Acupuncture appears to show results.

[BillyJoe]

Are you saying that having a study arm which identifies the natural course of an illness with no intervention is not of much value?

Not when you're doing a trial of acupuncture.

For that, all you need is two groups: a True Acupuncture group, and a Sham Acupuncture group.
All your control does is to reduce the numbers in the above two groups, thereby reducing the strength of the trial of acupuncture.

Your control group is great if you are doing a trial of the placebo effect, but it has no place in a trial of acupuncture.

But I think we have been through this at least three times now.

regards,
BillyJoe

 

[Skeptic Ginger]

The problem comes specifically when your acupuncture and sham arms have a similar outcome and you conclude both treatments work for the reasons placebos work when what might actually be occurring is neither had a placebo effect.

For many studies the outcome of, "no better than placebo" is sufficient to discount the treatment. For many of the CAM studies including acupuncture studies we are seeing claims that the fact something works by some placebo mechanism is still evidence the treatment is of value. That is why this is even an issue.

 

[Ivor the Engineer]

When doctors go on strike, the death rate goes down.

It's important to find out what happens when you do nothing.

 

[fls]

The problem comes specifically when your acupuncture and sham arms have a similar outcome and you conclude both treatments work for the reasons placebos work when what might actually be occurring is neither had a placebo effect.

When you say "placebo effect", what exactly are you talking about?

For many studies the outcome of, "no better than placebo" is sufficient to discount the treatment. For many of the CAM studies including acupuncture studies we are seeing claims that the fact something works by some placebo mechanism is still evidence the treatment is of value. That is why this is even an issue.

What did the untreated control group in the acupuncture study tell us, and how does it make this less of an issue?

Linda

 

[Ivor the Engineer]

<snip>

Your control group is great if you are doing a trial of the placebo effect, but it has no place in a trial of acupuncture.

<snip>

Yes it does. It could be that sticking needles in people is worse than doing nothing to them.* Having a baseline is important.

*I generally find this to be the case in my day to day life.

 

[BillyJoe]

Igor,

Do not just read the headlines....

When doctors go on strike, the death rate goes down.

Firstly, they did not go on strike, they merely declined to perform some of their usual activities:

...hundreds of thousands of visits to outpatient clinics have been cancelled or postponed along with tens of thousands of elective operations.

One effect of postponing tens of thousands of elective operations is to postpone the deaths due to elective surgery:

Elective surgery can bring about a great improvement in a patient's condition, but it can also mean disability and death in the weakest patients. And patients who do not undergo diagnosis or surgery now could decline or die in a few months due to the postponement."

Secondly, many activities were continued:

Public hospitals....have kept their emergency rooms, dialysis units, oncology departments, obstetric and neonatal departments, and other vital facilities working normally during the industrial action.

If the doctors had really gone on strike, imagine the resulting death toll.

regards,
BillyJoe

 

[fls]

Yes it does. It could be that sticking needles in people is worse than doing nothing to them.* Having a baseline is important.

I think you could make a case for that if acupuncture was a novel treatment. But the baseline for lack of harm is so well-established that the point is moot.

*I generally find this to be the case in my day to day life.

It probably doesn't help that I just read a footnote about Kinsey sticking a toothbrush up his urethra, but the mental image I just got was not very (shall we say) 'nice'.

Linda

 

[BillyJoe]

Yes it does. It could be that sticking needles in people is worse than doing nothing to them.* Having a baseline is important.

*I generally find this to be the case in my day to day life.

It has already been established that "sticking needles* in people" has a positive effect.

What we are interested in is comparing acupuncture (active group) to "sticking needles in people" (control group)


*of the type used in acupuncture

 

[Ivor the Engineer]

Linda,

when you say "the baseline for lack of harm is so well-established", do you mean that at the end of trial the acupuncture and needling groups have not got worse relative to the start of the trial, or that at the end of the trial the acupuncture and needling groups have at least kept up with a no-treatment group, allowing for the possibility a no-treatment group may improve over the duration of a trial?

For example, if at the end of a trial the acupuncture group had improved 10%, the needling group 5% and the no-treatment group 15%, is acupuncture/needling considered harmful? What if there were no no-treatment group?

 

[fls]

Linda,

when you say "the baseline for lack of harm is so well-established", do you mean that at the end of trial the acupuncture and needling groups have not got worse relative to the start of the trial, or that at the end of the trial the acupuncture and needling groups have at least kept up with a no-treatment group, allowing for the possibility a no-treatment group may improve over the duration of a trial?

I mean the thousands of studies on acupuncture with explicit and implicit comparisons to no treatment, and the information we have from general use.

For example, if at the end of a trial the acupuncture group had improved 10%, the needling group 5% and the no-treatment group 15%, is acupuncture/needling considered harmful? What if there were no no-treatment group?

It would be an unusual result. It's hard to draw reliable conclusions from one unusual result (especially when harm isn't the hypothesis being tested). In this setting, I don't think it would even be enough to draw the 'needs further study' recommendation.

Linda

 

[Ivor the Engineer]

So is it reasonable to assume, with the illnesses/conditions typically treated with acupuncture, a no-treatment group would show no change from the start of the trial to the end?

Or do trials without no-treatment groups use some kind of typical (i.e. possibly non-horizontal) baseline, derived from prior knowledge of the course of the specific illness/condition, when reporting the effects of treatment?

 

[fls]

So is it reasonable to assume, with the illnesses/conditions typically treated with acupuncture, a no-treatment group would show no change from the start of the trial to the end?

Or do trials without no-treatment groups use some kind of typical (i.e. possibly non-horizontal) baseline, derived from prior knowledge of the course of the specific illness/condition, when reporting the effects of treatment?

I don't think it's at all reasonable to assume that. I think it's reasonable to assume that various groups (if randomized (and sufficiently homogenous and large)) are likely to change in a similar manner in the absence of treatment.

Trials without no-treatment groups report on the effects of treatment compared to placebo.

Linda

 

[Ivor the Engineer]

I don't think it's at all reasonable to assume that. I think it's reasonable to assume that various groups (if randomized (and sufficiently homogenous and large)) are likely to change in a similar manner in the absence of treatment.

I agree.

Trials without no-treatment groups report on the effects of treatment compared to placebo.

Which is useful to potential users if the treatment the placebo group received can be roughly equated with doing nothing (e.g., sugar pills). But when the placebo group receives a procedure which is elaborate and far from trivial, it seems more risky to infer the likely benefit or harm of a treatment (compared to no-treatment) from the results of a study unless there was a no-treatment group.

 

[fls]

Which is useful to potential users if the treatment the placebo group received can be roughly equated with doing nothing (e.g., sugar pills). But when the placebo group receives a procedure which is elaborate and far from trivial, it seems more risky to infer the likely benefit or harm of a treatment (compared to no-treatment) from the results of a study unless there was a no-treatment group.

Why would you want to infer the likely benefit or harm of a treatment compared to no-treatment in the first place?

Linda

 

[Morrigan]

fls, I don't see how I can make it any clearer than I have. Maybe someone else would like to try.

I don't know why people don't understand you, because I do. If we don't test with further controls, but only with True acupuncture and Sham acupuncture, and get similar results, we're at a dead-end.

And you have yet to show us why that third group is necessary when what we want to know is whether acupuncture works.

If both true and sham acupuncture show similar results, what is the conclusion? You still don't know if acupuncture works or not.

That true acupuncture is no more effective than sham acupuncture.

That may not tell you that it works, though. It can be that both are placebos, or that both are beneficial due to the needle sticking, or something else.

 

[fls]

I don't know why people don't understand you, because I do.

I have had no problem understanding Sherman Bay, and as far as I can tell, Billy Joe hasn't either.

If we don't test with further controls, but only with True acupuncture and Sham acupuncture, and get similar results, we're at a dead-end.

If your question is "is acupuncture effective?", you are now able to answer "no". How is that a dead end?

If both true and sham acupuncture show similar results, what is the conclusion?

That acupuncture does not have a specific treatment effect.

You still don't know if acupuncture works or not.

Huh? You just demonstrated that it didn't.

That may not tell you that it works, though. It can be that both are placebos, or that both are beneficial due to the needle sticking, or something else.

Yes, as has been mentioned several times now, while the issue of meridians and acupoints may be discredited (i.e. acupuncture), you may wonder whether the actual insertion of needles has a specific effect (i.e. needling). As I mentioned previously, this can be tested by using a device with a retractable needle and comparing a group where the needle was inserted with a group where the needle was retracted. Your choice of control group depends upon what specific question(s) you are asking. Can you propose a question specific to acupuncture that would be answered (in a non-trivial way) only by a no-treatment control group?

Linda

 

[Ivor the Engineer]

Why would you want to infer the likely benefit or harm of a treatment compared to no-treatment in the first place?

Linda

I'm not sure I understand the question, but I'd say so the absolute benefit (or harm) of the treatment can be estimated.

I think it also depends on the effect size and the predictability of the course of the disease being treated.

 

[fls]

I'm not sure I understand the question, but I'd say so the absolute benefit (or harm) of the treatment can be estimated.

I think it also depends on the effect size and the predictability of the course of the disease being treated.

If you think this would be useful, why do you think that it is rarely done?

Linda

 

[Ivor the Engineer]

If you think this would be useful, why do you think that it is rarely done?

Linda

Probably because of ethical and practical reasons. For example, if there is an existing treatment that is believed to be beneficial, it would be unethical to withhold it. Similarly, if the testing of an existing treatment is believed to have a pedigree which includes a test against a no-treatment group, then there is little point in having another no-treatment group.

Another practical reason which springs to mind is that it would probably be difficult to recruit and retain a non-treatment group if the study period was more than a day or so. It would also be difficult to ensure the no-treatment group did not secretly get access to the treatment under study, or seek treatment elsewhere.

 

[fls]

Probably because of ethical and practical reasons. For example, if there is an existing treatment that is believed to be beneficial, it would be unethical to withhold it.

I did a quick survey of the RCT's (for any intervention) published in the NEJM this year. About half had placebo-controlled groups, but none of those had no treatment groups.

Similarly, if the testing of an existing treatment is believed to have a pedigree which includes a test against a no-treatment group, then there is little point in having another no-treatment group.

Since no-treatment groups are rare, the vast majority of existing treatments will have no pedigree.

Another practical reason which springs to mind is that it would probably be difficult to recruit and retain a non-treatment group if the study period was more than a day or so.

I don't understand this.

It would also be difficult to ensure the no-treatment group did not secretly get access to the treatment under study, or seek treatment elsewhere.

That does actually happen, so it is one of those things we keep track of.

Linda