat ignore the advice because of the study? 

It's that problem but writ large that makes decent dietary research hard to do. You'll notice a lot of studies are done over 1-3 months, because that's a reasonable amount of time to be able to provide three pre-packaged meals to your participants that you can control exactly (though those participants are probably going to eat other things: who goes to a birthday party without eating cake, for example?). 

So then we have to do animal studies, and our ancestral diet is very different from say a mouse or even a chimp, and is maybe not even what is best for us. 

Also, something that isn't really emphasised in this sort of discussion is the cultural value of food. Sitting and sharing meals with people is good for our mental wellbeing. 

I think some people also stick on "are eggs good or bad for you? is red meat healthy? are tomatoes good?" when this is kinda missing the point. No one thing should be such a big part of your diet that this information is gamebreaking. And I think everyone knows that vegetables are healthy and hamburgers aren't. 

So, where does that leave us? 

Fortunately, pretty much every country in the world has a team of dietitians who come together to make a guide for how to eat healthfully. I'm Australian so the Australian guide to healthy eating ( https://www.eatforhealth.gov.au/guidelines/australian-guide-healthy-eating ) and the associated material on the website is what I'm most familiar with. It boils down to eating a lot of vegetables (50% of your plate!), a moderate amount of grains, a small amount of lean protein, and eating fruit and calcium-rich foods. But the website has a lot of information on it in a very accessible format and I'd recommend it as a good starting point. 

But most importantly, and most relevantly for the lesswrong sort of person, is that it is basically impossible to study diets in anything resembling double blind randomised placebo controlled.

To be more exact, doing that is basically outlawed. You could run those studies in prisons but that wouldn't get past the ethical review board.

Indeed, since each body is a different ecosystem and each body exists inside larger, distinct ecosystems, you'll have to decide what your objective about health is and be satisfied with "good enough" and letting your body do its thing.

To illustrate the complications, people who descend from groups adapted to particular regions and who have now migrated or been forcibly relocated to others may have different dietary needs than the adapted, local groups. For example: someone living away from the equator traditionally consumed a lot more fish and milk derivat... (read more)

There are a number of places, assuming you don't just build one, where people are locked up and eat what they're given. No, that's not a perfect experiment, but it's better than no experiment at all.

I'd also imagine that coma patients are really cooperative experimental participants.

4ChristianKl4moThey don't tend to be cooperative about eating to the point where they have to be fed via tubes that bypass part of the eating process. They are also untypical for other reasons and are not willing to put in their daily steps.
1mad4moSetting aside the absolutely horrific ethical problems with experimenting on imprisoned people, imprisoned people do have access to commissary, would have access to special meals for religious or cultural reasons, and are likely to exchange food amongst themselves. And as said in the other reply, coma patients are fed through tubes and don't do things like exercise, so you wouldn't be able to determine the effect of protein intake on muscle growth or whatever, and they also have whatever condition put them in a coma in the first place, and I believe long term comas are rare. Oh, and no informed consent, because they're in a coma.
2Stuart Anderson4moLocked up and imprisoned are not synonyms. A good example are deep sea divers. They literally live in pressurised tubes and are fed via airlocks. Anywhere that has to ration or import food would also be a possibility. Whenever you can list reasons something won't work is also automatically a list of criteria to look for situations where those reasons are minimised or absent. I bet they don't share markers in their excrement and urine though. If you think people are liars or gaming a system you redesign the system in a manner that makes cheating very difficult or impossible. By suggesting this as something to think about rather than do we are now having a discussion about their lack of suitability for many experiments. They are not a general model. Had that crossed my mind I wouldn't have suggested them, but I've learned something. The question that springs to mind subsequent to the digestive differences of the comatose: do their brains consume less glucose? In the ICU we have a population that are generally extremely sick, some of which are not conscious, and some of which will almost certainly die. They're also getting constant obs and all their needs externally provided. People don't have to be in legit comas for years for us to be able to benefit medical science as a whole by figuring out how to make them die less. Diet could be a part of that, but we don't know unless we actually check (and of all the interventions possible on a patient what you feed them is likely to be least impactful). I can guarantee that there will be relatives giving consent to studies, either in the hope of improvement for the patient, or more rarely in the knowledge that the patient is going to die, so something good might as well come of it.

Advice for the average person with a typical western diet boils down to "would it kill you to eat a damn vegetable?"

I heard a hypothesis that all "diets that work" have one thing in common (which is probably the only reason they work) -- they recommend eating more vegetables than you were eating previously, but they achieve it mostly indirectly, by banning something else. Also, they indirectly make you eat less, by making you pay more attention to what you eat, and banning some of your previously favorite meals.

For example, vegetarian or vegan diet seems l... (read more)

I find Cochrane reviews to generally be of good quality, even if that means there findings are very often "we have reviewed all available data as of X year and we not able to draw any clear findings". Depending on your technical knowledge it might be useful to point out that they generally include a "plain language summary".

It is important to note that not everyone agrees with their findings (the abstract of https://pubmed.ncbi.nlm.nih.gov/16052203/ is well worth reading, not just as a criticism of Cochrane but as a comment of the field of research in general). I suppose one could reasonably argue that a combining a load of crap observational or small RTCs (with high drop out / low protocol adherence) is not going to teach you very much, yet this is what systematic reviews of the field tend to do.

You do occasionally see nice articles like https://pubmed.ncbi.nlm.nih.gov/32144378/ which follow 142 people for 2 years. It's still not a lot of people and not that long to draw conclusions for many different lifestyles and cultures across the entire human lifespan and across the world.

See https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD002128.pub4/full for an example of a Cochrane review, and for a list of advice see https://nutrition.cochrane.org/evidence

Or see this almost comical example, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7427685/ "12,133 records identified, 30 studies met inclusion criteria ... despite the large number of trials included in the review ... evidence for primary prevention on clinical endpoints is limited to one large trial with methodological issues ... [hence] there is still uncertainty regarding the effects of a Mediterranean-style diet on clinical endpoints and cardiovascular disease (CVD)"

I also second peoples point about examine.com - it's good.

Also, it's not really directly answering the question but I want to rant. Calories-in must equal calories-out but in a way that has so many caveats that it's not really a useful measure of anything.

  • Processing: A tree, if put in a calorimeter will tell you it has a lot of calories (350 Kcal per 100g for red oak), but if you eat wood, you will shit out wood, we cannot process lignin so cannot extract the calories. Potatoes are interesting, because uncooked, most people (I think this varies by person) cannot process the calories nearly as well, hence the same potato has a different amount of energy depending on how you cook it. Peanuts are calorie dense, but unless you really chew them into a paste (which most people don't) then you will not be extracting the full calories, hence grinding peanuts into peanut butter increases the effective calories. In short the calories you extract depends on whether you can digest the food, which partly depends on how you prepare it (and partly depends on what species of animal you are and partly on your gut microbiome).
  • Set point theory (https://pubmed.ncbi.nlm.nih.gov/30026913/) If you exercise more you will want to rest more and you will want to eat more. So if you burn off 100 calories more than you usually do then you might lower your energy expenditure in other ways to make up for it. Or you might eat more than you usually would. A similar thing is probably true if you change the amount you eat- if you eat more you might expend more energy (I know there are studies on this, I was asked to be part of one where I would have to eat and extra 50% calories per day, but I haven't read the studies). The way these interact is complex and there isn't a whole load of agreement on what is going on.

I've adopted a two-pronged approach:

  1. What did my ancestors eat?
  2. What does the science say?

The first approach involves tracing my lineage to figure out what they ate. My family is Russian and Ukrainian, so the idea is to figure out what the most traditional foods are. I assume my genetics is going to be optimized for that environment. For me this means more cabbage and more buckwheat. However, though potatoes are popular in Russian cooking, they're a modern addition.

 

The second approach is to figure out what vitamins, minerals, etc my body needs. Essentially, you treat the body like an old-school Ford factory where raw materials come in one end and the end product (healthy cells) comes out the other. Another way to think about this is that if I had to put a feeding tube into by blood system, what does modern science say the mix of ingredients should be? Yet another thought experiment: take a healthy person and figure out what elements from the periodic table of elements exist in the body and in what proportion. If a healthy body contains magnesium but it's not in your diet, clearly you'd see some health consequences from this.

 

Finally, synthesize the two approaches. For example, if you have darker skin, you'll probably need more Vitamin D if you live up North. It may also be useful to pay special attention to "weird" foods of your ancestors since they may contain a beneficial nutritional profile that filled a gap. Buckwheat was one such food for me. It's packed with nutrients, but eating the same foods may not be enough. The same food grown in different geographies will have different nutritional profiles. I know rice can have varying composition depending on where it's from. In theory, you can become nutritionally deficient even if you eat the same foods as your ancestors and it could come down to soil composition.

 

The hybrid approach may seem complex, but it's simple to start since you can start eating what your ancestors ate and then take a lab test for nutritional deficiency to see if you're lacking in anything. The benefit of the hybrid approach is that they balance out each other's weaknesses. Science can't say anything about a chemical it hasn't studied. Even if it's studied, the science may only make claims about what's been observed or what can be inferred based on known mechanics. Traditional foods tell you what your genetics are likely adapted to. This may include chemical soups that science hasn't gotten to yet. However, just because your ancestors ate a certain way doesn't mean it was optimal for them. Finding the perfect diet is a back and forth between science and tradition.

I recommend the book "The Renaissance Diet 2.0" which does a competent job summarizing the science.

I think a productive way to look at it is to look for absence of evidence, which is evidence of absence.

Much has been said about "the western diet" that is killing tons of people, but in reality, what we really know is that being obese is bad for you, as is having severe nutrient deficiencies. Otherwise not a whole lot much is sure.

Let's take an example. Studies on meat consumption barely find a significant effect on all-cause mortality. But most often they fail to control for things as basic as pre-existing obesity or caloric intake. And if you step back one second, it's rather obvious: people who voluntarily reduce their meat consumption (or abstain completely) tend to be quite health-focused. All the people with a junk diet are on the other side. That ought to tip the scales, but even then, the finding is minuscule.

This absence of conclusive evidence really does tell you something: diet is much less impactful (to your health) than people give it the credit for. Consider, for contrast, that being partnered adds, on average, years to your life (I suspect this finding might also have a control problem, but the magnitude of the finding is in another league).

Something where diet really has an impact is your day-to-day well-being. I don't have a solid proof, but it seems to me that if your diet makes you feel like crap, it might not be that great in the long run, and vice-versa (beware deficiencies though, which take a long time to show up).

Unsolicited diet advice:
- Control your calories (track! I guarantee you will be surprised and learn something)
- Eat enough veggies / supplement to avoid deficiencies. You don't need that much, and you don't need to hit the RDAs necessarily - they're incredibly hard to hit with food only, so I supplement to be on the safe side. Also, animal products do actually have a ton of nutrients, compared to what popular wisdom sometimes seems to imply.
- Eat enough proteins. It's really hard to eat too much proteins, it's incredibly good for a ton of things, and it's generally filling besides.
- Eat enough fibers (the better argument for eating more veggies). Personally, my digestive system is weird and I actually supplement this as well (as psyllium), makes a huge difference, but I expect this is quite personal.
- Don't sweat the rest, and enjoy your food!

Imho, these are the 80-20 (or really 98-2) rules of nutrition. I've never seen any evidence that any of the intricate fluff makes any difference.

... or at least at the population level. It might worth experimenting with your own potential intolerances/needs (e.g. like my need for a ton of extra fibers). But that's not something you'll get as general advice. I do think there must exist books or resources on the subject however.

I get the impression that much of the official advice of the past was actively harmful; the official advice has flip-flopped on many things over the past decades; things which were touted as healthy were later shown to be unhealthy, and vice versa. So presumably some of the current official advice is also actively harmful, and some of it is merely useless. But I don't know which.

This is not, in itself, a reason to stop trusting official advice on nutrition. On the science hierarchy of how certain we are of our understanding, nutrition science is very low and physics is very high. So we can treat all nutrition science recommendations as "we believe this slightly more than we believe the opposite". So it will take a little bit of counter evidence for the field to update its point of view to the opposite. But once that has happened, it will take more counter evidence to flip back. The flip flops will get rarer as evidence accumulates, and at every point in time it could still be that official recommendations are the best that can be made given the data that we currently have. I'm not saying this is the case, but it is consistent with the data of flip flops.

Unfortunately, the official advice lags our current understanding (by decades, maybe). I think that's a much bigger problem than the perceived flip flopping.

I don't have any advice for nutrition, but examine.com is pretty decent for information about supplements and vitamins, and the NIH guide on supplements is even better: https://ods.od.nih.gov/factsheets/list-all/

Have you considered the reason for this is that optimal diet is highly variable, both genetic and environmental.

For example whatever a 16yo school going boy from Jordan and a 56yo recently retired woman from Nebraska ought to eat to maximize health/performance/happiness might be entirely different and, if switched, have disastrous effects.

But these effects could also be present in much more similar individuals and even in the same individual over time. Most humans used to eat entirely different diets depending on the weather.

90% of it comes down to "eat more vegetables" and "cook what you think are balanced meals, from scratch." Also don't smoke and drink less alcohol. And if you are overweight, eat a little less.

But I also appreciate the nutritional advice on places like examine.com.

UpToDate is a good starting point.

4 comments, sorted by Highlighting new comments since Today at 10:55 PM

I am a Registered Dietitian from the United States. I have had to recognize that my education and practice in the field was not immune to the effects of US food policy, which includes commodity farming and corporate influence. The Dietary Guidelines for Americans (and I'm assuming other countries) are guidelines optimized for many things, health being only one.

I think the best advice is the simplest--as many have said, you probably aren't eating enough vegetables. You could do a lot worse than following similar diets to those who live the longest, in case there are some causal links there. The Mediterranean or Blue Zones type diets approach this.

Mark Bittman and David Katz wrote a nice article that I like on the topic: Here

Most importantly, realize our bodies are highly adaptable and there is no universally perfect diet.

My impression is that this is a perennial topic not because there isn't a good answer, but because people don't like the answer.

This is a perennial topic as it's incredibly difficult to study.

You can't do a double blind RCT on the effects of eating / not eating x on longevity.

As a result you're limited to short term clinical studies and epidemiological data. Both of which come with their own set of problems.

The official dietary guidelines are based on the best available science but there's still plenty of things we don't know for sure.

There's also issues of risk communication which further confuses things.

Processed meats for example are classed as a probable carcinogen based on observational data. You'll often see it stated that consuming processed meats increases your risk of colon cancer by 20% which sounds horrific. But that's relative risk.

The data actually shows that people who eat zero processed meat have a lifetime risk of colon cancer of 5%, whereas those that consume the highest levels (>50g per day) have a risk of 6%.

So while it's quite possibly true that there is a causal link and it's not just confounding factors, it's hardly a terrifying statistic, though the headlines typically make it sound like having a slice of bacon is akin to smoking a pack of cigarettes.

Plus, self-reported dietary information is garbage.