Epidemiology vs. Controlled Trials

[Jay Guindon]

So ever since I started getting into scientific research I've been wondering why so much of epidemiology is taken as scientific fact and used to make nutrition guidlines when there are so many confounding variables that go unaccounted for compared to controlled trials. As such I generally look at epidemiology as a good starting point to look at correlations and then see if a controlled trial, which has much less chance of confounding, corroborates the epidemiology. This has lead me to look more into epidemiological vs controlled studies and what I have found is that most epidemiologists believe that they can adjust for confounding factors and find cause and effect. For instance, a whole grain study that I cannot find right now, said that after adjusting for age, BMI, and "all other possible confounding factors, people who consumed more whole grains had less instance of cardiovascular disease." This researcher believed he had eliminated all confounding factors and thus whole grains must prevent heart disease. I have since found http://www.ncbi.nlm.nih.gov/pubmed/20307353 and some others that are randomised controlled trials on cardiovascular risk factors and whole grains and they show no benefit to cardio risk factors. I know cardio risk factors and actual disease are different, but I would trust more the controlled study that showed no change in risk factors vs an observation on people who have or do not have heart disease and their whole grain consumption.

Is it ever possible to adjust for confounding factors in an epidemiological study to the point where you can actually find cause and effect or do you always need a controlled trial to be sure?

Are controlled trials always more accurate in establishing cause and effect than epidemological studies?

[Neal Winkler]

First things first, if you are only looking at the abstract then you really haven't found anything. Abstracts are next to worthless.

As to your questions:

(1) A well conducted controlled trial is the ultimate evidence, but you can never be 100% sure when it comes to anything. An epidemiological trial that is very well done cannot quite show cause and effect but it can tell us many things.

(2) No, controlled trials are not always more accurate. A poorly conducted control trial can be worse than a well done prospective epidemiological trial.

[Jay Guindon]

Yes I am aware of the whole abstract thing. It does however give me enough information to decide whether I want to spend money on the article to dig deeper. In a lot of cases they do post some very rudimentary measures, and unless they are misreporting those, it does give you a little to work with when deciding if it's enough to research more thoroughly or to leave it be because I "kind of already knew that."

reduction in inflammatory parameters such as interleukin-6 (pre: 8.08+/-1.57 vs. post: 4.58+/-0.94 pg/mL), interleukin-8 (pre: 45.02+/-5.82 vs. post: 28.59+/-2.64 pg/mL), and tumour necrosis factor-alpha (pre: 53.58+/-25.67 vs. post: 32.09+/-17.42 pg/mL) whereas no significant differences in the placebo period were observed. With regard to haemorheological parameters, the test period significantly ameliorated erythrocytes' filtration rate (pre: 7.61+/-0.71% vs. post: 9.12+/-0.97%; p=0.03) with respect to the placebo period

This is enough info for me to not purchase such a study. While that may be kind of short sighted, I don't have the money to go spending on every study I might find slightly compelling.

Regarding the rest, that was what I thought but it's actually hard to find a thorough review of epidemiological studies vs controlled trials on google. You get a lot of people saying that they are equally good, one guy said that epidemological studies show the same things controlled trials find, etc. Then you get some people saying controlled trials are the only worhtwhile studies. Some people say controlled trials are a waste of time because an epidemiological study will show the same things. Everyone has their opinions of course depending on what method of study they are involved in. I figured I'd get a more balanced opinion here.

[Neal Winkler]

This is enough info for me to not purchase such a study. While that may be kind of short sighted, I don't have the money to go spending on every study I might find slightly compelling.

For sure. This is why science is basically inaccessible to the average person. Right now I'm in grad school so I have access to the schools online database and can order photo copies.

[Jay Guindon]

That's awesome!

[Joshua Naterman]

As far as causation goes, it can ONLY be established in controlled studies, and even then you need a lot of replication with no significant changes in results.

Epidemiology is actually really awesome, because it allows the open minded researcher to more readily identify possible confounds and also to generate really good hypotheses to test! It is also a lot more accessible to the general public, which means it is easier to mislead people with as well, though there seems to be little difficulty in misleading with controlled trials as well!