double blind tests

[Dr. A Sheikh]

when i went to ayurvedic school, the teacher said that double-blind testing is done for 'safety', that you cannot get any valid results unless you use subjects who have something in common.

he said how can fat pills be tested on skinny people? how can those results be valid?

he said double-blind tests don't work.

he said he had participated in them and they are hooey.

he said how can you test a pill for high blood pressure on people who have low blood pressure.

he said double-blind tests are just to test the safety of an item.

please do not respond with put-downs. please give me scientific answers, i need to ponder this.

Do you find it impossible stick with what you say?
Please, do it for us, just this once.

Homeopathy treats symptoms. So for Double blind test are carried out to figure out the efficacy of medicines on disease. We do not claim we treat diseases we treat only symptoms. If you someone could provide us 100 patient of single symptom then we are ready to try double blind tests over them if we fail we will leave homeopathy for ever. We are not against the dbt we are against the steps which are being followed in dbt.

 

[Mojo]

Homeopathy treats symptoms. So for Double blind test are carried out to figure out the efficacy of medicines on disease. We do not claim we treat diseases we treat only symptoms. If you someone could provide us 100 patient of single symptom then we are ready to try double blind tests over them if we fail we will leave homeopathy for ever. We are not against the dbt we are against the steps which are being followed in dbt.

In fact, you have made the following promise:

Proving how? this is our job!

We will prove it with double blind test. if we will fail then we will leave homeopathic practice for ever.

You can only demand to prove what we will claim. Our claim is quoted above.

You don't seem to be making much progress with the test so far. Is your CV up to date?

 

[Luke T.]

http://www.skepticwiki.org/wiki/index.php/Double-Blind

 

[Luke T.]

he said how can fat pills be tested on skinny people? how can those results be valid?

There is one other element of a double blind test which hasn't been mentioned yet.

For instance, you don't want someone who is participating in the study to start the Atkins diet while they are taking the diet pill test. You don't want them doing anything, other than taking that pill, which could be attributed to the success or failure of their weight loss.

ETA: That's also why placebos have inactive ingredients.

 

[Jorghnassen]

I'd just like to point out that double-blind testing doesn't mean testing against a placebo necessarily, especially when there already exists an effective treatment (in which case it can be unethical to do a study of new treatment vs placebo).

 

[kevin]

Actually, it's not really true to say that placebo is the same as doing nothing.

Yeah, I was being a smart a.. for that part.

It would be interesting to know what the placebo effect would be of a substance that the users knew had no pharmacological or physiological effect, and that it was placebo only. That would make a good psychological study.

heh, a double-blind placebo vs. placebo test where the only difference is what the doctor tells you.

 

[Jeff Corey]

An other point. Partcipants must be randomly assigned to the different groups. Otherwise, a threat to internal validity, subject selection, could be a confounding factor. Preexisting differences between the groups could produce effects that could be mistaken for the effect of treatment.

 

[blutoski]

I'd just like to point out that double-blind testing doesn't mean testing against a placebo necessarily, especially when there already exists an effective treatment (in which case it can be unethical to do a study of new treatment vs placebo).

Right, that would be "double blind, placebo, control" study.

Also: an earlier post suggested that placebos contain no ingredients, but this is not entirely true. Part of my job is to design placebos that have 'side effects' to match the experimental dose. This improves blinding, as the patients will not be able to identify their cohort. (aka: confounding)


I think the teacher mentioned at the start of the thread is confusing controlled clinical trials with population studies. In population studies, you look at everybody, but you're not really doing an experiment. In clinical trials, you select subjects who have the condition you want to treat, and you're doing an experiment.

 

[cbish]

I don't know if this has been mentioned but the Dr. doesn't know if a particular patient is getting treatment or placebo. The Dr. doesn't set up the "pills". Someone else does. Records which is which. Gives them to Dr. to be distributed. Compare notes after.

To summarize; Lab tech sets up 50 treatment pills, 50 sugar pills. Records which is which. Gives to DR. Never to be heard from again until the end. DR distributes unknown pills to patients recording what pill number a patient received. Collect data. Compare at the end. Lab tech doesn't know who got what pill. Dr. doesn't know what he gave to who.

 

[patnray]

One thing which has been hinted at, but not stated explicitly, is that the unblinding is not done until after the results are analyzed, so the analyst is not biased. He reports that "Group A had on average an 80% reduction in symptoms compared to Group B". Only then does everyone find out whether Group A got the drug or the placebo.

All quacks have an endless list of excuses why double blind testing does not work for their treatments.

 

[Dr. MAS]

MAS organization is not against the DBT. MAS organization is ready to carry out any experiment by adopting any technique including DBT.

 

[Jeff Corey]

So now we taught you how to do it, go ahead and do it.

 

[DevilsAdvocate]

for example, he said if you are gonna test something about sound, you have to find people who hear.

he says double-blind testing has things that do not fit into it's parameters, and it cannot be used to test everything.

I'm not sure what does not fit into the parameters of a double-blind test other than cases where there a physical limitations of blinding a participant. For example, in testing surgery you can't blind the surgeon from the operation that he or she is performing. But the principal still stands to blind as much as possible.

You teacher is only showing that just because a test is double-blind does not mean that it is a good test. Even double-blind tests can be bad. A bad test protocol, whether double-blind, single-blind, or not blinded, is still a bad test protocol.

But even with a bad test protocol, it is easy to see the advantages of double-blind. Let’s take your example of a test about sound with test subjects that can’t hear:

I have a hypothesis that a certain tone gives people a sense of joy. To test this, test subjects will listen to three tones then select which one gave them the most felling of joy.

For a double-blind test, the experimenter would be “blinded” from the subject (he is outside of the room and so has no effect on the test subject) and the subject is “blinded” from which tone is being played (the subject doesn’t know which tone is the tone that is supposed to elicit joy). If the subjects are all deaf, we will probably get a hodge-podge of data that doesn’t show any significant difference in the chosen tone.

But what happens if it is only single-blinded? In this case, the experimenter is not blinded from knowing what is going on. He is in the room with the test subject and knows which tone is the test tone. This opens the possibility for the experimenter to influence the subject. For example the experimenter may smile when the test tone is played. The subjects may fell that the test tone gives more joy because of the experimenter’s smile.

Similarly, the test subject could not be blinded. In this case, the subject knows which tone is suppose to give joy. The subjects may then select that tone over other, simply because they fell that they should.

In even less controlled environments, there could be no blinding and even encouragement. For example the experimenter plays the two placebo tones and tells the subject what a downer they are, then says, “and here comes the wonderful exciting tone of joy! Hey, wasn’t that nice? Now which one gave you joy?”

So even deaf subjects that can’t hear the tones at all could be influenced to pick (or in other circumstances reject) the test tone because of influences other than the test tone. This would lead to the obviously false conclusion that the tone had an effect—even on deaf people!

So even with a bad test protocol, double-blind is the best test.

Of course double-blind is really only a sort of minimum. Blinding is simply the method for testing. Any test of any type of anything at all is a comparison of two different things—the test and the control. The best test of any thing is to create conditions where only the thing being tested changes. That means eliminating as much as possible all conditions that could influence the test. Double-blind simply means at least eliminating the influence of the subject and experimenter from knowing whether they are the test or the control. The basic premise of removing influences from a test is to have everything blinded and random.

Of course the tests your teacher mention are really NOT invalid. But you have to understand WHAT has been tested. A test of fat pills on skinny people does not prove how the pills work on ALL people. But it does give a conclusion: the pills have such-and-such effect on skinny people. If the pills have no effect, then you have proven something about what was tested: the pills have no effect on skinny people. The test was a very valid test of what it tested. Drawing the conclusion that fat pills have no effect on anyone is just a flat out unreasonable. This would be like concluding that because chocolate is not harmful to you that it is good for you cat or dog.

Tests only test what is tested. Tests can be good or bad. But double-blind tests will ALWAYS be more conclusive than non-double-blind.