4 ms·

I'm just gonna say it: epidemiological studies suck and we need something better. Humans and their lives are just too multi-variate, even with nearly ~500k peo

I'm just gonna say it: epidemiological studies suck and we need something better. Humans and their lives are just too multi-variate, even with nearly ~500k people in a study. All you end up with are correlations – I've basically never seen a study like this that manages to demonstrate causality convincingly.

nextos 6y ago
To infer causality you need interventions. Although there are some methods to infer causal relationships from observational data (e.g. see Pearl or Spirtes' work, which is ignored by most epidemiologists). I think the biggest problem is that they are not measuring low level molecular variables like metabolites, or abundances of gut bacteria. Hence, it's pretty hard to infer a causal chain. And since controlling for covariates is so hard, inferring something useful is difficult.
- hntrader 6y ago
  > controlling for covariates is so hard Especially when the purported effects are small/clinically insignificant.
diydsp 6y ago
The gold standard to establish causation is the 2x blind random placebo-controlled study. There are many of them assessing various types of foods, from meat to lychee to soy and everything in between. Many biomarkers in the body can be studied such as stress hormones, markers of inflammation, lipid panels, arterial wall thickness, etc. If you poke around on youtube, you can find smart people presenting this research! Most of them are health influencers. Personally, I follow Drs. Gregor, Rhonda Patrick, Fung, Mike Hetzel Thoms Delauer and the rest of this particular cohort. They all have their separate biases and areas of interest, such as fasting, muscle gain, veganism, grass-fed beef, ketosis, etc, but if you find the intersection of their work, you'll have an excellent idea what is good and bad for you and how much you can get away with! To distill two of the most valuable things I've learned: 1. The Mediterranean diet seems to be the healthiest and closest to the Western and American diets, although its actual form has some key differences from the received wisdom. In particular, its pasta is a whole-grain variety and the fish consumption is less than once per day. 2. Most people are not fat-adapted and their bodies do a poor job of metabolizing fat! Typically this requires fasting or other forms of intermittent ketosis to get the body producing fat-burning enzymes. This is likely the source of a lot of conflicting info on fat. It's perfectly excellent for you - if you are fat-adapted, you will use it for energy and won't experience 3pm brainfog. But if you're not, it will very quickly create adipose tissue (body fat!) that continuously emits cytokines (poison) that sabotage your health. 3. Stress alters our metabolism for the worse! Since so many of us are stressed we're not performing proper digestion and not getting the right benefits from what we eat! Breathing exercises, etc. can help improve this.
- Bakary 6y ago
  The Mediterranean diet is only superficially close to the American one. It would be more accurate to speak of separate lifestyles than diets. The relationship to food that Americans have is almost completely alien to what most Italians experience, although with the spread of American culture this might not last beyond a few generations. It's a bit like what used to be called the 'French paradox' in that American observers mistook a tradition of rich foods as the constant ingestion of such food.
- fastball 6y ago
  Double-blind random placebo-controlled doesn't automatically demonstrate causation either. Humans are just too multi-variate.
  - bigfudge 6y ago
    It would if people could stick to the protocols in the studies as designed. In practice RCTs only enable us to estimate the effect of the _offer_ of treatment (the intention to treat estimate) rather than making a direct causal estimate of the treatment itself. Missing data is a particular source of bias in this regard (if people dropped out, you can't know why or the counterfactual of what you would have observed had they stayed in the trial).
argvargc 6y ago
John Robbins (would-be heir to the Baskin & Robbins empire) wrote a book examining the social, exercise and diet cultures of four of the known longest-living cohesive groups of people in the world. These were/are isolated or otherwise somehow physically-bounded societies who regularly have members living healthy lives at over 100 years of age (for example, no cancer, no degraded eyesight etc), and in a much higher proportion than is normally expected. He was curious as to whether these groups of people, known to live a long time and even be healthy in their old age, had anything in common. And they did. They all exercised a lot. They all had very strong social and community connections. The all ate a mostly-vegan diet, punctuated with small amounts of animal product (in quantities varying from a portion every few weeks, to a few times a week but for only a couple of months of the year - for seasonal reasons). He attempted to also prove, that taking any one of the above three commonalities away, was enough to invalid the results. It's an interesting book, for sure. Unfortunately, it has the terrible, baity title of: "Healthy At 100", and an even worse sub-title I won't repeat. But other than that mis-step (possibly at the hands of a publishing company), it's a good, well-referenced read.

nextos 6y ago

To infer causality you need interventions. Although there are some methods to infer causal relationships from observational data (e.g. see Pearl or Spirtes' work, which is ignored by most epidemiologists). I think the biggest problem is that they are not measuring low level molecular variables like metabolites, or abundances of gut bacteria. Hence, it's pretty hard to infer a causal chain. And since controlling for covariates is so hard, inferring something useful is difficult.

hntrader 6y ago
> controlling for covariates is so hard Especially when the purported effects are small/clinically insignificant.

diydsp 6y ago

The gold standard to establish causation is the 2x blind random placebo-controlled study. There are many of them assessing various types of foods, from meat to lychee to soy and everything in between. Many biomarkers in the body can be studied such as stress hormones, markers of inflammation, lipid panels, arterial wall thickness, etc. If you poke around on youtube, you can find smart people presenting this research! Most of them are health influencers. Personally, I follow Drs. Gregor, Rhonda Patrick, Fung, Mike Hetzel Thoms Delauer and the rest of this particular cohort. They all have their separate biases and areas of interest, such as fasting, muscle gain, veganism, grass-fed beef, ketosis, etc, but if you find the intersection of their work, you'll have an excellent idea what is good and bad for you and how much you can get away with! To distill two of the most valuable things I've learned: 1. The Mediterranean diet seems to be the healthiest and closest to the Western and American diets, although its actual form has some key differences from the received wisdom. In particular, its pasta is a whole-grain variety and the fish consumption is less than once per day. 2. Most people are not fat-adapted and their bodies do a poor job of metabolizing fat! Typically this requires fasting or other forms of intermittent ketosis to get the body producing fat-burning enzymes. This is likely the source of a lot of conflicting info on fat. It's perfectly excellent for you - if you are fat-adapted, you will use it for energy and won't experience 3pm brainfog. But if you're not, it will very quickly create adipose tissue (body fat!) that continuously emits cytokines (poison) that sabotage your health. 3. Stress alters our metabolism for the worse! Since so many of us are stressed we're not performing proper digestion and not getting the right benefits from what we eat! Breathing exercises, etc. can help improve this.

Bakary 6y ago
The Mediterranean diet is only superficially close to the American one. It would be more accurate to speak of separate lifestyles than diets. The relationship to food that Americans have is almost completely alien to what most Italians experience, although with the spread of American culture this might not last beyond a few generations. It's a bit like what used to be called the 'French paradox' in that American observers mistook a tradition of rich foods as the constant ingestion of such food.
fastball 6y ago
Double-blind random placebo-controlled doesn't automatically demonstrate causation either. Humans are just too multi-variate.
- bigfudge 6y ago
  It would if people could stick to the protocols in the studies as designed. In practice RCTs only enable us to estimate the effect of the _offer_ of treatment (the intention to treat estimate) rather than making a direct causal estimate of the treatment itself. Missing data is a particular source of bias in this regard (if people dropped out, you can't know why or the counterfactual of what you would have observed had they stayed in the trial).

argvargc 6y ago

John Robbins (would-be heir to the Baskin & Robbins empire) wrote a book examining the social, exercise and diet cultures of four of the known longest-living cohesive groups of people in the world. These were/are isolated or otherwise somehow physically-bounded societies who regularly have members living healthy lives at over 100 years of age (for example, no cancer, no degraded eyesight etc), and in a much higher proportion than is normally expected. He was curious as to whether these groups of people, known to live a long time and even be healthy in their old age, had anything in common. And they did. They all exercised a lot. They all had very strong social and community connections. The all ate a mostly-vegan diet, punctuated with small amounts of animal product (in quantities varying from a portion every few weeks, to a few times a week but for only a couple of months of the year - for seasonal reasons). He attempted to also prove, that taking any one of the above three commonalities away, was enough to invalid the results. It's an interesting book, for sure. Unfortunately, it has the terrible, baity title of: "Healthy At 100", and an even worse sub-title I won't repeat. But other than that mis-step (possibly at the hands of a publishing company), it's a good, well-referenced read.