The systematic reviews of randomized controlled trials also say that vegetables and fruits (and, generally, minimally processed whole plant foods) are good for reducing all-cause mortality and heart disease. It’s not just the cohort studies. (1, 2)
The large cohort studies are great because they have large sample sizes, look at people living in normal conditions, and look at the most important outcomes (all-cause mortality, heart disease) over long periods of time.
The randomized controlled trials are great because they address confounding, directly. They’re smaller, the controlled feeding studies usually look at people in unusual situations, like living in wards, and they look at shorter periods of time. But they address confounding, which is a huge benefit. Causal inference methods can in principle address confounding too, but that is more difficult and we are not at that level.
In practice, systematic reviews of randomized controlled trials and systematic reviews of large cohort studies reach mostly the same conclusions. I would agree with not believing claims solely based on correlational studies, but that is not necessary since the randomized trials exist.
I do buy that people have various specific nutritional requirements, and that not eating vegetables and fruits means you risk having deficits in various places. The same is true of basically any exclusionary diet chosen for whatever reason, and especially true for e.g. vegans.
In practice, the only thing that seems to be an actual issue is fiber.
Saturated fat is a big issue! It increases your risk of heart disease, and that’s one of the most important causes of death. Nutrition researchers and physicians and cardiologists have been agreeing on this for 20–25 years and there are systematic reviews of controlled trials. Plant foods other than coconut just have a lot less saturated fat than non-plant foods.
Another big issue is the nutrient density (density relative to energy), particularly in underconsumed nutrients. The minimally processed whole plant foods are more dense overall, relative to energy, in the essential vitamins, minerals, omega−3 and omega−6, dietary fiber, relative to the dietary reference intakes. Food composition databases don’t account for bioavailability, but nevertheless. This is especially true for vegetables rather than nuts, fruits, and grain. If you think eggs have high nutrient density, try beating mushrooms, spinach, broccoli—the most nutrient-dense foods are plant foods. This means plant foods allow you to construct a better diet, even if it doesn’t happen automatically.
The NASEM’s adequate intake for fiber is higher than the recommendations of countries outside North America, yet the dose-response curves just show the health benefits for dietary fiber continuing to get higher until the values are high enough that there is no data to know whether they continue going up. Fiber consumption is low even relative to the lower recommendations.
Protein quality doesn’t seem to be an important concern. All plant foods contain all the essential amino acids. Yes, protein quality is lower overall than in typical animal foods, but that matters little because it’s trivial to meet the recommended dietary allowances for all the essential amino acids by eating beans or tofu, and the systematic reviews support plant protein having better health outcomes than animal protein (the systematic reviews tell you what actually matters, and apparently what actually matters is reducing TMAO levels and saturated fat).
The main actual issues with minimally processed whole plant foods are vitamins D and B12. D can technically be managed with UV-irradiated mushrooms, but they are expensive, so supplementing D is the more likely choice. B12 is not solvable. It is present (and bioavailable) in mushrooms but the quantities are small, so either supplementing B12 or adding seafood (likely the least bad animal food) is necessary.
These lines of argument do have the problem that our health decisions should be based primarily on systematic reviews of dietary patterns, rather than individual foods or individual nutrients, and they should be based on health outcomes and especially the most important ones (all-cause mortality, heart disease, cancer).
Government Food Labels Are Often Obvious Nonsense
Yes, you can get more detailed and accurate data using Cronometer, or foodb.ca for the missing nutrients (and way too much information). Thankfully, most minimally processed foods are covered, so the inaccurate FDA nutrition facts can be ignored. Classification systems like Nova and Nutri-Score also don’t seem worthwhile to me due to inaccuracy and oversimplification.
The systematic reviews of randomized controlled trials also say that vegetables and fruits (and, generally, minimally processed whole plant foods) are good for reducing all-cause mortality and heart disease. It’s not just the cohort studies. (1, 2)
The large cohort studies are great because they have large sample sizes, look at people living in normal conditions, and look at the most important outcomes (all-cause mortality, heart disease) over long periods of time.
The randomized controlled trials are great because they address confounding, directly. They’re smaller, the controlled feeding studies usually look at people in unusual situations, like living in wards, and they look at shorter periods of time. But they address confounding, which is a huge benefit. Causal inference methods can in principle address confounding too, but that is more difficult and we are not at that level.
In practice, systematic reviews of randomized controlled trials and systematic reviews of large cohort studies reach mostly the same conclusions. I would agree with not believing claims solely based on correlational studies, but that is not necessary since the randomized trials exist.
Saturated fat is a big issue! It increases your risk of heart disease, and that’s one of the most important causes of death. Nutrition researchers and physicians and cardiologists have been agreeing on this for 20–25 years and there are systematic reviews of controlled trials. Plant foods other than coconut just have a lot less saturated fat than non-plant foods.
Another big issue is the nutrient density (density relative to energy), particularly in underconsumed nutrients. The minimally processed whole plant foods are more dense overall, relative to energy, in the essential vitamins, minerals, omega−3 and omega−6, dietary fiber, relative to the dietary reference intakes. Food composition databases don’t account for bioavailability, but nevertheless. This is especially true for vegetables rather than nuts, fruits, and grain. If you think eggs have high nutrient density, try beating mushrooms, spinach, broccoli—the most nutrient-dense foods are plant foods. This means plant foods allow you to construct a better diet, even if it doesn’t happen automatically.
The NASEM’s adequate intake for fiber is higher than the recommendations of countries outside North America, yet the dose-response curves just show the health benefits for dietary fiber continuing to get higher until the values are high enough that there is no data to know whether they continue going up. Fiber consumption is low even relative to the lower recommendations.
Protein quality doesn’t seem to be an important concern. All plant foods contain all the essential amino acids. Yes, protein quality is lower overall than in typical animal foods, but that matters little because it’s trivial to meet the recommended dietary allowances for all the essential amino acids by eating beans or tofu, and the systematic reviews support plant protein having better health outcomes than animal protein (the systematic reviews tell you what actually matters, and apparently what actually matters is reducing TMAO levels and saturated fat).
The main actual issues with minimally processed whole plant foods are vitamins D and B12. D can technically be managed with UV-irradiated mushrooms, but they are expensive, so supplementing D is the more likely choice. B12 is not solvable. It is present (and bioavailable) in mushrooms but the quantities are small, so either supplementing B12 or adding seafood (likely the least bad animal food) is necessary.
These lines of argument do have the problem that our health decisions should be based primarily on systematic reviews of dietary patterns, rather than individual foods or individual nutrients, and they should be based on health outcomes and especially the most important ones (all-cause mortality, heart disease, cancer).
Yes, you can get more detailed and accurate data using Cronometer, or foodb.ca for the missing nutrients (and way too much information). Thankfully, most minimally processed foods are covered, so the inaccurate FDA nutrition facts can be ignored. Classification systems like Nova and Nutri-Score also don’t seem worthwhile to me due to inaccuracy and oversimplification.