Diet Research: Excellent post by Diego

[Guest guest]

Diego:

Your recent post was excellent. Too often, many scientists and health

professionals do not listen to the anecdotal experiences of their patients

and of people in general. They discount such experiences because it

contradicts what they've been taught or what they've read in research. I

also think it has something to do with ego; many scientists/health

professionals (just like most people) do not want to admit they might be

wrong about something or even acknowledge the possibility of being wrong.

Thus, they will turn a blind eye to people's experiences and keep preaching

what they've always preached.

You are also correct in your comments about research. Results of research

are simply based on probabilities. Researchers are trying to take a sample

of a population and then make some type of inference about that population.

They report the probability of whether the differences they observe in their

samples are due to random sampling variation. That is all. Just because a

difference is not statistically significant doesn't mean there was no

difference. It just means that the probability that the results were due to

random sampling variation is too high to call it significant.

In all research the magical value is 0.05 (which was established from quality

control research...not science as you or I may think of it). So if we get a

value of 0.08 or 0.11, or better yet 0.054, we can't say there was a treatment

effect. But obviously, there very well may have been. It only means now

that there is an 8% chance or an 11% chance or a 5.4% chance that the

difference we observed was due to random sampling variation and not due to

the treatment.

This is why it is critical that researchers report their P

values, so that people can draw their own conclusions about the results of a

study. But how many studies do you read where they report the P value for

their nonsignificant results? Very few. And how many studies report the

confidence intervals, which tell you much more than the P values do (such as

how reliable the conclusions are)? Again, very few.

With a lot of research, you also have the issue of the Type II error, which

is where researchers conclude their was no treatment effect when there

really was. Sample size is one factor that plays a role in this. Freiman

et al. (1978) determined that out of 71 randomized controlled trials which

reported no effect of treatment, 67 of the trials had greater than a 10%

risk of missing a true 25% therapeutic improvement, and 50 of the trials

could have missed a 50% improvement. Many of these studies had inadequate

sample sizes to adequately test their hypotheses.

Again, this is why it is important for researchers to give confidence intervals

so you can see how reliable their data is. How many researchers also report the

statistical power of their study (i.e., the chances of having a type II error)?

Again, very few.

I think it is important for people to be open minded and listen to the

anecdotal experiences of others. If enough people are reporting positive

experiences with ketogenic diets, then maybe some of the scientists and

health professionals are wrong. Maybe these professionals need to take a

closer look at the research on which their popular opinions are based.

Maybe the research is flawed. Almost every study has some type of flaw (if

not numerous flaws) in it. Maybe the way professionals are interpreting the

existing research is flawed.

Much of the dietary recommendations by health

professionals are based off of epidemiological research. But

interpretations from epidemiological research alone is problematic. For

example, the 7-countries study was a famous epidemiological study on which

the recommendation of a high-carb, low-fat diet is based. However,

controlled trials have not been able to support this study. What a lot of

people don't know is that, as one increases their carb intake and reduces

their fat intake at the same time, HDL concentrations decrease, and an

extremely significant inverse correlation between HDL concentrations and

heart disease risk has been established in the literature.

Again, Diego, that was an excellent post and thanks for taking the time to

present your opinion on this issue.

Krieger

Graduate student, exercise science

Washington State University