The Atkins Diet and the scientific method II

How would I design a trial to test whether or not the Atkins Diet 'works'? The following is the answer to that question, with the caveat that it may not be ?practical? or ?feasble?; it is simply the standard of proof that I require.

Before I go any farther, I have to decide what 'works' means. Most people agree that eating fewer calories than your body expends, a negative calorie balance, results in weight loss. Atkins supporters make the additional claim that a low carbohydrate negative calorie balance results in greater weight loss than does a low fat (conventional) negative calorie balance.

Thus, by asking "Does the Atkins Diet 'work'?" I am asking, "Does the Atkins Diet result in greater weight loss than a conventional diet?"

The independent variable is low carbohydrate composition of the diet calories.

As empirical science necessitates a control to be able to draw any conclusions from the data, it is important that I have a true control in the study. Remember that the only difference between the test group and the control group should be the independent variable. Therefore, the only difference between the two groups can be the low carbohydrate composition of the diet calories.

Two groups - 200 people in each group.[1] Each person has a Body Mass Index (BMI) of 30 or more. All 400 subjects stay for 18 months in a study "compound" where they are closely monitored for diet and exercise.[2] All of their meals will be provided for them. Each subject will have to walk/run 2 miles a day. In other words, they cannot eat something without me knowing it, and they cannot exercise without me knowing it.

Group 1: the Atkins group - 200 subjects (100 male, 100 female). Each subject gets a daily total caloric intake based on a BMI of 24 (upper end of normal range). Those calories are low carbohydrate, high fat, and high protein by composition.[3]

Group 2: the Control group - 200 subjects (100 male, 100 female). Each subject gets a daily total caloric intake based on a BMI of 24 (upper end of normal range). Those calories are high carbohydrate, low fat, and high protein by composition.

Every day, the weight of each subject gets taken, up till the end of the study in 18 months. At the end of the study, I analyze the results to see if there are any statistically significant differences between the average weight loss of the subjects in the two groups at various times during the 18 month period.

That is my design for the experiment. Your job is to find any and all flaws with it.

-----------------------------------------------------------------------------------------

[1] Why so many people in each group? To minimize the random differences between the groups that could result in less than a true control. Consider the extreme case in which only 5 people are assigned to each group. There is a much greater chance that the differences between the two groups are not limited to the independent variable. Perhaps in one group 4 of the subjects just happened to have diabetes whereas in the other group only 1 subject has diabetes. The control is gone. Or consider the even more extreme example in which each group has only 1 subject. Nobody could truthfully claim that a control exists in that situation. With increasing numbers of subjects, it is statistically more improbable that random differences between the subjects are significant enough to result in a lack of a control.

[2] Again, I realize that it may not be ?practical? to have people take time off from work and isolate them in compound, but this post is about the standard of proof, not what is practical.

[3] I have decided against blinding the study subjects because I can keep a close eye on them in the compound. There will be little chance of them cheating on their diets or exercise. Thus, their food will look as it normally does, i.e., steak will look like steak.

Share this

A study has already been

A study has already been done. You can find it at:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=12761365&dopt=Abstract

Oblomov, Thanks. Actually,

Oblomov,
Thanks. Actually, multiple studies have been done, of varying degrees of merit with how well they performed the scientific method. My two posts on the Atkins Diet were made not as a search for the efficacy of the Atikins Diet, but rather to bring about discussion on the scientific method itself. I'm not primarily interested in whether or not the Atkins Diet 'works', but rather how one would go about finding out if it works.

BTW, I don't see a problem

BTW,

I don't see a problem with your study design, except to say that BMI may not be the best measure of obesity, since BMI can increase if the proportion of muscle tissue increases. Maybe the right way to do this is to find the total fat weight by caliper measurement or water displacement.

"Does the Atkins Diet result

"Does the Atkins Diet result in greater weight loss than a conventional diet?"

There may be other factors you want to consider, such as the enjoyment of the diet itself. Even if a conventional diet results in greater weight loss than Atkins, if Atkins is more enjoyable, people are more likely to stick with the diet rather than cheat, which would make Atkins a better diet overall even if it is theoretically less effective than a conventional diet.

In other words, your standard might be too strict if we wish to determine whether Atkins is a better diet overall.

Also, we might want to consider other health effects - "externalities," the economist in me says - when determining which diet is better overall. If Atkins results in hair loss, bad breath, and is difficult to sustain over a long period of time, its benefits may be overstated.

Micha, I don't think the

Micha,

I don't think the enjoyment of the diet affects the study itself, which simply seeks to determine whether or not the composition of diet calories effects weight loss.

The enjoyment of the diet does not play a part in the study because, in my "compound" the subjects don't get a choice in what they eat; they eat what I feed them. It's a standard collectivist setup for the greater good of society.

The conclusions I take from the results are when the enjoyment of the diet comes into play. If I conclude, "People will definitely lose weight faster on the Atkins diet" I would probably be non-rigorous as they might not enjoy the diet.

Also, we might want to consider other health effects - "externalities," the economist in me says - when determining which diet is better overall. If Atkins results in hair loss, bad breath, and is difficult to sustain over a long period of time, its benefits may be overstated.

Truedat. I'm sure that other than weight loss itself, I could also measure other outcomes - bodyfat composition, kidney function, blood pressure, hair loss, subjective symptoms, etc.

But I don't think this effects whether or not I can conclude that the diet 'works', i.e., whether or not the composition of the calories effects weight loss when on a caloric deficit diet.

I don't see a problem with

I don't see a problem with your study design, except to say that BMI may not be the best measure of obesity, since BMI can increase if the proportion of muscle tissue increases. Maybe the right way to do this is to find the total fat weight by caliper measurement or water displacement.

Good point. I actually struggled in figuring out which parameter to use for inclusion in the study. Should I use people who are 10% overweight? If so, how do I figure out what their normal weight is? Should I use only extremely obese people? Should I only use people over 200 lbs? In the end, I decided to use BMI. It's not perfect, but it's standardized for height. With enough people in the study, the 'outliers' that are exceptions to the BMI standards will not play a major role.

Perhaps using bodyfat composition would be better. But then I think the outcome measurement should also be bodyfat composition, not weight loss.

Each subject will have to

Each subject will have to walk/run 2 miles a day. In other words, they cannot eat something without me knowing it, and they cannot exercise without me knowing it.

When keeping these patients in isolation, how do you prevent them from doing any other activities besides your 2 mile walk/run? What if some of the subjects do jumping jacks on their own but they don't do it regularly? How do you account for that?

When keeping these patients

When keeping these patients in isolation, how do you prevent them from doing any other activities besides your 2 mile walk/run? What if some of the subjects do jumping jacks on their own but they don't do it regularly? How do you account for that?

I guess it depends on how tyrannical I want to be. If I wanted really strict control, I would have to put a camera on every part of the "compound" including the subjects' private quarters. Anyone who does jumping jacks gets punished...

...and trust me, nobody wants to get punished. *cold stare*

When keeping these patients

When keeping these patients in isolation, how do you prevent them from doing any other activities besides your 2 mile walk/run? What if some of the subjects do jumping jacks on their own but they don't do it regularly? How do you account for that?

That is the purpose of getting large numbers of people to participate. Statistical analysis can then be used to "control" for uncontrollable variables.

However, if the point is

However, if the point is that people are more willing to stay on the Atkins diet and are therefore more likely to lose weight/keep it off, your study doesn't prove anything at all.

In not blinding the study

In not blinding the study subjects, you open the study to the placebo effect. That is, if the subjects in general believe one diet will work better than the other, it may have an effect on the results - who knows if there is a psychological component to dieting.

In not blinding the study

In not blinding the study subjects, you open the study to the placebo effect. That is, if the subjects in general believe one diet will work better than the other, it may have an effect on the results - who knows if there is a psychological component to dieting.

Yes I am an Atkins diet

Yes I am an Atkins diet survivor. I have currently loss 30 kilos in the last 9 months. The first 3 months I was on the Atkins Diet but I couldn't sustain it. Although I lost 10 kilos I just couldn;t sustain it.

The good thing the Atkins did though is gave me the confident that losing weight is possible. Now I'm on more a balanced diet and lots of exercise. Its working great