Placebo effect

[Dancing David]

Question: Have you seen the study where they give people a sugar pill, alternately tell them that it's a stimulant or a depressant, and then have them rate how they feel on standard forms (and have a nurse take their vital stats)?

This design is immune to most forms of error, but nevertheless patients told they're on a stimulant feel stimulated (and their heartrate etc do go up) and patients told they're on a depressant feel depressed (and ditto in reverse).

Classic placebo effect.

It is best to cite the study, for one then we can see the trail size, what control were used and when the study occured.

This is just anecdotal at this point.

 

[saizai]

It is best to cite the study, for one then we can see the trail size, what control were used and when the study occured.

This is just anecdotal at this point.

I don't remember offhand. It was cited in my cognitive neuroscience course. I can look it up if necessary.

 

[saizai]

http://www.annals.org/cgi/content/full/136/6/471#R55-11 has a pretty good list of studies though

 

[blutoski]

And all you have done is fold it into the placebo effect, and then treat both of those as a baseline. That works for evaluating the effect of a treatment, but means that you can not determine placebo effect vs observer bias, as both are present in the placebo group. It just removes its effect from being solely on those getting treatment, but it does not elliminate it.

So once again, you can not tell if a particular apparent effect on the placebo group was an actual effect or a result of bais. It does not matter which it was for the group recieving the treatment as they need to be comparied to both effects, but saying "there was a real reduction in X in this placebo group" becomes harder to say.

Yes, so you need to have another group in the study that measures baseline beneath placebo. This is best achieved with a non-treatment group. See my earlier posts and examples.

 

[ponderingturtle]

Yes, so you need to have another group in the study that measures baseline beneath placebo. This is best achieved with a non-treatment group. See my earlier posts and examples.

But why would they have the same observer bais as the placebo group?

 

[blutoski]

But why would they have the same observer bais as the placebo group?

I'm not sure I understand what you mean? Do you mean the expectations of the researchers? It'd be double-blinded, so the researchers' bias would be uniform across all participants.

If you mean the participants expectations, I would assume they'd be different, which is the point if you're trying to elicit the effect of the placebo alone.

 

[ponderingturtle]

I'm not sure I understand what you mean? Do you mean the expectations of the researchers? It'd be double-blinded, so the researchers' bias would be uniform across all participants.

If you mean the participants expectations, I would assume they'd be different, which is the point if you're trying to elicit the effect of the placebo alone.

How do you double blind a treatment vs non treatment group to tell the difference between placebo and no treatment?

IF you are measureing a placebo vs treatment group, in that you have both the placebo effect and bais being in both groups, but you want to keep the bais but not the placebo effect from a third group.

You need to get the placebo effect, but you also need to control for observer bais. HOw do you do that with a double blind study? All you do is have two placebo groups

 

[blutoski]

How do you double blind a treatment vs non treatment group to tell the difference between placebo and no treatment?

You can't. Why would you want to? Patient expectations is the proposed method of action. But you *can* blind the experimenter, to maintain control of this factor.

IF you are measureing a placebo vs treatment group, in that you have both the placebo effect and bais being in both groups, but you want to keep the bais but not the placebo effect from a third group.

You need to get the placebo effect, but you also need to control for observer bais. HOw do you do that with a double blind study? All you do is have two placebo groups

I wanted you to be more specific about what you meant by 'control for observer bias'. The theory is that placebo effect works by manipulating the patient's knowledge. You don't want to control for this: it's the variable.

One group gets the placebo; the other group gets nothing. The difference is the placebo. What are you not understanding?

 

[ponderingturtle]

You can't. Why would you want to? Patient expectations is the proposed method of action. But you *can* blind the experimenter, to maintain control of this factor.

BUt you are then not controling for bais in the behalf of the patient, divorced from any real improvement. Also the non treatement patients could unintentionaly cause the observer to get different impressions.

Doing that might reduce bais, but it would still be there.

I wanted you to be more specific about what you meant by 'control for observer bias'. The theory is that placebo effect works by manipulating the patient's knowledge. You don't want to control for this: it's the variable.

Not all of it. The non treatment group would influence the observer differently than the placebo group, in the same way that single blind placebo testing influences the patient's perceptions indirrectly.

One group gets the placebo; the other group gets nothing. The difference is the placebo. What are you not understanding?

How you properly control for just the placebo effect and not all the bais that is normally part of the placebo group and is used to measure the treatment group by.

 

[Cuddles]

BUt you are then not controling for bais in the behalf of the patient, divorced from any real improvement. Also the non treatement patients could unintentionaly cause the observer to get different impressions.

Doing that might reduce bais, but it would still be there.


Not all of it. The non treatment group would influence the observer differently than the placebo group, in the same way that single blind placebo testing influences the patient's perceptions indirrectly.



How you properly control for just the placebo effect and not all the bais that is normally part of the placebo group and is used to measure the treatment group by.

The whole point of a double-blind trial is that the people who assess the outcome have no way of knowing what treatment (or lack of it) was given to which patient. Usually this is done by having an entirely seperate group of people who will administer treatment and who have no contact with those who will evaluate it. If the observer has any way of being influenced by the patient, either directly or indirectly, then the trial is not blinded and is meaningless. If you are worried that the patients might tell the observer that they are not being treated the obvious way around this is to have a questionaire with yes/no answers - "Do you feel better than yesterday?". If you give them no chance to say anything that could give anything away and the observers are never involved with the patients, or any person or place involved, then there can be no bias.

 

[ponderingturtle]

The whole point of a double-blind trial is that the people who assess the outcome have no way of knowing what treatment (or lack of it) was given to which patient. Usually this is done by having an entirely seperate group of people who will administer treatment and who have no contact with those who will evaluate it. If the observer has any way of being influenced by the patient, either directly or indirectly, then the trial is not blinded and is meaningless. If you are worried that the patients might tell the observer that they are not being treated the obvious way around this is to have a questionaire with yes/no answers - "Do you feel better than yesterday?". If you give them no chance to say anything that could give anything away and the observers are never involved with the patients, or any person or place involved, then there can be no bias.

And you can't not do a double-blind trial to test the placebo effect, only at most single blind(the observer, not the patient). A double blind test lumps observer error and placebo together and measures the treatment against it.

As you have to reduce the blinding of the experiment to have a no treatment vs placebo groups you are getting into observer bias issues

 

[Cuddles]

And you can't not do a double-blind trial to test the placebo effect, only at most single blind(the observer, not the patient). A double blind test lumps observer error and placebo together and measures the treatment against it.

As you have to reduce the blinding of the experiment to have a no treatment vs placebo groups you are getting into observer bias issues

No you're not. Single blind means the observer cannot have any bias, otherwise it's not blind. Single blind trials aren't usually used because of patient bias, but in this case that is exactly what you are looking for so it is a perfect setup.

 

[blutoski]

BUt you are then not controling for bais in the behalf of the patient, divorced from any real improvement. Also the non treatement patients could unintentionaly cause the observer to get different impressions.

Doing that might reduce bais, but it would still be there.


Not all of it. The non treatment group would influence the observer differently than the placebo group, in the same way that single blind placebo testing influences the patient's perceptions indirrectly.



How you properly control for just the placebo effect and not all the bais that is normally part of the placebo group and is used to measure the treatment group by.

I sincerely do not understand what you're getting at. I think we're being too vague. Let's say I'm trying to test the effect of an antiretroviral drug.

I divide it into three groups:

A: no treatment
B: placebo tablets
C: experimental tablets

The tablets are identical in size, shape, flavour, &c, and the side effects of the drug are mimicked by the placebo. Let's say nausea.

The viral load is measured by having the hospital bloodwork team (aka: the vampires) take daily samples, and send them to the lab in coded vials. The lab records the results.

At the end of the experiment, the code is broken, and the results analyzed.

I really don't see any bias complications.

 

[ponderingturtle]

No you're not. Single blind means the observer cannot have any bias, otherwise it's not blind. Single blind trials aren't usually used because of patient bias, but in this case that is exactly what you are looking for so it is a perfect setup.

So then homeopathic medicine does work on animals as those where single blinded experiments(the animals did not know if they where getting the homeopathic treatment or not) so there was no observer bias to account for their behavior.

 

[ponderingturtle]

I sincerely do not understand what you're getting at. I think we're being too vague. Let's say I'm trying to test the effect of an antiretroviral drug.

I divide it into three groups:

A: no treatment
B: placebo tablets
C: experimental tablets

The tablets are identical in size, shape, flavour, &c, and the side effects of the drug are mimicked by the placebo. Let's say nausea.

The viral load is measured by having the hospital bloodwork team (aka: the vampires) take daily samples, and send them to the lab in coded vials. The lab records the results.

At the end of the experiment, the code is broken, and the results analyzed.

I really don't see any bias complications.

Becuase you have chosen a set up that is makeing quantitative measurements, and not "do you feel better" questions.

 

[Groovydoc]

Sorry if this sounds cynical, but....

I realize this thread is about using the placebo effect, which is quite a fascinating topic, but I can tell you as a practicing physician, as a practical standpoint, we have a hard enough time gettting people to do things we KNOW are going to be helpful. Just look around at all the people who continue to drink, smoke, do meth, weigh 300+ pounds, etc. Thats half the battle. Then, after 40-60 years of that, their multiple medical problems have so many treatments PROVEN to help but half the people don't want to do them or have trouble remembering to follow through. Then, as I often see, there are those on 15+ meds. Intentionally adding a placebo to this mix generally holds no interest to me, nor to patients for the most part.

 

[Dancing David]

So then homeopathic medicine does work on animals as those where single blinded experiments(the animals did not know if they where getting the homeopathic treatment or not) so there was no observer bias to account for their behavior.

Whoa, only if there is a statisticaly significant variation vs. untreated animals in a large pool of trials. And the experimenter has to do everything the same for both groups of animals, except one gets the dose and the others don't.

All the homeopathy 'results' is usualy based upon single trial anecdotes and is therefore not a controlled study.

 

[Dancing David]

Becuase you have chosen a set up that is makeing quantitative measurements, and not "do you feel better" questions.

So you use a post, pretest survey. But you have to make sure the nontreatment group goes through the same protocolo, ie sitting in the waiting area and talking to the nurse, they just don't get a 'treatment'.

 

[saizai]

Whoa, only if there is a statisticaly significant variation vs. untreated animals in a large pool of trials. And the experimenter has to do everything the same for both groups of animals, except one gets the dose and the others don't.

So, homeopathic treatment vs. regular deionized water then?


doc - It's not "adding a placebo"; that implies a placebo-only treatment of some kind, e.g. a sugar pill. It's increasing the placebo *effect*, aka context effects, to improve the effectiveness of all treatments.

 

[blutoski]

Becuase you have chosen a set up that is makeing quantitative measurements, and not "do you feel better" questions.

Sure. That's why I figured it would be valuable to get a specific example, instead of continuing to wrestle with vagueness.

re: spectral questions: try distributing questionnaires instead of face-to-face interviews.