4 ms·

On the other hand, as I explain above, epidemiological studies are essentially useless for anything other than formulating hypotheses (see the debacle with horm

by zngtk4 17y ago

On the other hand, as I explain above, epidemiological studies are essentially useless for anything other than formulating hypotheses (see the debacle with hormone replacement therapy). It baffles me that anyone would entirely change their diet and lifestyle based on an epidemiological study.

Tichy 17y ago
So how could they overlook the effects of the hormone treatment if they studied so many women? Shouldn't there be some statistical observations?
- zngtk4 17y ago
  Absolutely, and the observations suggested that hormone treatment was possibly even beneficial. That's why it's completely true that correlation != causation. You can try to "control" for differences in the observed groups, and epidemiologists try this all the time, but unless they already understand everything involved, it's impossible to get this completely right. And sometimes the things they try to control for are themselves based on other epidemiological studies. This is the main reason we are always hearing in the press about how a "study shows that X is bad for us" then a "study shows that X is good for us". The majority of the time when there are such conflicting studies, it is the result of epidemiology or poorly controlled studies.
  - Tichy 17y ago
    So there was some other factor that by chance made the hormones benefitial for the women in the study? Or all the women somehow where more healthy than "normal" women to begin with? How does that happen? And how could one control for it, wouldn't you need thousands of identical twins?
grandalf 17y ago
The evidence is not just from the epidemiological study, there have been lots of follow-up studies to examine the processes in detail. I would argue that a broad, epidemiological study is the most persuasive sort of evidence to use when considering whether to make drastic changes. Otherwise you're dealing with information like "omega 3s prevent x, y, and z" so you end up taking an Omega 3 supplement along with the dozen other supplements while eating Wendy's for lunch every day. Thinking that health can be understood on the basis of "taking supplement x is good" is the root of the current supplement gold rush (people sell tons of snake oil substances that contain "omega 3's" etc.) and a highly irrational way to behave.
- zngtk4 17y ago
  >follow-up studies to examine the processes in detail. Examining the processes isn't sufficient to provide proof unless it's a completely thorough examination. Biological science isn't at that point yet, and so what this usually means is that group A performs an observational study, group B finds a process that would explain group A's results, assuming group A's results are true. That's different from showing that group A's results were true. Human, and animal, physiology is very complicated, and it's just plain wrong to assert that we know that eating meat decreases longevity. We don't. There is some evidence to believe it, and some evidence to not believe it at this point (both epidemiological studies and at the molecular understanding of the processes involved). Epidemiological may be persuasive, but it's misleading because you take a single variable out of hundreds that may not even be known. So you may get Texans who become vegetarians based on longevity studies in some region of China, but then live an even shorter life because their lifestyles and the particular vegetarian foods that they eat aren't as conducive to longevity. This history of epidemiological studies are full of cases like this.
  - grandalf 17y ago
    I don't dispute your logical points, but I think there is still a rather strong case for eating a plant-based diet. Clearly, humans can eat meat (during most of human evolution calories were scarce and eating meat conferred a selection advantage)... But why would you reason that all foods are equally beneficial and harmful? To me, broad epidemiological studies can offer clues on classes of foods and some of the costs/benefits that they confer.
    - zngtk4 17y ago
      I'm not reasoning that all foods are equally beneficial and harmful, and I accept that it may very well be that a plant-based diet is superior to a meat-based diet. I reject the idea, however, that we know enough to say that this is true, and I absolutely reject the idea of significantly changing behavior based solely on epidemiological studies -- the epidemiological studies should be the starting point for much further study. To date, most large-scale controlled studies on diet have unfortunately been shelved. It should entirely be possible to separate multiple groups on multiple different diets and see what happens over a period of, say, ten years -- one diet as vegetarian, one diet as meat/fruit/nut but no grains, a fish+vegetarian diet, while holding other variables constant (such as calories). This is what we need to do to know, and this is what has consistently been proposed, but not done.

by zngtk4 17y ago

Tichy 17y ago

So how could they overlook the effects of the hormone treatment if they studied so many women? Shouldn't there be some statistical observations?

zngtk4 17y ago
Absolutely, and the observations suggested that hormone treatment was possibly even beneficial. That's why it's completely true that correlation != causation. You can try to "control" for differences in the observed groups, and epidemiologists try this all the time, but unless they already understand everything involved, it's impossible to get this completely right. And sometimes the things they try to control for are themselves based on other epidemiological studies. This is the main reason we are always hearing in the press about how a "study shows that X is bad for us" then a "study shows that X is good for us". The majority of the time when there are such conflicting studies, it is the result of epidemiology or poorly controlled studies.
- Tichy 17y ago
  So there was some other factor that by chance made the hormones benefitial for the women in the study? Or all the women somehow where more healthy than "normal" women to begin with? How does that happen? And how could one control for it, wouldn't you need thousands of identical twins?

grandalf 17y ago

The evidence is not just from the epidemiological study, there have been lots of follow-up studies to examine the processes in detail. I would argue that a broad, epidemiological study is the most persuasive sort of evidence to use when considering whether to make drastic changes. Otherwise you're dealing with information like "omega 3s prevent x, y, and z" so you end up taking an Omega 3 supplement along with the dozen other supplements while eating Wendy's for lunch every day. Thinking that health can be understood on the basis of "taking supplement x is good" is the root of the current supplement gold rush (people sell tons of snake oil substances that contain "omega 3's" etc.) and a highly irrational way to behave.

zngtk4 17y ago
>follow-up studies to examine the processes in detail. Examining the processes isn't sufficient to provide proof unless it's a completely thorough examination. Biological science isn't at that point yet, and so what this usually means is that group A performs an observational study, group B finds a process that would explain group A's results, assuming group A's results are true. That's different from showing that group A's results were true. Human, and animal, physiology is very complicated, and it's just plain wrong to assert that we know that eating meat decreases longevity. We don't. There is some evidence to believe it, and some evidence to not believe it at this point (both epidemiological studies and at the molecular understanding of the processes involved). Epidemiological may be persuasive, but it's misleading because you take a single variable out of hundreds that may not even be known. So you may get Texans who become vegetarians based on longevity studies in some region of China, but then live an even shorter life because their lifestyles and the particular vegetarian foods that they eat aren't as conducive to longevity. This history of epidemiological studies are full of cases like this.
- grandalf 17y ago
  I don't dispute your logical points, but I think there is still a rather strong case for eating a plant-based diet. Clearly, humans can eat meat (during most of human evolution calories were scarce and eating meat conferred a selection advantage)... But why would you reason that all foods are equally beneficial and harmful? To me, broad epidemiological studies can offer clues on classes of foods and some of the costs/benefits that they confer.
  - zngtk4 17y ago
    I'm not reasoning that all foods are equally beneficial and harmful, and I accept that it may very well be that a plant-based diet is superior to a meat-based diet. I reject the idea, however, that we know enough to say that this is true, and I absolutely reject the idea of significantly changing behavior based solely on epidemiological studies -- the epidemiological studies should be the starting point for much further study. To date, most large-scale controlled studies on diet have unfortunately been shelved. It should entirely be possible to separate multiple groups on multiple different diets and see what happens over a period of, say, ten years -- one diet as vegetarian, one diet as meat/fruit/nut but no grains, a fish+vegetarian diet, while holding other variables constant (such as calories). This is what we need to do to know, and this is what has consistently been proposed, but not done.