I’m about to launch a series of posts about nutrition that came from reading I did for my own health due to my heart valve issue, and also that went into my book. I was hesitant to go into such detail here because I really think the “no-junk” guidelines I talked about previously are a good start. There’s a danger of getting too hung up on the details, it can actually lead to an eating disorder called orthorexia.
I am a amateur on the subject of nutrition, although I believe a pretty well-read one. If you want to learn from a professional, I’d highly recommend fellow blogger Dr. Christine Rosenbloom’s site. She has a PhD in nutrition and many years experience, and has also written the excellent recent book Food and Fitness After 50 (with co-author Dr. Bob Murray, an exercise physiologist).
But there is a lot of conflicting information “out there” (media, internet, etc) about nutrition so I thought my efforts to put it all into perspective might be helpful, bearing in mind that this is a layman’s assessment (although a well-read and hopefully reasonably intelligent one).
As I discussed in my eating story, I became interested in healthy eating when I was about 45 and noticed a little “middle age spread” for the first time, and have been reading up on it ever since. I became more interested when I got diagnosed with Aortic stenosis. In trying to figure things out for myself, I have read extensively on various ways of eating, with an open mind. I do not take what any of the authors say at face value but always try to chase down the underlying science. This probably amounted to reading dozens of books and hundreds of scientific papers over the course of about 18 months. What can I tell you, I’m a nerd and enjoy reading stuff like this. In future posts on nutrition, I’ll go over what I have learned and what works for me. Don’t worry it will be as condensed as possible. I am not trying to convince you that my way of eating is the best, or debunk anyone else’s, just present the evidence as I see it. I welcome comments if anyone thinks I have missed important references or misinterpreted anything.
My technical training, while not directly relevant to nutrition, does make me comfortable wading through scientific literature in general. Here I want to give a little background on how I go about interpreting nutrition science because not all references are as high quality as others.
On Interpreting Scientific Studies
We all learned this at around the sixth grade but it’s worth a review. Even though I’ve followed this method doing engineering research for years, I had to refresh my memory how they managed to stretch it to 6 steps. The most important thing in the scientific method is that we keep going back from step 5 to step 3:
One of the important things is the quality of the data that is observed in these steps. I’ll hit the high notes about that to see how to tell the quality of a study.
In nutritional science, the most obvious data to start with is population data. For example, we’ll see in a future post under “current (or recent) healthy populations” that many populations around the world are healthy on their traditional lifestyles and diet, and become considerably less healthy when adopting a more modern diet and way of life. Traditionally they have lower animal product consumption, lower consumption of processed foods, and higher amount of physical activity, compared to typical modern societies, A hypothesis that something is healthier about this combination fits the data nicely.
It could be that it has nothing to do with the animal products or the exercise, it’s just the processed food, or that it’s just the exercise, etc., or that it really does require the combination of all three. Suppose we instead offered the simpler hypothesis that it’s healthier to eat less animal products. The data clearly does not yet prove that, so we would need more. This is typical with population studies, that there are multiple factors that need to be teased apart by further evidence.
The gold standard nutritional study is an interventional study on humans, with a large enough sample size. Suppose five people in group 1 ate less meat and their health got better, five people in group 2, the control, did not eat less meat and their health stayed the same. That’s not enough of a sample size to be statistically significant (prove that the difference is not just random). A sample size of 100 in each group would be more compelling. From statistics we can compare the magnitude of the difference we’ve observed vs the sample size to determine if the result is statistically significant.
Ideally the intervention would be double blind, where neither the scientists nor the people being studied (the subjects) know who is getting the intervention. This prevents the placebo effect, caused by the patient’s belief in the intervention. You can do double-blind with something like a supplement pill: group 1 would get the supplement, group 2 gets a “sugar pill”. You can sometimes also do it for single ingredients, for example by baking them into a cake that is fed to the subjects.
It’s a lot harder with diet or lifestyle changes. If you put group 1 on the Atkins diet and group 2 on Pritikin, that’s pretty hard to fake, everybody knows which diet they’re on.
Interventional studies on animals that have enough anatomical similarities to humans can offer clues but not proof. Here’s an obvious example: If you feed foods with cholesterol to rabbits, they will get heart disease, pretty quickly. Does that prove the same will happen it humans? No, because rabbits are 100% herbivores whose digestive system has no way of handling cholesterol.
I didn’t make this example up, by the way, such experiments really were run on rabbits in the 1950s and people really did suggest it meant it proves the same thing would happen in humans. Nonsense. If you do the same experiments with weasels, you’d probably see no effect at all, because they are carnivores whose digestion handles cholesterol just fine. If you did it with rats, who are omnivores, and saw an effect, that would be more intriguing. But it still wouldn’t be proof. Rats could just be missing something like an important liver enzyme that humans have, that is important for handling cholesterol. You’d need to reproduce the result with humans to prove anything. I’m not saying studies with animals aren’t useful, they are much faster and give important clues. There’s a fascinating example given by Dr. Valter Longo: Some important clues about the connection between diet and longevity were first discovered in yeast. They were then reproduced using mice. They were finally reproduced in humans, proving the validity . The yeast and mice steps, while not in themselves providing proof, were important because they were able to be conducted a lot faster.
Some interventional studies on humans can be done in a relatively short time. You can take a group of people and put them on two different diets and see who loses the most weight in 6 months. Ideally, this would be done in a setting where the scientists controlled the food, and made sure the subjects didn’t have access to snacks. A metabolic ward study where the subjects stay at a facility is the best example, and such studies have been done, but are quite expensive.
Instead, often two groups are given instructions to follow a diet, and then periodically interviewed to ask what they ate. This is obviously a lot lower quality than a ward study because people will vary on how well they comply with the diets (and possibly how honestly they answer the questions), but often this is the type of study done because it is easier and less costly. Some studies are much longer term. If you want to see the effect of an intervention on a chronic disease like heart disease it can take years. If you want to see the effects of an intervention on longevity it takes decades, so that type of study is much more rare.
Probably the next best thing to an intervention study is a longitudinal study, where a significant number of people are followed for a long period of time. Measurements like blood cholesterol, triglycerides, etc. are periodically rechecked, and the people are periodically interviewed about what they are eating, smoking habits, exercise levels, etc. This goes on for many years and you follow their longevity as well as risks of developing various diseases.
An example is the Adventist Health Study . Now you have to do statistical analysis to try to get clues or trends. When you have multiple similar studies you can do statistics on all of them called a meta-analysis. If properly done this is more compelling than the results of any single study.
The statistics needs to be interpreted carefully. For example, you might find that “people who drank more coffee had a higher risk of lung cancer”. Can we conclude coffee causes lung cancer? This is a classic example of “correlation is not causation”. It turns out heavy smokers often have a caffeine habit as well as a nicotine habit. Smoking caused the lung cancer, but coffee drinking accidentally correlated with it.
On the other hand, if coffee had correlated negatively with lung cancer, that would make it unlikely that coffee caused lung cancer. This is because of step 5 of the scientific method: The hypothesis that coffee causes lung cancer does not fit the data.
An example is a hypothesis about legumes, which is currently bandied about quite a bit (try googling “legumes antinutrients”). Legumes contain phytic acid, which is called an “antinutrient” because it can interfere with the body absorbing other nutrients. So we could hypothesize “legumes contain phytic acid, an antinutrient, so legumes will lead to a poorer health”. But there are data from populations around the world showing the opposite. For example, it turns out that elderly people who consume the most legumes live the longest. It is a very strong correlation, more than any other variable looked at . Also, other studies have shown that legumes correlate with reduced disease risk. Our hypothesis that legume consumption leads to poor health does not fit either of these pieces of data, so it is unlikely.
Another important thing to check when looking at a study is who funded the study. In the US, if the funding is from the National Institutes of Health, there are no strings attached. But if the funding is from an industry group, such as the American Beverage Association, the results of the study are more suspect. Statistics have shown that authors of a study are much more likely to come up with results that are favorable to sources of funding like that. This does not mean a lack of integrity on the scientists’ part, the funding source could just introduce an unconscious bias.
It’s a lot of work chasing down the data but it’s worth it. Sometime authors use a lot of scientific terms and may have a scientific degree. And they may be putting forth a very plausible theory. But they still need evidence to back up what they are saying, for it to be scientific.
- Longo, V, The Longevity Diet: Discover the New Science Behind Stem Cell Activation and Regeneration to Slow Aging, Fight Disease, and Optimize Weight, Penguin Group , 2018
- Orlich, M, et al, “Vegetarian Dietary Patterns and Mortality in Adventist Health Study 2”, JAMA Intern Med., 2013
- Darmadi-Blackberry, I, et al, “Legumes: the most important dietary predictor of survival in older people of different ethnicities.”, Asia Pac J Clin Nutr., 2004